int main(int argc,
char **argv)
{
char filename[BUF_SIZE];
char *dirin, *rootname, *fname;
l_int32 i, firstpage, npages, nfiles;
l_float32 thresh, weight;
JBDATA *data;
JBCLASSER *classer;
SARRAY *safiles;
PIX *pix, *pixt;
PIXA *pixa, *pixadb;
static char mainName[] = "jbcorrelation";
if (argc != 5 && argc != 7)
return ERROR_INT(" Syntax: jbcorrelation dirin thresh weight "
"rootname [firstpage, npages]", mainName, 1);
dirin = argv[1];
thresh = atof(argv[2]);
weight = atof(argv[3]);
rootname = argv[4];
if (argc == 5) {
firstpage = 0;
npages = 0;
}
else {
firstpage = atoi(argv[5]);
npages = atoi(argv[6]);
}
#if 0
/*--------------------------------------------------------------*/
jbCorrelation(dirin, thresh, weight, COMPONENTS, rootname,
firstpage, npages, 1);
/*--------------------------------------------------------------*/
#else
/*--------------------------------------------------------------*/
safiles = getSortedPathnamesInDirectory(dirin, NULL, firstpage, npages);
nfiles = sarrayGetCount(safiles);
sarrayWriteStream(stderr, safiles);
/* Classify components on requested pages */
startTimer();
classer = jbCorrelationInit(COMPONENTS, 0, 0, thresh, weight);
jbAddPages(classer, safiles);
fprintf(stderr, "Time to generate classes: %6.3f sec\n", stopTimer());
/* Save and write out the result */
data = jbDataSave(classer);
jbDataWrite(rootname, data);
fprintf(stderr, "Number of classes: %d\n", classer->nclass);
/* Render the pages from the classifier data.
* Use debugflag == FALSE to omit outlines of each component. */
pixa = jbDataRender(data, FALSE);
/* Write the pages out */
npages = pixaGetCount(pixa);
if (npages != nfiles)
fprintf(stderr, "npages = %d, nfiles = %d, not equal!\n",
npages, nfiles);
for (i = 0; i < npages; i++) {
pix = pixaGetPix(pixa, i, L_CLONE);
snprintf(filename, BUF_SIZE, "%s.%05d", rootname, i);
fprintf(stderr, "filename: %s\n", filename);
pixWrite(filename, pix, IFF_PNG);
pixDestroy(&pix);
}
#if DISPLAY_DIFFERENCE
fname = sarrayGetString(safiles, 0, 0);
pixt = pixRead(fname);
pix = pixaGetPix(pixa, 0, L_CLONE);
pixXor(pixt, pixt, pix);
pixWrite("junk_output_diff", pixt, IFF_PNG);
pixDestroy(&pix);
pixDestroy(&pixt);
#endif /* DISPLAY_DIFFERENCE */
#if DEBUG_TEST_DATA_IO
{ JBDATA *newdata;
PIX *newpix;
PIXA *newpixa;
l_int32 same, iofail;
/* Read the data back in and render the pages */
newdata = jbDataRead(rootname);
newpixa = jbDataRender(newdata, FALSE);
iofail = FALSE;
for (i = 0; i < npages; i++) {
pix = pixaGetPix(pixa, i, L_CLONE);
newpix = pixaGetPix(newpixa, i, L_CLONE);
pixEqual(pix, newpix, &same);
if (!same) {
iofail = TRUE;
fprintf(stderr, "pix on page %d are unequal!\n", i);
}
pixDestroy(&pix);
pixDestroy(&newpix);
}
if (iofail)
fprintf(stderr, "read/write for jbdata fails\n");
else
fprintf(stderr, "read/write for jbdata succeeds\n");
jbDataDestroy(&newdata);
pixaDestroy(&newpixa);
}
#endif /* DEBUG_TEST_DATA_IO */
#if RENDER_DEBUG
/* Use debugflag == TRUE to see outlines of each component. */
pixadb = jbDataRender(data, TRUE);
/* Write the debug pages out */
npages = pixaGetCount(pixadb);
for (i = 0; i < npages; i++) {
pix = pixaGetPix(pixadb, i, L_CLONE);
snprintf(filename, BUF_SIZE, "%s.db.%05d", rootname, i);
fprintf(stderr, "filename: %s\n", filename);
pixWrite(filename, pix, IFF_PNG);
pixDestroy(&pix);
}
pixaDestroy(&pixadb);
#endif /* RENDER_DEBUG */
#if DISPLAY_ALL_INSTANCES
/* display all instances, organized by template */
pix = pixaaDisplayByPixa(classer->pixaa,
X_SPACING, Y_SPACING, MAX_OUTPUT_WIDTH);
pixWrite("output_instances", pix, IFF_PNG);
pixDestroy(&pix);
#endif /* DISPLAY_ALL_INSTANCES */
pixaDestroy(&pixa);
sarrayDestroy(&safiles);
jbClasserDestroy(&classer);
jbDataDestroy(&data);
/*--------------------------------------------------------------*/
#endif
return 0;
}
/*
* pixaWriteCompressedToPS()
*
* Input: pixa (any set of images)
* fileout (output ps file)
* res (of input image)
* level (compression: 2 or 3)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) This generates a PS file of multiple page images, all
* with bounding boxes.
* (2) It compresses to:
* cmap + level2: jpeg
* cmap + level3: flate
* 1 bpp: tiffg4
* 2 or 4 bpp + level2: jpeg
* 2 or 4 bpp + level3: flate
* 8 bpp: jpeg
* 16 bpp: flate
* 32 bpp: jpeg
* (3) To generate a pdf, use: ps2pdf <infile.ps> <outfile.pdf>
*/
l_int32
pixaWriteCompressedToPS(PIXA *pixa,
const char *fileout,
l_int32 res,
l_int32 level)
{
char *tname, *g4_name, *jpeg_name, *png_name;
l_int32 i, n, firstfile, index, writeout, d;
PIX *pix, *pixt;
PIXCMAP *cmap;
PROCNAME("pixaWriteCompressedToPS");
if (!pixa)
return ERROR_INT("pixa not defined", procName, 1);
if (!fileout)
return ERROR_INT("fileout not defined", procName, 1);
if (level != 2 && level != 3) {
L_ERROR("only levels 2 and 3 permitted; using level 2\n", procName);
level = 2;
}
n = pixaGetCount(pixa);
firstfile = TRUE;
index = 0;
lept_mkdir("lept/comp");
g4_name = genTempFilename("/tmp/lept/comp", "temp.tif", 0, 0);
jpeg_name = genTempFilename("/tmp/lept/comp", "temp.jpg", 0, 0);
png_name = genTempFilename("/tmp/lept/comp", "temp.png", 0, 0);
for (i = 0; i < n; i++) {
writeout = TRUE;
pix = pixaGetPix(pixa, i, L_CLONE);
d = pixGetDepth(pix);
cmap = pixGetColormap(pix);
if (d == 1) {
tname = g4_name;
pixWrite(tname, pix, IFF_TIFF_G4);
} else if (cmap) {
if (level == 2) {
pixt = pixConvertForPSWrap(pix);
tname = jpeg_name;
pixWrite(tname, pixt, IFF_JFIF_JPEG);
pixDestroy(&pixt);
} else { /* level == 3 */
tname = png_name;
pixWrite(tname, pix, IFF_PNG);
}
} else if (d == 16) {
if (level == 2)
L_WARNING("d = 16; must write out flate\n", procName);
tname = png_name;
pixWrite(tname, pix, IFF_PNG);
} else if (d == 2 || d == 4) {
if (level == 2) {
pixt = pixConvertTo8(pix, 0);
tname = jpeg_name;
pixWrite(tname, pixt, IFF_JFIF_JPEG);
pixDestroy(&pixt);
} else { /* level == 3 */
tname = png_name;
pixWrite(tname, pix, IFF_PNG);
}
} else if (d == 8 || d == 32) {
tname = jpeg_name;
pixWrite(tname, pix, IFF_JFIF_JPEG);
} else { /* shouldn't happen */
L_ERROR("invalid depth: %d\n", procName, d);
writeout = FALSE;
}
pixDestroy(&pix);
if (writeout)
writeImageCompressedToPSFile(tname, fileout, res,
&firstfile, &index);
}
LEPT_FREE(g4_name);
LEPT_FREE(jpeg_name);
LEPT_FREE(png_name);
return 0;
}
l_int32
DoComparisonDwa1(PIX *pixs,
PIX *pixt1,
PIX *pixt2,
PIX *pixt3,
PIX *pixt4,
PIX *pixt5,
PIX *pixt6,
l_int32 isize)
{
l_int32 fact1, fact2, size;
selectComposableSizes(isize, &fact1, &fact2);
size = fact1 * fact2;
fprintf(stderr, "..%d..", size);
if (TIMING) startTimer();
pixDilateCompBrickExtendDwa(pixt1, pixs, size, 1);
pixDilateCompBrickExtendDwa(pixt3, pixs, 1, size);
pixDilateCompBrickExtendDwa(pixt5, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Dwa: %7.3f sec\n", stopTimer());
if (TIMING) startTimer();
pixDilateCompBrick(pixt2, pixs, size, 1);
pixDilateCompBrick(pixt4, pixs, 1, size);
pixDilateCompBrick(pixt6, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Rop: %7.3f sec\n", stopTimer());
PixCompareDwa(size, "dilate", pixt1, pixt2, pixt3, pixt4, pixt5, pixt6);
if (TIMING) startTimer();
pixErodeCompBrickExtendDwa(pixt1, pixs, size, 1);
pixErodeCompBrickExtendDwa(pixt3, pixs, 1, size);
pixErodeCompBrickExtendDwa(pixt5, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Dwa: %7.3f sec\n", stopTimer());
if (TIMING) startTimer();
pixErodeCompBrick(pixt2, pixs, size, 1);
pixErodeCompBrick(pixt4, pixs, 1, size);
pixErodeCompBrick(pixt6, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Rop: %7.3f sec\n", stopTimer());
PixCompareDwa(size, "erode", pixt1, pixt2, pixt3, pixt4, pixt5, pixt6);
if (TIMING) startTimer();
pixOpenCompBrickExtendDwa(pixt1, pixs, size, 1);
pixOpenCompBrickExtendDwa(pixt3, pixs, 1, size);
pixOpenCompBrickExtendDwa(pixt5, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Dwa: %7.3f sec\n", stopTimer());
if (TIMING) startTimer();
pixOpenCompBrick(pixt2, pixs, size, 1);
pixOpenCompBrick(pixt4, pixs, 1, size);
pixOpenCompBrick(pixt6, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Rop: %7.3f sec\n", stopTimer());
PixCompareDwa(size, "open", pixt1, pixt2, pixt3, pixt4, pixt5, pixt6);
if (TIMING) startTimer();
pixCloseCompBrickExtendDwa(pixt1, pixs, size, 1);
pixCloseCompBrickExtendDwa(pixt3, pixs, 1, size);
pixCloseCompBrickExtendDwa(pixt5, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Dwa: %7.3f sec\n", stopTimer());
if (TIMING) startTimer();
pixCloseSafeCompBrick(pixt2, pixs, size, 1);
pixCloseSafeCompBrick(pixt4, pixs, 1, size);
pixCloseSafeCompBrick(pixt6, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Rop: %7.3f sec\n", stopTimer());
PixCompareDwa(size, "close", pixt1, pixt2, pixt3, pixt4, pixt5, pixt6);
#if 0
pixWrite("/tmp/junkpixt3.png", pixt3, IFF_PNG);
pixWrite("/tmp/junkpixt4.png", pixt4, IFF_PNG);
pixXor(pixt3, pixt3, pixt4);
pixWrite("/tmp/junkxor.png", pixt3, IFF_PNG);
#endif
return 0;
}
int main(int argc,
char **argv) {
char buf[32];
char *filein, *fileout, *fontdir, *textstr;
l_int32 n, i, maxdepth, ntext, border, lossless, display, showtext;
l_float32 scalefact;
L_BMF *bmf;
PIX *pix1, *pix2, *pix3, *pix4, *pixd;
PIXA *pixa, *pixad;
static char mainName[] = "displaypixa";
if (argc != 3 && argc != 4 && argc != 7 && argc != 8) {
fprintf(stderr, "Syntax error in displaypixa:\n"
" displaypixa filein fileout [showtext]\n"
" displaypixa filein scalefact border"
" lossless disp fileout [showtext]\n");
return 1;
}
filein = argv[1];
if ((pixa = pixaRead(filein)) == NULL)
return ERROR_INT("pixa not made", mainName, 1);
pixaCountText(pixa, &ntext);
if (argc == 3 || argc == 4)
fileout = argv[2];
if (argc == 4)
showtext = atoi(argv[3]);
/* Simple specification; no output text */
if (argc == 3 ||
(argc == 4 && (ntext == 0 || showtext == 0))) { /* no text output */
pixaVerifyDepth(pixa, &maxdepth);
pixd = pixaDisplayTiledInRows(pixa, maxdepth, 1400, 1.0, 0, 10, 0);
pixDisplay(pixd, 100, 100);
if (pixGetDepth(pixd) == 1)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
pixDestroy(&pixd);
pixaDestroy(&pixa);
return 0;
}
/* Simple specification with output text */
if (argc == 4) { /* showtext == 1 && ntext > 0 */
n = pixaGetCount(pixa);
bmf = bmfCreate(NULL, 6);
pixad = pixaCreate(n);
for (i = 0; i < n; i++) {
pix1 = pixaGetPix(pixa, i, L_CLONE);
pix2 = pixConvertTo32(pix1);
pix3 = pixAddBorderGeneral(pix2, 10, 10, 5, 5, 0xffffff00);
textstr = pixGetText(pix1);
if (textstr && strlen(textstr) > 0) {
snprintf(buf, sizeof(buf), "%s", textstr);
pix4 = pixAddSingleTextblock(pix3, bmf, buf, 0xff000000,
L_ADD_BELOW, NULL);
} else {
pix4 = pixClone(pix3);
}
pixaAddPix(pixad, pix4, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
}
bmfDestroy(&bmf);
pixaVerifyDepth(pixad, &maxdepth);
pixd = pixaDisplayTiledInRows(pixad, maxdepth, 1400, 1.0, 0, 10, 0);
pixDisplay(pixd, 100, 100);
if (pixGetDepth(pixd) == 1)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
pixDestroy(&pixd);
pixaDestroy(&pixa);
pixaDestroy(&pixad);
return 0;
}
/* Full specification */
scalefact = atof(argv[2]);
border = atoi(argv[3]);
lossless = atoi(argv[4]);
display = atoi(argv[5]);
fileout = argv[6];
showtext = (argc == 8) ? atoi(argv[7]) : 0;
if (showtext && ntext == 0)
L_INFO("No text found in any of the pix\n", mainName);
bmf = (showtext && ntext > 0) ? bmfCreate(NULL, 6) : NULL;
n = pixaGetCount(pixa);
pixad = pixaCreate(n);
for (i = 0; i < n; i++) {
pix1 = pixaGetPix(pixa, i, L_CLONE);
pix2 = pixConvertTo32(pix1);
pix3 = pixAddBorderGeneral(pix2, 10, 10, 5, 5, 0xffffff00);
textstr = pixGetText(pix1);
if (bmf && textstr && strlen(textstr) > 0) {
snprintf(buf, sizeof(buf), "%s", textstr);
pix4 = pixAddSingleTextblock(pix3, bmf, buf, 0xff000000,
L_ADD_BELOW, NULL);
} else {
pix4 = pixClone(pix3);
}
pixaAddPix(pixad, pix4, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
}
bmfDestroy(&bmf);
pixaVerifyDepth(pixad, &maxdepth);
pixd = pixaDisplayTiledInRows(pixad, maxdepth, 1400, scalefact,
0, 10, border);
if (display) pixDisplay(pixd, 20, 20);
if (pixGetDepth(pixd) == 1 || lossless)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
pixDestroy(&pixd);
pixaDestroy(&pixa);
pixaDestroy(&pixad);
return 0;
}
main(int argc,
char **argv)
{
l_int32 d;
PIX *pixs, *pixc, *pixr, *pixg, *pixb, *pixsg, *pixsm, *pixd;
PIXA *pixa;
static char mainName[] = "livre_adapt";
if (argc != 1)
exit(ERROR_INT(" Syntax: livre_adapt", mainName, 1));
/* Read the image in at 150 ppi. */
pixDisplayWrite(NULL, -1);
if ((pixs = pixRead("brothers.150.jpg")) == NULL)
exit(ERROR_INT("pix not made", mainName, 1));
pixDisplayWriteFormat(pixs, 2, IFF_JFIF_JPEG);
/* Normalize for uneven illumination on RGB image */
pixBackgroundNormRGBArraysMorph(pixs, NULL, 4, 5, 200,
&pixr, &pixg, &pixb);
pixd = pixApplyInvBackgroundRGBMap(pixs, pixr, pixg, pixb, 4, 4);
pixDisplayWriteFormat(pixd, 2, IFF_JFIF_JPEG);
pixDestroy(&pixr);
pixDestroy(&pixg);
pixDestroy(&pixb);
pixDestroy(&pixd);
/* Convert the RGB image to grayscale. */
pixsg = pixConvertRGBToLuminance(pixs);
pixDisplayWriteFormat(pixsg, 2, IFF_JFIF_JPEG);
/* Remove the text in the fg. */
pixc = pixCloseGray(pixsg, 25, 25);
pixDisplayWriteFormat(pixc, 2, IFF_JFIF_JPEG);
/* Smooth the bg with a convolution. */
pixsm = pixBlockconv(pixc, 15, 15);
pixDisplayWriteFormat(pixsm, 2, IFF_JFIF_JPEG);
pixDestroy(&pixc);
/* Normalize for uneven illumination on gray image. */
pixBackgroundNormGrayArrayMorph(pixsg, NULL, 4, 5, 200, &pixg);
pixc = pixApplyInvBackgroundGrayMap(pixsg, pixg, 4, 4);
pixDisplayWriteFormat(pixc, 2, IFF_JFIF_JPEG);
pixDestroy(&pixg);
/* Increase the dynamic range. */
pixd = pixGammaTRC(NULL, pixc, 1.0, 30, 180);
pixDisplayWriteFormat(pixd, 2, IFF_JFIF_JPEG);
pixDestroy(&pixc);
/* Threshold to 1 bpp. */
pixb = pixThresholdToBinary(pixd, 120);
pixDisplayWriteFormat(pixb, 2, IFF_PNG);
pixDestroy(&pixd);
pixDestroy(&pixb);
/* Generate the output image */
pixa = pixaReadFiles("/tmp", "junk_write_display");
pixd = pixaDisplayTiledAndScaled(pixa, 8, 350, 4, 0, 25, 2);
pixWrite("/tmp/adapt.jpg", pixd, IFF_JFIF_JPEG);
pixDisplayWithTitle(pixd, 100, 100, NULL, 1);
pixDestroy(&pixd);
pixDestroy(&pixs);
pixDestroy(&pixsg);
return 0;
}
/*!
* pixConnCompBB()
*
* Input: pixs (1 bpp)
* connectivity (4 or 8)
* Return: boxa, or null on error
*
* Notes:
* (1) Finds bounding boxes of 4- or 8-connected components
* in a binary image.
* (2) This works on a copy of the input pix. The c.c. are located
* in raster order and erased one at a time. In the process,
* the b.b. is computed and saved.
*/
BOXA *
pixConnCompBB(PIX *pixs,
l_int32 connectivity)
{
l_int32 h, iszero;
l_int32 x, y, xstart, ystart;
PIX *pixt;
BOX *box;
BOXA *boxa;
L_STACK *stack, *auxstack;
PROCNAME("pixConnCompBB");
if (!pixs || pixGetDepth(pixs) != 1)
return (BOXA *)ERROR_PTR("pixs undefined or not 1 bpp", procName, NULL);
if (connectivity != 4 && connectivity != 8)
return (BOXA *)ERROR_PTR("connectivity not 4 or 8", procName, NULL);
pixZero(pixs, &iszero);
if (iszero)
return boxaCreate(1); /* return empty boxa */
if ((pixt = pixCopy(NULL, pixs)) == NULL)
return (BOXA *)ERROR_PTR("pixt not made", procName, NULL);
h = pixGetHeight(pixs);
if ((stack = lstackCreate(h)) == NULL)
return (BOXA *)ERROR_PTR("stack not made", procName, NULL);
if ((auxstack = lstackCreate(0)) == NULL)
return (BOXA *)ERROR_PTR("auxstack not made", procName, NULL);
stack->auxstack = auxstack;
if ((boxa = boxaCreate(0)) == NULL)
return (BOXA *)ERROR_PTR("boxa not made", procName, NULL);
xstart = 0;
ystart = 0;
while (1)
{
if (!nextOnPixelInRaster(pixt, xstart, ystart, &x, &y))
break;
if ((box = pixSeedfillBB(pixt, stack, x, y, connectivity)) == NULL)
return (BOXA *)ERROR_PTR("box not made", procName, NULL);
boxaAddBox(boxa, box, L_INSERT);
xstart = x;
ystart = y;
}
#if DEBUG
pixCountPixels(pixt, &iszero, NULL);
fprintf(stderr, "Number of remaining pixels = %d\n", iszero);
pixWrite("junkremain", pixt1, IFF_PNG);
#endif /* DEBUG */
/* Cleanup, freeing the fillsegs on each stack */
lstackDestroy(&stack, TRUE);
pixDestroy(&pixt);
return boxa;
}
int main(int argc,
char **argv)
{
l_int32 i, n, w, h, nactual, imax;
BOX *box;
BOXA *boxa;
PIX *pixs, *pixd, *pix;
PIXA *pixas, *pixat, *pixac;
L_PTRA *papix, *pabox, *papix2, *pabox2;
static char mainName[] = "ptra1_reg";
if (argc != 1)
return ERROR_INT(" Syntax: ptra1_reg", mainName, 1);
setLeptDebugOK(1);
pixac = pixaCreate(0);
if ((pixs = pixRead("lucasta.1.300.tif")) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
pixGetDimensions(pixs, &w, &h, NULL);
boxa = pixConnComp(pixs, &pixas, 8);
pixDestroy(&pixs);
boxaDestroy(&boxa);
n = pixaGetCount(pixas);
/* Fill ptras with clones and reconstruct */
fprintf(stderr, "Fill with clones and reconstruct\n");
MakePtrasFromPixa(pixas, &papix, &pabox, L_CLONE);
pixat = ReconstructPixa(papix, pabox, CHOOSE_RECON);
ptraDestroy(&papix, 0, 1);
ptraDestroy(&pabox, 0, 1);
DisplayResult(pixac, &pixat, w, h, 1);
/* Remove every other one for the first half;
* with compaction at each removal */
fprintf(stderr, "Remove every other in 1st half, with compaction\n");
MakePtrasFromPixa(pixas, &papix, &pabox, L_COPY);
for (i = 0; i < n / 2; i++) {
if (i % 2 == 0) {
pix = (PIX *)ptraRemove(papix, i, L_COMPACTION);
box = (BOX *)ptraRemove(pabox, i, L_COMPACTION);
pixDestroy(&pix);
boxDestroy(&box);
}
}
pixat = ReconstructPixa(papix, pabox, CHOOSE_RECON);
ptraDestroy(&papix, 0, 1);
ptraDestroy(&pabox, 0, 1);
DisplayResult(pixac, &pixat, w, h, 0);
/* Remove every other one for the entire set,
* but without compaction at each removal */
fprintf(stderr,
"Remove every other in 1st half, without & then with compaction\n");
MakePtrasFromPixa(pixas, &papix, &pabox, L_COPY);
for (i = 0; i < n; i++) {
if (i % 2 == 0) {
pix = (PIX *)ptraRemove(papix, i, L_NO_COMPACTION);
box = (BOX *)ptraRemove(pabox, i, L_NO_COMPACTION);
pixDestroy(&pix);
boxDestroy(&box);
}
}
ptraCompactArray(papix); /* now do the compaction */
ptraCompactArray(pabox);
pixat = ReconstructPixa(papix, pabox, CHOOSE_RECON);
ptraDestroy(&papix, 0, 1);
ptraDestroy(&pabox, 0, 1);
DisplayResult(pixac, &pixat, w, h, 0);
/* Fill ptras using insert at head, and reconstruct */
fprintf(stderr, "Insert at head and reconstruct\n");
papix = ptraCreate(n);
pabox = ptraCreate(n);
for (i = 0; i < n; i++) {
pix = pixaGetPix(pixas, i, L_CLONE);
box = pixaGetBox(pixas, i, L_CLONE);
ptraInsert(papix, 0, pix, L_MIN_DOWNSHIFT);
ptraInsert(pabox, 0, box, L_FULL_DOWNSHIFT);
}
pixat = ReconstructPixa(papix, pabox, CHOOSE_RECON);
ptraDestroy(&papix, 0, 1);
ptraDestroy(&pabox, 0, 1);
DisplayResult(pixac, &pixat, w, h, 1);
/* Reverse the arrays by swapping */
fprintf(stderr, "Reverse by swapping\n");
MakePtrasFromPixa(pixas, &papix, &pabox, L_CLONE);
for (i = 0; i < n / 2; i++) {
ptraSwap(papix, i, n - i - 1);
ptraSwap(pabox, i, n - i - 1);
}
ptraCompactArray(papix); /* already compact; shouldn't do anything */
ptraCompactArray(pabox);
pixat = ReconstructPixa(papix, pabox, CHOOSE_RECON);
ptraDestroy(&papix, 0, 1);
ptraDestroy(&pabox, 0, 1);
DisplayResult(pixac, &pixat, w, h, 0);
/* Remove at the top of the array and push the hole to the end
* by neighbor swapping (!). This is O(n^2), so it's not a
* recommended way to copy a ptra. [joke] */
fprintf(stderr,
"Remove at top, pushing hole to end by swapping -- O(n^2)\n");
MakePtrasFromPixa(pixas, &papix, &pabox, L_CLONE);
papix2 = ptraCreate(0);
pabox2 = ptraCreate(0);
while (1) {
ptraGetActualCount(papix, &nactual);
if (nactual == 0) break;
ptraGetMaxIndex(papix, &imax);
pix = (PIX *)ptraRemove(papix, 0, L_NO_COMPACTION);
box = (BOX *)ptraRemove(pabox, 0, L_NO_COMPACTION);
ptraAdd(papix2, pix);
ptraAdd(pabox2, box);
for (i = 1; i <= imax; i++) {
ptraSwap(papix, i - 1, i);
ptraSwap(pabox, i - 1, i);
}
}
ptraCompactArray(papix); /* should be empty */
ptraCompactArray(pabox); /* ditto */
pixat = ReconstructPixa(papix, pabox, CHOOSE_RECON);
ptraDestroy(&papix, 0, 1);
ptraDestroy(&pabox, 0, 1);
DisplayResult(pixac, &pixat, w, h, 1); /* nothing there */
pixat = ReconstructPixa(papix2, pabox2, CHOOSE_RECON);
ptraDestroy(&papix2, 0, 1);
ptraDestroy(&pabox2, 0, 1);
DisplayResult(pixac, &pixat, w, h, 0);
/* Remove and insert one position above, allowing minimum downshift.
* If you specify L_AUTO_DOWNSHIFT, because there is only 1 hole,
* it will do a full downshift at each insert. This is a
* situation where the heuristic (expected number of holes)
* fails to do the optimal thing. */
fprintf(stderr, "Remove and insert one position above (min downshift)\n");
MakePtrasFromPixa(pixas, &papix, &pabox, L_CLONE);
for (i = 1; i < n; i++) {
pix = (PIX *)ptraRemove(papix, i, L_NO_COMPACTION);
box = (BOX *)ptraRemove(pabox, i, L_NO_COMPACTION);
ptraInsert(papix, i - 1, pix, L_MIN_DOWNSHIFT);
ptraInsert(pabox, i - 1, box, L_MIN_DOWNSHIFT);
}
pixat = ReconstructPixa(papix, pabox, CHOOSE_RECON);
ptraDestroy(&papix, 0, 1);
ptraDestroy(&pabox, 0, 1);
DisplayResult(pixac, &pixat, w, h, 1);
/* Remove and insert one position above, but this time
* forcing a full downshift at each step. */
fprintf(stderr, "Remove and insert one position above (full downshift)\n");
MakePtrasFromPixa(pixas, &papix, &pabox, L_CLONE);
for (i = 1; i < n; i++) {
pix = (PIX *)ptraRemove(papix, i, L_NO_COMPACTION);
box = (BOX *)ptraRemove(pabox, i, L_NO_COMPACTION);
ptraInsert(papix, i - 1, pix, L_AUTO_DOWNSHIFT);
ptraInsert(pabox, i - 1, box, L_AUTO_DOWNSHIFT);
}
/* ptraCompactArray(papix);
ptraCompactArray(pabox); */
pixat = ReconstructPixa(papix, pabox, CHOOSE_RECON);
ptraDestroy(&papix, 0, 1);
ptraDestroy(&pabox, 0, 1);
DisplayResult(pixac, &pixat, w, h, 0);
pixd = pixaDisplay(pixac, 0, 0);
pixDisplay(pixd, 100, 100);
pixWrite("/tmp/junkptra1.png", pixd, IFF_PNG);
pixDestroy(&pixd);
pixaDestroy(&pixac);
pixaDestroy(&pixas);
return 0;
}
main(int argc,
char **argv)
{
char *filein, *fileout;
char buffer[512];
const char *psfile = "/tmp/junk_split_image.ps";
l_int32 nx, ny, i, w, h, d, ws, hs, n, res, ignore;
l_float32 scale;
PIX *pixs, *pixt, *pixr;
PIXA *pixa;
static char mainName[] = "splitimage2pdf";
if (argc != 5)
return ERROR_INT(" Syntax: splitimage2pdf filein nx ny fileout",
mainName, 1);
filein = argv[1];
nx = atoi(argv[2]);
ny = atoi(argv[3]);
fileout = argv[4];
lept_rm(NULL, "junk_split_image.ps");
if ((pixs = pixRead(filein)) == NULL)
exit(ERROR_INT("pixs not made", mainName, 1));
d = pixGetDepth(pixs);
if (d == 1 )
lept_rm(NULL, "junk_split_image.tif");
else if (d == 8 || d == 32)
lept_rm(NULL, "junk_split_image.jpg");
else
return ERROR_INT("d not in {1,8,32} bpp", mainName, 1);
ws = pixGetWidth(pixs);
hs = pixGetHeight(pixs);
if (ny * ws > nx * hs)
pixr = pixRotate90(pixs, 1);
else
pixr = pixClone(pixs);
pixa = pixaSplitPix(pixr, nx, ny, 0, 0);
n = pixaGetCount(pixa);
res = 300;
for (i = 0; i < n; i++) {
pixt = pixaGetPix(pixa, i, L_CLONE);
w = pixGetWidth(pixt);
h = pixGetHeight(pixt);
scale = L_MIN(FILL_FACTOR * 2550 / w, FILL_FACTOR * 3300 / h);
if (d == 1) {
pixWrite("/tmp/junk_split_image.tif", pixt, IFF_TIFF_G4);
if (i == 0)
convertG4ToPS("/tmp/junk_split_image.tif", psfile,
"w", 0, 0, 300, scale, 1, FALSE, TRUE);
else
convertG4ToPS("/tmp/junk_split_image.tif", psfile,
"a", 0, 0, 300, scale, 1, FALSE, TRUE);
}
else {
pixWrite("/tmp/junk_split_image.jpg", pixt, IFF_JFIF_JPEG);
if (i == 0)
convertJpegToPS("/tmp/junk_split_image.jpg", psfile,
"w", 0, 0, 300, scale, 1, TRUE);
else
convertJpegToPS("/tmp/junk_split_image.jpg", psfile,
"a", 0, 0, 300, scale, 1, TRUE);
}
pixDestroy(&pixt);
}
sprintf(buffer, "ps2pdf %s %s", psfile, fileout);
ignore = system(buffer);
pixaDestroy(&pixa);
pixDestroy(&pixr);
pixDestroy(&pixs);
return 0;
}
int main_find_pattern(int argc,
char **argv)
{
char *filein, *fileout, *patternfile;
l_int32 w, h, i, n;
BOX *box, *boxe;
BOXA *boxa1, *boxa2;
PIX *pixs, *pixp, *pixpe;
PIX *pixd, *pixt1, *pixt2, *pixhmt;
SEL *sel_2h, *sel;
static char mainName[] = "findpattern1";
filein = "feyn.tif";
patternfile = "char.tif";
fileout = "result.findpattern1";
if ((pixs = pixRead(filein)) == NULL)
printf("pixs not made\n");
if ((pixp = pixRead(patternfile)) == NULL)
printf("pixp not made\n");
w = pixGetWidth(pixp);
h = pixGetHeight(pixp);
/* generate the hit-miss Sel with runs */
sel = pixGenerateSelWithRuns(pixp, NumHorLines, NumVertLines, 0,
MinRunlength, 7, 7, 0, 0, &pixpe);
/* display the Sel two ways */
selWriteStream(stderr, sel);
pixt1 = pixDisplayHitMissSel(pixpe, sel, 9, HitColor, MissColor);
pixDisplay(pixt1, 200, 200);
pixWrite("junkpixt", pixt1, IFF_PNG);
/* use the Sel to find all instances in the page */
startTimer();
pixhmt = pixHMT(NULL, pixs, sel);
fprintf(stderr, "Time to find patterns = %7.3f\n", stopTimer());
/* small erosion to remove noise; typically not necessary if
* there are enough elements in the Sel */
sel_2h = selCreateBrick(1, 2, 0, 0, SEL_HIT);
pixt2 = pixErode(NULL, pixhmt, sel_2h);
/* display the result visually by placing the Sel at each
* location found */
pixd = pixDilate(NULL, pixt2, sel);
pixWrite(fileout, pixd, IFF_TIFF_G4);
/* display outut with an outline around each located pattern */
boxa1 = pixConnCompBB(pixt2, 8);
n = boxaGetCount(boxa1);
boxa2 = boxaCreate(n);
for (i = 0; i < n; i++) {
box = boxaGetBox(boxa1, i, L_COPY);
boxe = boxCreate(box->x - w / 2, box->y - h / 2, w + 4, h + 4);
boxaAddBox(boxa2, boxe, L_INSERT);
pixRenderBox(pixs, boxe, 4, L_FLIP_PIXELS);
boxDestroy(&box);
}
pixWrite("junkoutline", pixs, IFF_TIFF_G4);
//boxaWriteStream(stderr, boxa2); //TODO ???
pixDestroy(&pixs);
pixDestroy(&pixp);
pixDestroy(&pixpe);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixhmt);
pixDestroy(&pixd);
selDestroy(&sel);
selDestroy(&sel_2h);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
printf("\n---\nEND\n");
getchar();
return 0;
}
// This method dumps a debug image to the specified location.
void ShiroRekhaSplitter::DumpDebugImage(const char* filename) const {
pixWrite(filename, debug_image_, IFF_PNG);
}
l_int32 main(int argc,
char **argv)
{
l_int32 i, n;
l_float32 a, b, c, d, e;
NUMA *nax, *nafit;
PIX *pixs, *pixn, *pixg, *pixb, *pixt1, *pixt2;
PIXA *pixa;
PTA *pta, *ptad;
PTAA *ptaa1, *ptaa2;
pixs = pixRead("cat-35.jpg");
/* Normalize for varying background and binarize */
pixn = pixBackgroundNormSimple(pixs, NULL, NULL);
pixg = pixConvertRGBToGray(pixn, 0.5, 0.3, 0.2);
pixb = pixThresholdToBinary(pixg, 130);
pixDestroy(&pixn);
pixDestroy(&pixg);
/* Get the textline centers */
pixa = pixaCreate(6);
ptaa1 = dewarpGetTextlineCenters(pixb, 0);
pixt1 = pixCreateTemplate(pixs);
pixSetAll(pixt1);
pixt2 = pixDisplayPtaa(pixt1, ptaa1);
pixWrite("/tmp/textline1.png", pixt2, IFF_PNG);
pixDisplayWithTitle(pixt2, 0, 100, "textline centers 1", 1);
pixaAddPix(pixa, pixt2, L_INSERT);
pixDestroy(&pixt1);
/* Remove short lines */
fprintf(stderr, "Num all lines = %d\n", ptaaGetCount(ptaa1));
ptaa2 = dewarpRemoveShortLines(pixb, ptaa1, 0.8, 0);
pixt1 = pixCreateTemplate(pixs);
pixSetAll(pixt1);
pixt2 = pixDisplayPtaa(pixt1, ptaa2);
pixWrite("/tmp/textline2.png", pixt2, IFF_PNG);
pixDisplayWithTitle(pixt2, 300, 100, "textline centers 2", 1);
pixaAddPix(pixa, pixt2, L_INSERT);
pixDestroy(&pixt1);
n = ptaaGetCount(ptaa2);
fprintf(stderr, "Num long lines = %d\n", n);
ptaaDestroy(&ptaa1);
pixDestroy(&pixb);
/* Long lines over input image */
pixt1 = pixCopy(NULL, pixs);
pixt2 = pixDisplayPtaa(pixt1, ptaa2);
pixWrite("/tmp/textline3.png", pixt2, IFF_PNG);
pixDisplayWithTitle(pixt2, 600, 100, "textline centers 3", 1);
pixaAddPix(pixa, pixt2, L_INSERT);
pixDestroy(&pixt1);
/* Quadratic fit to curve */
pixt1 = pixCopy(NULL, pixs);
for (i = 0; i < n; i++) {
pta = ptaaGetPta(ptaa2, i, L_CLONE);
ptaGetArrays(pta, &nax, NULL);
ptaGetQuadraticLSF(pta, &a, &b, &c, &nafit);
fprintf(stderr, "Quadratic: a = %10.6f, b = %7.3f, c = %7.3f\n",
a, b, c);
ptad = ptaCreateFromNuma(nax, nafit);
pixDisplayPta(pixt1, pixt1, ptad);
ptaDestroy(&pta);
ptaDestroy(&ptad);
numaDestroy(&nax);
numaDestroy(&nafit);
}
pixWrite("/tmp/textline4.png", pixt1, IFF_PNG);
pixDisplayWithTitle(pixt1, 900, 100, "textline centers 4", 1);
pixaAddPix(pixa, pixt1, L_INSERT);
/* Cubic fit to curve */
pixt1 = pixCopy(NULL, pixs);
for (i = 0; i < n; i++) {
pta = ptaaGetPta(ptaa2, i, L_CLONE);
ptaGetArrays(pta, &nax, NULL);
ptaGetCubicLSF(pta, &a, &b, &c, &d, &nafit);
fprintf(stderr, "Cubic: a = %10.6f, b = %10.6f, c = %7.3f, d = %7.3f\n",
a, b, c, d);
ptad = ptaCreateFromNuma(nax, nafit);
pixDisplayPta(pixt1, pixt1, ptad);
ptaDestroy(&pta);
ptaDestroy(&ptad);
numaDestroy(&nax);
numaDestroy(&nafit);
}
pixWrite("/tmp/textline5.png", pixt1, IFF_PNG);
pixDisplayWithTitle(pixt1, 1200, 100, "textline centers 5", 1);
pixaAddPix(pixa, pixt1, L_INSERT);
/* Quartic fit to curve */
pixt1 = pixCopy(NULL, pixs);
for (i = 0; i < n; i++) {
pta = ptaaGetPta(ptaa2, i, L_CLONE);
ptaGetArrays(pta, &nax, NULL);
ptaGetQuarticLSF(pta, &a, &b, &c, &d, &e, &nafit);
fprintf(stderr,
"Quartic: a = %7.3f, b = %7.3f, c = %9.5f, d = %7.3f, e = %7.3f\n",
a, b, c, d, e);
ptad = ptaCreateFromNuma(nax, nafit);
pixDisplayPta(pixt1, pixt1, ptad);
ptaDestroy(&pta);
ptaDestroy(&ptad);
numaDestroy(&nax);
numaDestroy(&nafit);
}
pixWrite("/tmp/textline6.png", pixt1, IFF_PNG);
pixDisplayWithTitle(pixt1, 1500, 100, "textline centers 6", 1);
pixaAddPix(pixa, pixt1, L_INSERT);
pixaConvertToPdf(pixa, 300, 0.5, L_JPEG_ENCODE, 75,
"LS fittings to textlines", "/tmp/dewarp_fittings.pdf");
pixaDestroy(&pixa);
pixDestroy(&pixs);
ptaaDestroy(&ptaa2);
return 0;
}
main(int argc,
char **argv)
{
l_int32 i, n, ws, hs, w, h, rval, gval, bval, order;
l_float32 *mat1, *mat2, *mat3;
l_float32 matd[9];
BOX *box, *boxt;
BOXA *boxa, *boxat, *boxa1, *boxa2, *boxa3, *boxa4, *boxa5;
PIX *pix, *pixs, *pixb, *pixc, *pixt, *pixt1, *pixt2, *pixt3;
PIXA *pixa;
static char mainName[] = "xformbox_reg";
/* ----------------------------------------------------------- *
* Test hash rendering in 3 modes *
* ----------------------------------------------------------- */
pixs = pixRead("feyn.tif");
box = boxCreate(461, 429, 1393, 342);
pixt1 = pixClipRectangle(pixs, box, NULL);
boxa = pixConnComp(pixt1, NULL, 8);
n = boxaGetCount(boxa);
pixt2 = pixConvertTo8(pixt1, 1);
pixt3 = pixConvertTo32(pixt1);
for (i = 0; i < n; i++) {
boxt = boxaGetBox(boxa, i, L_CLONE);
rval = (1413 * i) % 256;
gval = (4917 * i) % 256;
bval = (7341 * i) % 256;
pixRenderHashBox(pixt1, boxt, 8, 2, i % 4, 1, L_SET_PIXELS);
pixRenderHashBoxArb(pixt2, boxt, 7, 2, i % 4, 1, rval, gval, bval);
pixRenderHashBoxBlend(pixt3, boxt, 7, 2, i % 4, 1, rval, gval, bval,
0.5);
boxDestroy(&boxt);
}
pixDisplay(pixt1, 0, 0);
pixDisplay(pixt2, 0, 300);
pixDisplay(pixt3, 0, 570);
pixWrite("/tmp/junkpixt1.png", pixt1, IFF_PNG);
pixWrite("/tmp/junkpixt2.png", pixt2, IFF_PNG);
pixWrite("/tmp/junkpixt3.png", pixt3, IFF_PNG);
boxaDestroy(&boxa);
boxDestroy(&box);
pixDestroy(&pixs);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
/* ----------------------------------------------------------- *
* Test box transforms with either translation or scaling *
* combined with rotation, using the simple 'ordered' *
* function. Show that the order of the operations does *
* not matter; different hashing schemes end up in the *
* identical boxes. *
* ----------------------------------------------------------- */
pix = pixRead("feyn.tif");
box = boxCreate(420, 360, 1500, 465);
pixt = pixClipRectangle(pix, box, NULL);
pixs = pixAddBorderGeneral(pixt, 0, 200, 0, 0, 0);
boxDestroy(&box);
pixDestroy(&pix);
pixDestroy(&pixt);
boxa = pixConnComp(pixs, NULL, 8);
n = boxaGetCount(boxa);
pixa = pixaCreate(0);
pixt = pixConvertTo32(pixs);
for (i = 0; i < 3; i++) {
if (i == 0)
order = L_TR_SC_RO;
else if (i == 1)
order = L_TR_RO_SC;
else
order = L_SC_TR_RO;
boxat = boxaTransformOrdered(boxa, SHIFTX_2, SHIFTY_2, 1.0, 1.0,
450, 250, ROTATION_2, order);
RenderTransformedBoxa(pixt, boxat, i);
boxaDestroy(&boxat);
}
pixSaveTiled(pixt, pixa, 1, 1, 30, 32);
pixDestroy(&pixt);
pixt = pixConvertTo32(pixs);
for (i = 0; i < 3; i++) {
if (i == 0)
order = L_RO_TR_SC;
else if (i == 1)
order = L_RO_SC_TR;
else
order = L_SC_RO_TR;
boxat = boxaTransformOrdered(boxa, SHIFTX_2, SHIFTY_2, 1.0, 1.0,
450, 250, ROTATION_2, order);
RenderTransformedBoxa(pixt, boxat, i + 4);
boxaDestroy(&boxat);
}
pixSaveTiled(pixt, pixa, 1, 1, 30, 0);
pixDestroy(&pixt);
pixt = pixConvertTo32(pixs);
for (i = 0; i < 3; i++) {
if (i == 0)
order = L_TR_SC_RO;
else if (i == 1)
order = L_SC_RO_TR;
else
order = L_SC_TR_RO;
boxat = boxaTransformOrdered(boxa, 0, 0, SCALEX_2, SCALEY_2,
450, 250, ROTATION_2, order);
RenderTransformedBoxa(pixt, boxat, i + 8);
boxaDestroy(&boxat);
}
pixSaveTiled(pixt, pixa, 1, 1, 30, 0);
pixDestroy(&pixt);
pixt = pixConvertTo32(pixs);
for (i = 0; i < 3; i++) {
if (i == 0)
order = L_RO_TR_SC;
else if (i == 1)
order = L_RO_SC_TR;
else
order = L_TR_RO_SC;
boxat = boxaTransformOrdered(boxa, 0, 0, SCALEX_2, SCALEY_2,
450, 250, ROTATION_2, order);
RenderTransformedBoxa(pixt, boxat, i + 16);
boxaDestroy(&boxat);
}
pixSaveTiled(pixt, pixa, 1, 1, 30, 0);
pixDestroy(&pixt);
pixt = pixaDisplay(pixa, 0, 0);
pixWrite("/tmp/junkxform1.png", pixt, IFF_PNG);
pixDisplay(pixt, 1000, 0);
pixDestroy(&pixt);
pixDestroy(&pixs);
boxaDestroy(&boxa);
pixaDestroy(&pixa);
/* ----------------------------------------------------------- *
* Do more testing of box and pta transforms. Show that *
* resulting boxes are identical by three methods. *
* ----------------------------------------------------------- */
/* Set up pix and boxa */
pixa = pixaCreate(0);
pix = pixRead("lucasta.1.300.tif");
pixTranslate(pix, pix, 70, 0, L_BRING_IN_WHITE);
pixt = pixCloseBrick(NULL, pix, 14, 5);
pixOpenBrick(pixt, pixt, 1, 2);
boxa = pixConnComp(pixt, NULL, 8);
pixs = pixConvertTo32(pix);
pixc = pixCopy(NULL, pixs);
RenderTransformedBoxa(pixc, boxa, 113);
pixSaveTiled(pixc, pixa, 2, 1, 30, 32);
pixDestroy(&pix);
pixDestroy(&pixc);
pixDestroy(&pixt);
/* (a) Do successive discrete operations: shift, scale, rotate */
pixt1 = pixTranslate(NULL, pixs, SHIFTX_3, SHIFTY_3, L_BRING_IN_WHITE);
boxa1 = boxaTranslate(boxa, SHIFTX_3, SHIFTY_3);
pixc = pixCopy(NULL, pixt1);
RenderTransformedBoxa(pixc, boxa1, 213);
pixSaveTiled(pixc, pixa, 2, 0, 30, 32);
pixDestroy(&pixc);
pixt2 = pixScale(pixt1, SCALEX_3, SCALEY_3);
boxa2 = boxaScale(boxa1, SCALEX_3, SCALEY_3);
pixc = pixCopy(NULL, pixt2);
RenderTransformedBoxa(pixc, boxa2, 313);
pixSaveTiled(pixc, pixa, 2, 1, 30, 32);
pixDestroy(&pixc);
pixGetDimensions(pixt2, &w, &h, NULL);
pixt3 = pixRotateAM(pixt2, ROTATION_3, L_BRING_IN_WHITE);
boxa3 = boxaRotate(boxa2, w / 2, h / 2, ROTATION_3);
pixc = pixCopy(NULL, pixt3);
RenderTransformedBoxa(pixc, boxa3, 413);
pixSaveTiled(pixc, pixa, 2, 0, 30, 32);
pixDestroy(&pixc);
/* (b) Set up and use the composite transform */
mat1 = createMatrix2dTranslate(SHIFTX_3, SHIFTY_3);
mat2 = createMatrix2dScale(SCALEX_3, SCALEY_3);
mat3 = createMatrix2dRotate(w / 2, h / 2, ROTATION_3);
l_productMat3(mat3, mat2, mat1, matd, 3);
boxa4 = boxaAffineTransform(boxa, matd);
pixc = pixCopy(NULL, pixt3);
RenderTransformedBoxa(pixc, boxa4, 513);
pixSaveTiled(pixc, pixa, 2, 1, 30, 32);
pixDestroy(&pixc);
/* (c) Use the special 'ordered' function */
pixGetDimensions(pixs, &ws, &hs, NULL);
boxa5 = boxaTransformOrdered(boxa, SHIFTX_3, SHIFTY_3,
SCALEX_3, SCALEY_3,
ws / 2, hs / 2, ROTATION_3, L_TR_SC_RO);
pixc = pixCopy(NULL, pixt3);
RenderTransformedBoxa(pixc, boxa5, 613);
pixSaveTiled(pixc, pixa, 2, 0, 30, 32);
pixDestroy(&pixc);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
boxaDestroy(&boxa4);
boxaDestroy(&boxa5);
lept_free(mat1);
lept_free(mat2);
lept_free(mat3);
pixt = pixaDisplay(pixa, 0, 0);
pixWrite("/tmp/junkxform2.png", pixt, IFF_PNG);
pixDisplay(pixt, 1000, 300);
pixDestroy(&pixt);
pixDestroy(&pixs);
boxaDestroy(&boxa);
pixaDestroy(&pixa);
return 0;
}
int main(int argc,
char **argv)
{
l_int32 i, j;
l_float32 f;
l_uint32 redval, greenval;
PIX *pixs, *pixd, *pix0, *pix1, *pix2;
static char mainName[] = "locminmax_reg";
if (argc != 1)
return ERROR_INT("syntax: locminmax_reg", mainName, 1);
pixs = pixCreate(500, 500, 8);
for (i = 0; i < 500; i++) {
for (j = 0; j < 500; j++) {
f = 128.0 + 26.3 * sin(0.0438 * (l_float32)i);
f += 33.4 * cos(0.0712 * (l_float32)i);
f += 18.6 * sin(0.0561 * (l_float32)j);
f += 23.6 * cos(0.0327 * (l_float32)j);
pixSetPixel(pixs, j, i, (l_int32)f);
}
}
pixDisplay(pixs, 0, 0);
pixWrite("/tmp/junkpattern.png", pixs, IFF_PNG);
startTimer();
/* pixSelectedLocalExtrema(pixs, 1, &pix1, &pix2); */
pixLocalExtrema(pixs, 0, 0, &pix1, &pix2);
fprintf(stderr, "Time for extrema: %7.3f\n", stopTimer());
composeRGBPixel(255, 0, 0, &redval);
composeRGBPixel(0, 255, 0, &greenval);
pixd = pixConvertTo32(pixs);
pixPaintThroughMask(pixd, pix2, 0, 0, greenval);
pixPaintThroughMask(pixd, pix1, 0, 0, redval);
pixDisplay(pixd, 510, 0);
pixWrite("/tmp/junkpixd.png", pixd, IFF_PNG);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pixs);
pixDestroy(&pixd);
pix0 = pixRead("karen8.jpg");
pixs = pixBlockconv(pix0, 10, 10);
pixDisplay(pixs, 0, 400);
pixWrite("/tmp/junkconv.png", pixs, IFF_PNG);
startTimer();
/* pixSelectedLocalExtrema(pixs, 1, &pix1, &pix2); */
pixLocalExtrema(pixs, 50, 100, &pix1, &pix2);
fprintf(stderr, "Time for extrema: %7.3f\n", stopTimer());
composeRGBPixel(255, 0, 0, &redval);
composeRGBPixel(0, 255, 0, &greenval);
pixd = pixConvertTo32(pixs);
pixPaintThroughMask(pixd, pix2, 0, 0, greenval);
pixPaintThroughMask(pixd, pix1, 0, 0, redval);
pixDisplay(pixd, 350, 400);
pixWrite("/tmp/junkpixd2.png", pixd, IFF_PNG);
pixDestroy(&pix0);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pixs);
pixDestroy(&pixd);
return 0;
}
main(int argc,
char **argv)
{
char *str;
l_int32 i, j, same, ok;
l_float32 sum, avediff, rmsdiff;
L_KERNEL *kel1, *kel2, *kel3, *kel4, *kelx, *kely;
BOX *box;
PIX *pix, *pixs, *pixb, *pixg, *pixr, *pixd, *pixp, *pixt;
PIX *pixt1, *pixt2, *pixt3;
PIXA *pixa;
SARRAY *sa;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
pixa = pixaCreate(0);
/* Test creating from a string */
kel1 = kernelCreateFromString(5, 5, 2, 2, kdatastr);
pixd = kernelDisplayInPix(kel1, 41, 2);
pixWrite("/tmp/pixkern.png", pixd, IFF_PNG);
regTestCheckFile(rp, "/tmp/pixkern.png"); /* 0 */
pixSaveTiled(pixd, pixa, 1, 1, 20, 8);
pixDestroy(&pixd);
kernelDestroy(&kel1);
/* Test read/write for kernel. Note that both get
* compared to the same golden file, which is
* overwritten with a copy of /tmp/kern2.kel */
kel1 = kernelCreateFromString(5, 5, 2, 2, kdatastr);
kernelWrite("/tmp/kern1.kel", kel1);
regTestCheckFile(rp, "/tmp/kern1.kel"); /* 1 */
kel2 = kernelRead("/tmp/kern1.kel");
kernelWrite("/tmp/kern2.kel", kel2);
regTestCheckFile(rp, "/tmp/kern2.kel"); /* 2 */
regTestCompareFiles(rp, 1, 2); /* 3 */
kernelDestroy(&kel1);
kernelDestroy(&kel2);
/* Test creating from a file */
sa = sarrayCreate(0);
sarrayAddString(sa, (char *)"# small 3x3 kernel", L_COPY);
sarrayAddString(sa, (char *)"3 5", L_COPY);
sarrayAddString(sa, (char *)"1 2", L_COPY);
sarrayAddString(sa, (char *)"20.5 50 80 50 20", L_COPY);
sarrayAddString(sa, (char *)"82. 120 180 120 80", L_COPY);
sarrayAddString(sa, (char *)"22.1 50 80 50 20", L_COPY);
str = sarrayToString(sa, 1);
l_binaryWrite("/tmp/kernfile.kel", "w", str, strlen(str));
kel2 = kernelCreateFromFile("/tmp/kernfile.kel");
pixd = kernelDisplayInPix(kel2, 41, 2);
pixSaveTiled(pixd, pixa, 1, 1, 20, 0);
pixWrite("/tmp/ker1.png", pixd, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker1.png"); /* 4 */
pixDestroy(&pixd);
sarrayDestroy(&sa);
lept_free(str);
kernelDestroy(&kel2);
/* Test creating from a pix */
pixt = pixCreate(5, 3, 8);
pixSetPixel(pixt, 0, 0, 20);
pixSetPixel(pixt, 1, 0, 50);
pixSetPixel(pixt, 2, 0, 80);
pixSetPixel(pixt, 3, 0, 50);
pixSetPixel(pixt, 4, 0, 20);
pixSetPixel(pixt, 0, 1, 80);
pixSetPixel(pixt, 1, 1, 120);
pixSetPixel(pixt, 2, 1, 180);
pixSetPixel(pixt, 3, 1, 120);
pixSetPixel(pixt, 4, 1, 80);
pixSetPixel(pixt, 0, 0, 20);
pixSetPixel(pixt, 1, 2, 50);
pixSetPixel(pixt, 2, 2, 80);
pixSetPixel(pixt, 3, 2, 50);
pixSetPixel(pixt, 4, 2, 20);
kel3 = kernelCreateFromPix(pixt, 1, 2);
pixd = kernelDisplayInPix(kel3, 41, 2);
pixSaveTiled(pixd, pixa, 1, 0, 20, 0);
pixWrite("/tmp/ker2.png", pixd, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker2.png"); /* 5 */
pixDestroy(&pixd);
pixDestroy(&pixt);
kernelDestroy(&kel3);
/* Test convolution with kel1 */
pixs = pixRead("test24.jpg");
pixg = pixScaleRGBToGrayFast(pixs, 3, COLOR_GREEN);
pixSaveTiled(pixg, pixa, 1, 1, 20, 0);
kel1 = kernelCreateFromString(5, 5, 2, 2, kdatastr);
pixd = pixConvolve(pixg, kel1, 8, 1);
pixSaveTiled(pixd, pixa, 1, 0, 20, 0);
pixWrite("/tmp/ker3.png", pixd, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker3.png"); /* 6 */
pixDestroy(&pixs);
pixDestroy(&pixg);
pixDestroy(&pixd);
kernelDestroy(&kel1);
/* Test convolution with flat rectangular kel; also test
* block convolution with tiling. */
pixs = pixRead("test24.jpg");
pixg = pixScaleRGBToGrayFast(pixs, 3, COLOR_GREEN);
kel2 = makeFlatKernel(11, 11, 5, 5);
pixd = pixConvolve(pixg, kel2, 8, 1);
pixSaveTiled(pixd, pixa, 1, 1, 20, 0);
pixWrite("/tmp/ker4.png", pixd, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker4.png"); /* 7 */
pixt = pixBlockconv(pixg, 5, 5);
pixSaveTiled(pixt, pixa, 1, 0, 20, 0);
pixWrite("/tmp/ker5.png", pixt, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker5.png"); /* 8 */
if (rp->display)
pixCompareGray(pixd, pixt, L_COMPARE_ABS_DIFF, GPLOT_X11, NULL,
NULL, NULL, NULL);
pixt2 = pixBlockconvTiled(pixg, 5, 5, 3, 6);
pixSaveTiled(pixt2, pixa, 1, 0, 20, 0);
pixWrite("/tmp/ker5a.png", pixt2, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker5a.png"); /* 9 */
pixDestroy(&pixt2);
ok = TRUE;
for (i = 1; i <= 7; i++) {
for (j = 1; j <= 7; j++) {
if (i == 1 && j == 1) continue;
pixt2 = pixBlockconvTiled(pixg, 5, 5, j, i);
pixEqual(pixt2, pixd, &same);
if (!same) {
fprintf(stderr," Error for nx = %d, ny = %d\n", j, i);
ok = FALSE;
}
pixDestroy(&pixt2);
}
}
if (ok)
fprintf(stderr, "OK: Tiled results identical to pixConvolve()\n");
else
fprintf(stderr, "ERROR: Tiled results not identical to pixConvolve()\n");
pixDestroy(&pixs);
pixDestroy(&pixg);
pixDestroy(&pixd);
pixDestroy(&pixt);
kernelDestroy(&kel2);
/* Do another flat rectangular test; this time with white at edge.
* About 1% of the pixels near the image edge differ by 1 between
* the pixConvolve() and pixBlockconv(). For what it's worth,
* pixConvolve() gives the more accurate result; namely, 255 for
* pixels at the edge. */
pix = pixRead("pageseg1.tif");
box = boxCreate(100, 100, 2260, 3160);
pixb = pixClipRectangle(pix, box, NULL);
pixs = pixScaleToGray4(pixb);
kel3 = makeFlatKernel(7, 7, 3, 3);
startTimer();
pixt = pixConvolve(pixs, kel3, 8, 1);
fprintf(stderr, "Generic convolution time: %5.3f sec\n", stopTimer());
pixSaveTiled(pixt, pixa, 1, 1, 20, 0);
pixWrite("/tmp/conv1.png", pixt, IFF_PNG);
regTestCheckFile(rp, "/tmp/conv1.png"); /* 10 */
startTimer();
pixt2 = pixBlockconv(pixs, 3, 3);
fprintf(stderr, "Flat block convolution time: %5.3f sec\n", stopTimer());
pixSaveTiled(pixt2, pixa, 1, 0, 20, 0);
pixWrite("/tmp/conv2.png", pixt2, IFF_PNG); /* ditto */
regTestCheckFile(rp, "/tmp/conv2.png"); /* 11 */
pixCompareGray(pixt, pixt2, L_COMPARE_ABS_DIFF, GPLOT_PNG, NULL,
&avediff, &rmsdiff, NULL);
#ifndef _WIN32
sleep(1); /* give gnuplot time to write out the file */
#else
Sleep(1000);
#endif /* _WIN32 */
pixp = pixRead("/tmp/grayroot.png");
pixSaveTiled(pixp, pixa, 1, 0, 20, 0);
pixWrite("/tmp/conv3.png", pixp, IFF_PNG);
regTestCheckFile(rp, "/tmp/conv3.png"); /* 12 */
fprintf(stderr, "Ave diff = %6.4f, RMS diff = %6.4f\n", avediff, rmsdiff);
if (avediff <= 0.01)
fprintf(stderr, "OK: avediff = %6.4f <= 0.01\n", avediff);
else
fprintf(stderr, "Bad?: avediff = %6.4f > 0.01\n", avediff);
pixDestroy(&pixt);
pixDestroy(&pixt2);
pixDestroy(&pixs);
pixDestroy(&pixp);
pixDestroy(&pix);
pixDestroy(&pixb);
boxDestroy(&box);
kernelDestroy(&kel3);
/* Do yet another set of flat rectangular tests, this time
* on an RGB image */
pixs = pixRead("test24.jpg");
kel4 = makeFlatKernel(7, 7, 3, 3);
startTimer();
pixt1 = pixConvolveRGB(pixs, kel4);
fprintf(stderr, "Time 7x7 non-separable: %7.3f sec\n", stopTimer());
pixWrite("/tmp/conv4.jpg", pixt1, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/conv4.jpg"); /* 13 */
kelx = makeFlatKernel(1, 7, 0, 3);
kely = makeFlatKernel(7, 1, 3, 0);
startTimer();
pixt2 = pixConvolveRGBSep(pixs, kelx, kely);
fprintf(stderr, "Time 7x1,1x7 separable: %7.3f sec\n", stopTimer());
pixWrite("/tmp/conv5.jpg", pixt2, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/conv5.jpg"); /* 14 */
startTimer();
pixt3 = pixBlockconv(pixs, 3, 3);
fprintf(stderr, "Time 7x7 blockconv: %7.3f sec\n", stopTimer());
pixWrite("/tmp/conv6.jpg", pixt3, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/conv6.jpg"); /* 15 */
regTestComparePix(rp, pixt1, pixt2); /* 16 */
regTestCompareSimilarPix(rp, pixt2, pixt3, 15, 0.0005, 0); /* 17 */
pixDestroy(&pixs);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
kernelDestroy(&kel4);
kernelDestroy(&kelx);
kernelDestroy(&kely);
/* Test generation and convolution with gaussian kernel */
pixs = pixRead("test8.jpg");
pixSaveTiled(pixs, pixa, 1, 1, 20, 0);
kel1 = makeGaussianKernel(5, 5, 3.0, 5.0);
kernelGetSum(kel1, &sum);
fprintf(stderr, "Sum for gaussian kernel = %f\n", sum);
kernelWrite("/tmp/gauss.kel", kel1);
pixt = pixConvolve(pixs, kel1, 8, 1);
pixt2 = pixConvolve(pixs, kel1, 16, 0);
pixSaveTiled(pixt, pixa, 1, 0, 20, 0);
pixSaveTiled(pixt2, pixa, 1, 0, 20, 0);
pixWrite("/tmp/ker6.png", pixt, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker6.png"); /* 18 */
pixDestroy(&pixt);
pixDestroy(&pixt2);
pixt = kernelDisplayInPix(kel1, 25, 2);
pixSaveTiled(pixt, pixa, 1, 0, 20, 0);
pixDestroy(&pixt);
kernelDestroy(&kel1);
pixDestroy(&pixs);
/* Test generation and convolution with separable gaussian kernel */
pixs = pixRead("test8.jpg");
pixSaveTiled(pixs, pixa, 1, 1, 20, 0);
makeGaussianKernelSep(5, 5, 3.0, 5.0, &kelx, &kely);
kernelGetSum(kelx, &sum);
fprintf(stderr, "Sum for x gaussian kernel = %f\n", sum);
kernelGetSum(kely, &sum);
fprintf(stderr, "Sum for y gaussian kernel = %f\n", sum);
kernelWrite("/tmp/gauss.kelx", kelx);
kernelWrite("/tmp/gauss.kely", kely);
pixt = pixConvolveSep(pixs, kelx, kely, 8, 1);
pixt2 = pixConvolveSep(pixs, kelx, kely, 16, 0);
pixSaveTiled(pixt, pixa, 1, 0, 20, 0);
pixSaveTiled(pixt2, pixa, 1, 0, 20, 0);
pixWrite("/tmp/ker7.png", pixt, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker7.png"); /* 19 */
pixDestroy(&pixt);
pixDestroy(&pixt2);
pixt = kernelDisplayInPix(kelx, 25, 2);
pixSaveTiled(pixt, pixa, 1, 0, 20, 0);
pixDestroy(&pixt);
pixt = kernelDisplayInPix(kely, 25, 2);
pixSaveTiled(pixt, pixa, 1, 0, 20, 0);
pixDestroy(&pixt);
kernelDestroy(&kelx);
kernelDestroy(&kely);
pixDestroy(&pixs);
/* Test generation and convolution with diff of gaussians kernel */
/* pixt = pixRead("marge.jpg");
pixs = pixConvertRGBToLuminance(pixt);
pixDestroy(&pixt); */
pixs = pixRead("test8.jpg");
pixSaveTiled(pixs, pixa, 1, 1, 20, 0);
kel1 = makeDoGKernel(7, 7, 1.5, 2.7);
kernelGetSum(kel1, &sum);
fprintf(stderr, "Sum for DoG kernel = %f\n", sum);
kernelWrite("/tmp/dog.kel", kel1);
pixt = pixConvolve(pixs, kel1, 8, 0);
/* pixInvert(pixt, pixt); */
pixSaveTiled(pixt, pixa, 1, 0, 20, 0);
pixWrite("/tmp/ker8.png", pixt, IFF_PNG);
regTestCheckFile(rp, "/tmp/ker8.png"); /* 20 */
pixDestroy(&pixt);
pixt = kernelDisplayInPix(kel1, 20, 2);
pixSaveTiled(pixt, pixa, 1, 0, 20, 0);
pixDestroy(&pixt);
kernelDestroy(&kel1);
pixDestroy(&pixs);
pixd = pixaDisplay(pixa, 0, 0);
pixDisplayWithTitle(pixd, 100, 100, NULL, rp->display);
pixWrite("/tmp/kernel.jpg", pixd, IFF_JFIF_JPEG);
pixDestroy(&pixd);
pixaDestroy(&pixa);
regTestCleanup(rp);
return 0;
}
int main(int argc,
char **argv)
{
l_int32 i, w, h;
l_float32 factor, scale;
BOX *box;
FILE *fp1;
PIX *pixs, *pixt;
PIXA *pixa;
SARRAY *sa;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
factor = 0.95;
/* Uncompressed PS with scaling but centered on the page */
pixs = pixRead("feyn-fract.tif");
pixGetDimensions(pixs, &w, &h, NULL);
scale = L_MIN(factor * 2550 / w, factor * 3300 / h);
fp1 = lept_fopen("/tmp/regout/psio0.ps", "wb+");
pixWriteStreamPS(fp1, pixs, NULL, 300, scale);
lept_fclose(fp1);
regTestCheckFile(rp, "/tmp/regout/psio0.ps"); /* 0 */
pixDestroy(&pixs);
/* Uncompressed PS with scaling, with LL corner at (1500, 1500) mils */
pixs = pixRead("weasel4.11c.png");
pixGetDimensions(pixs, &w, &h, NULL);
scale = L_MIN(factor * 2550 / w, factor * 3300 / h);
box = boxCreate(1500, 1500, (l_int32)(1000 * scale * w / 300),
(l_int32)(1000 * scale * h / 300));
fp1 = lept_fopen("/tmp/regout/psio1.ps", "wb+");
pixWriteStreamPS(fp1, pixs, box, 300, 1.0);
lept_fclose(fp1);
regTestCheckFile(rp, "/tmp/regout/psio1.ps"); /* 1 */
boxDestroy(&box);
pixDestroy(&pixs);
/* DCT compressed PS with LL corner at (300, 1000) pixels */
pixs = pixRead("marge.jpg");
pixt = pixConvertTo32(pixs);
pixWrite("/tmp/regout/psio2.jpg", pixt, IFF_JFIF_JPEG);
convertJpegToPS("/tmp/regout/psio2.jpg", "/tmp/regout/psio3.ps",
"w", 300, 1000, 0, 4.0, 1, 1);
regTestCheckFile(rp, "/tmp/regout/psio2.jpg"); /* 2 */
regTestCheckFile(rp, "/tmp/regout/psio3.ps"); /* 3 */
pixDestroy(&pixt);
pixDestroy(&pixs);
/* For each page, apply tiff g4 image first; then jpeg or png over it */
convertG4ToPS("feyn.tif", "/tmp/regout/psio4.ps", "w",
0, 0, 0, 1.0, 1, 1, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio4.ps",
"a", 500, 100, 300, 2.0, 1, 0);
convertFlateToPS("weasel4.11c.png", "/tmp/regout/psio4.ps",
"a", 300, 400, 300, 6.0, 1, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio4.ps",
"a", 100, 800, 300, 1.5, 1, 1);
convertG4ToPS("feyn.tif", "/tmp/regout/psio4.ps",
"a", 0, 0, 0, 1.0, 2, 1, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio4.ps",
"a", 1000, 700, 300, 2.0, 2, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio4.ps",
"a", 100, 200, 300, 2.0, 2, 1);
convertG4ToPS("feyn.tif", "/tmp/regout/psio4.ps",
"a", 0, 0, 0, 1.0, 3, 1, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio4.ps",
"a", 200, 200, 300, 2.0, 3, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio4.ps",
"a", 200, 900, 300, 2.0, 3, 1);
regTestCheckFile(rp, "/tmp/regout/psio4.ps"); /* 4 */
/* Now apply jpeg first; then paint through a g4 mask.
* For gv, the first image with a b.b. determines the
* window size for the canvas, so we put down the largest
* image first. If we had rendered a small image first,
* gv and evince will not show the entire page. However, after
* conversion to pdf, everything works fine, regardless of the
* order in which images are placed into the PS. That is
* because the pdf interpreter is robust to bad hints, ignoring
* the page hints and computing the bounding box from the
* set of images rendered on the page.
*
* Concatenate several pages, with colormapped png, color
* jpeg and tiffg4 images (with the g4 image acting as a mask
* that we're painting black through. If the text layer
* is painted first, the following images occlude it; otherwise,
* the images remain in the background of the text. */
pixs = pixRead("wyom.jpg");
pixt = pixScaleToSize(pixs, 2528, 3300);
pixWrite("/tmp/regout/psio5.jpg", pixt, IFF_JFIF_JPEG);
pixDestroy(&pixs);
pixDestroy(&pixt);
convertJpegToPS("/tmp/regout/psio5.jpg", "/tmp/regout/psio5.ps",
"w", 0, 0, 300, 1.0, 1, 0);
convertFlateToPS("weasel8.240c.png", "/tmp/regout/psio5.ps",
"a", 100, 100, 300, 5.0, 1, 0);
convertFlateToPS("weasel8.149g.png", "/tmp/regout/psio5.ps",
"a", 200, 300, 300, 5.0, 1, 0);
convertFlateToPS("weasel4.11c.png", "/tmp/regout/psio5.ps",
"a", 300, 500, 300, 5.0, 1, 0);
convertG4ToPS("feyn.tif", "/tmp/regout/psio5.ps",
"a", 0, 0, 0, 1.0, 1, 1, 1);
convertJpegToPS("marge.jpg", "/tmp/regout/psio5.ps",
"a", 500, 100, 300, 2.0, 2, 0);
convertFlateToPS("weasel4.11c.png", "/tmp/regout/psio5.ps",
"a", 300, 400, 300, 6.0, 2, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio5.ps",
"a", 100, 800, 300, 1.5, 2, 0);
convertG4ToPS("feyn.tif", "/tmp/regout/psio5.ps",
"a", 0, 0, 0, 1.0, 2, 1, 1);
convertJpegToPS("marge.jpg", "/tmp/regout/psio5.ps",
"a", 500, 100, 300, 2.0, 3, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio5.ps",
"a", 100, 800, 300, 2.0, 3, 0);
convertG4ToPS("feyn.tif", "/tmp/regout/psio5.ps",
"a", 0, 0, 0, 1.0, 3, 1, 1);
convertJpegToPS("marge.jpg", "/tmp/regout/psio5.ps",
"a", 700, 700, 300, 2.0, 4, 0);
convertFlateToPS("weasel8.149g.png", "/tmp/regout/psio5.ps",
"a", 400, 400, 300, 5.0, 4, 0);
convertG4ToPS("feyn.tif", "/tmp/regout/psio5.ps",
"a", 0, 0, 0, 1.0, 4, 1, 0);
convertFlateToPS("weasel8.240c.png", "/tmp/regout/psio5.ps",
"a", 100, 220, 300, 5.0, 4, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio5.ps",
"a", 100, 200, 300, 2.0, 4, 1);
convertJpegToPS("marge.jpg", "/tmp/regout/psio5.ps",
"a", 200, 200, 300, 1.5, 5, 0);
convertFlateToPS("weasel8.240c.png", "/tmp/regout/psio5.ps",
"a", 140, 80, 300, 7.0, 5, 0);
convertG4ToPS("feyn.tif", "/tmp/regout/psio5.ps",
"a", 0, 0, 0, 1.0, 5, 1, 0);
convertFlateToPS("weasel8.149g.png", "/tmp/regout/psio5.ps",
"a", 280, 310, 300, 5.0, 4, 0);
convertJpegToPS("marge.jpg", "/tmp/regout/psio5.ps",
"a", 200, 900, 300, 2.0, 5, 1);
regTestCheckFile(rp, "/tmp/regout/psio5.ps"); /* 5 */
/* Generation using segmentation masks */
convertSegmentedPagesToPS(".", "lion-page", 10, ".", "lion-mask", 10,
0, 100, 2.0, 0.8, 190, "/tmp/regout/psio6.ps");
regTestCheckFile(rp, "/tmp/regout/psio6.ps"); /* 6 */
/* PS generation for embeddding */
convertJpegToPSEmbed("tetons.jpg", "/tmp/regout/psio7.ps");
regTestCheckFile(rp, "/tmp/regout/psio7.ps"); /* 7 */
convertG4ToPSEmbed("feyn-fract.tif", "/tmp/regout/psio8.ps");
regTestCheckFile(rp, "/tmp/regout/psio8.ps"); /* 8 */
convertFlateToPSEmbed("weasel8.240c.png", "/tmp/regout/psio9.ps");
regTestCheckFile(rp, "/tmp/regout/psio9.ps"); /* 9 */
/* Writing compressed from a pixa */
sa = sarrayCreate(0);
for (i = 0; i < 11; i++)
sarrayAddString(sa, WeaselNames[i], L_COPY);
pixa = pixaReadFilesSA(sa);
pixaWriteCompressedToPS(pixa, "/tmp/regout/psio10.ps", 0, 3);
regTestCheckFile(rp, "/tmp/regout/psio10.ps"); /* 10 */
pixaDestroy(&pixa);
sarrayDestroy(&sa);
return regTestCleanup(rp);
}
/*!
* pixDisplayWithTitle()
*
* Input: pix (1, 2, 4, 8, 16, 32 bpp)
* x, y (location of display frame)
* title (<optional> on frame; can be NULL);
* dispflag (1 to write, else disabled)
* Return: 0 if OK; 1 on error
*
* Notes:
* (1) See notes for pixDisplay().
* (2) This displays the image if dispflag == 1.
*/
l_int32
pixDisplayWithTitle(PIX *pixs,
l_int32 x,
l_int32 y,
const char *title,
l_int32 dispflag)
{
char *tempname;
char buffer[L_BUF_SIZE];
static l_int32 index = 0; /* caution: not .so or thread safe */
l_int32 w, h, d, ignore;
l_float32 ratw, rath, ratmin;
PIX *pixt;
#ifndef _WIN32
l_int32 wt, ht;
#else
char pathname[_MAX_PATH];
#endif /* _WIN32 */
PROCNAME("pixDisplayWithTitle");
if (dispflag != 1) return 0;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (ChosenDisplayProg != L_DISPLAY_WITH_XV &&
ChosenDisplayProg != L_DISPLAY_WITH_XLI &&
ChosenDisplayProg != L_DISPLAY_WITH_XZGV &&
ChosenDisplayProg != L_DISPLAY_WITH_IV)
return ERROR_INT("no program chosen for display", procName, 1);
pixGetDimensions(pixs, &w, &h, &d);
if (w <= MAX_DISPLAY_WIDTH && h <= MAX_DISPLAY_HEIGHT) {
if (d == 16) /* take MSB */
pixt = pixConvert16To8(pixs, 1);
else
pixt = pixClone(pixs);
}
else {
ratw = (l_float32)MAX_DISPLAY_WIDTH / (l_float32)w;
rath = (l_float32)MAX_DISPLAY_HEIGHT / (l_float32)h;
ratmin = L_MIN(ratw, rath);
if (ratmin < 0.125 && d == 1)
pixt = pixScaleToGray8(pixs);
else if (ratmin < 0.25 && d == 1)
pixt = pixScaleToGray4(pixs);
else if (ratmin < 0.33 && d == 1)
pixt = pixScaleToGray3(pixs);
else if (ratmin < 0.5 && d == 1)
pixt = pixScaleToGray2(pixs);
else
pixt = pixScale(pixs, ratmin, ratmin);
if (!pixt)
return ERROR_INT("pixt not made", procName, 1);
}
if (index == 0) {
snprintf(buffer, L_BUF_SIZE, "rm -f /tmp/junk_display.*");
ignore = system(buffer);
}
index++;
if (pixGetDepth(pixt) < 8 ||
(w < MAX_SIZE_FOR_PNG && h < MAX_SIZE_FOR_PNG)) {
snprintf(buffer, L_BUF_SIZE, "/tmp/junk_display.%03d.png", index);
pixWrite(buffer, pixt, IFF_PNG);
}
else {
snprintf(buffer, L_BUF_SIZE, "/tmp/junk_display.%03d.jpg", index);
pixWrite(buffer, pixt, IFF_JFIF_JPEG);
}
tempname = stringNew(buffer);
#ifndef _WIN32
/* Unix */
if (ChosenDisplayProg == L_DISPLAY_WITH_XV) {
if (title)
snprintf(buffer, L_BUF_SIZE,
"xv -quit -geometry +%d+%d -name \"%s\" %s &",
x, y, title, tempname);
else
snprintf(buffer, L_BUF_SIZE,
"xv -quit -geometry +%d+%d %s &", x, y, tempname);
}
else if (ChosenDisplayProg == L_DISPLAY_WITH_XLI) {
if (title)
snprintf(buffer, L_BUF_SIZE,
"xli -dispgamma 1.0 -quiet -geometry +%d+%d -title \"%s\" %s &",
x, y, title, tempname);
else
snprintf(buffer, L_BUF_SIZE,
"xli -dispgamma 1.0 -quiet -geometry +%d+%d %s &",
x, y, tempname);
}
else if (ChosenDisplayProg == L_DISPLAY_WITH_XZGV) {
/* no way to display title */
pixGetDimensions(pixt, &wt, &ht, NULL);
snprintf(buffer, L_BUF_SIZE,
"xzgv --geometry %dx%d+%d+%d %s &", wt + 10, ht + 10,
x, y, tempname);
}
ignore = system(buffer);
#else /* _WIN32 */
/* Windows: L_DISPLAY_WITH_IV */
_fullpath(pathname, tempname, sizeof(pathname));
if (title)
snprintf(buffer, L_BUF_SIZE,
"i_view32.exe \"%s\" /pos=(%d,%d) /title=\"%s\"",
pathname, x, y, title);
else
snprintf(buffer, L_BUF_SIZE, "i_view32.exe \"%s\" /pos=(%d,%d)",
pathname, x, y);
ignore = system(buffer);
#endif /* _WIN32 */
pixDestroy(&pixt);
FREE(tempname);
return 0;
}
/*!
* pixConnCompPixa()
*
* Input: pixs (1 bpp)
* &pixa (<return> pixa of each c.c.)
* connectivity (4 or 8)
* Return: boxa, or null on error
*
* Notes:
* (1) This finds bounding boxes of 4- or 8-connected components
* in a binary image, and saves images of each c.c
* in a pixa array.
* (2) It sets up 2 temporary pix, and for each c.c. that is
* located in raster order, it erases the c.c. from one pix,
* then uses the b.b. to extract the c.c. from the two pix using
* an XOR, and finally erases the c.c. from the second pix.
* (3) A clone of the returned boxa (where all boxes in the array
* are clones) is inserted into the pixa.
* (4) If the input is valid, this always returns a boxa and a pixa.
* If pixs is empty, the boxa and pixa will be empty.
*/
BOXA *
pixConnCompPixa(PIX *pixs,
PIXA **ppixa,
l_int32 connectivity)
{
l_int32 h, iszero;
l_int32 x, y, xstart, ystart;
PIX *pixt1, *pixt2, *pixt3, *pixt4;
PIXA *pixa;
BOX *box;
BOXA *boxa;
L_STACK *stack, *auxstack;
PROCNAME("pixConnCompPixa");
if (!ppixa)
return (BOXA *)ERROR_PTR("&pixa not defined", procName, NULL);
*ppixa = NULL;
if (!pixs || pixGetDepth(pixs) != 1)
return (BOXA *)ERROR_PTR("pixs undefined or not 1 bpp", procName, NULL);
if (connectivity != 4 && connectivity != 8)
return (BOXA *)ERROR_PTR("connectivity not 4 or 8", procName, NULL);
pixa = pixaCreate(0);
*ppixa = pixa;
pixZero(pixs, &iszero);
if (iszero)
return boxaCreate(1); /* return empty boxa */
if ((pixt1 = pixCopy(NULL, pixs)) == NULL)
return (BOXA *)ERROR_PTR("pixt1 not made", procName, NULL);
if ((pixt2 = pixCopy(NULL, pixs)) == NULL)
return (BOXA *)ERROR_PTR("pixt2 not made", procName, NULL);
h = pixGetHeight(pixs);
if ((stack = lstackCreate(h)) == NULL)
return (BOXA *)ERROR_PTR("stack not made", procName, NULL);
if ((auxstack = lstackCreate(0)) == NULL)
return (BOXA *)ERROR_PTR("auxstack not made", procName, NULL);
stack->auxstack = auxstack;
if ((boxa = boxaCreate(0)) == NULL)
return (BOXA *)ERROR_PTR("boxa not made", procName, NULL);
xstart = 0;
ystart = 0;
while (1)
{
if (!nextOnPixelInRaster(pixt1, xstart, ystart, &x, &y))
break;
if ((box = pixSeedfillBB(pixt1, stack, x, y, connectivity)) == NULL)
return (BOXA *)ERROR_PTR("box not made", procName, NULL);
boxaAddBox(boxa, box, L_INSERT);
/* Save the c.c. and remove from pixt2 as well */
pixt3 = pixClipRectangle(pixt1, box, NULL);
pixt4 = pixClipRectangle(pixt2, box, NULL);
pixXor(pixt3, pixt3, pixt4);
pixRasterop(pixt2, box->x, box->y, box->w, box->h, PIX_SRC ^ PIX_DST,
pixt3, 0, 0);
pixaAddPix(pixa, pixt3, L_INSERT);
pixDestroy(&pixt4);
xstart = x;
ystart = y;
}
#if DEBUG
pixCountPixels(pixt1, &iszero, NULL);
fprintf(stderr, "Number of remaining pixels = %d\n", iszero);
pixWrite("junkremain", pixt1, IFF_PNG);
#endif /* DEBUG */
/* Remove old boxa of pixa and replace with a clone copy */
boxaDestroy(&pixa->boxa);
pixa->boxa = boxaCopy(boxa, L_CLONE);
/* Cleanup, freeing the fillsegs on each stack */
lstackDestroy(&stack, TRUE);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
return boxa;
}
/*!
* pixDisplayWriteFormat()
*
* Input: pix (1, 2, 4, 8, 16, 32 bpp)
* reduction (-1 to reset/erase; 0 to disable;
* otherwise this is a reduction factor)
* format (IFF_PNG or IFF_JFIF_JPEG)
* Return: 0 if OK; 1 on error
*
* Notes:
* (1) This writes files if reduction > 0. These can be displayed using
* pixDisplayMultiple("/tmp/junk_write_display*");
* (2) All previously written files can be erased by calling with
* reduction < 0; the value of pixs is ignored.
* (3) If reduction > 1 and depth == 1, this does a scale-to-gray
* reduction.
* (4) This function uses a static internal variable to number
* output files written by a single process. Behavior
* with a shared library may be unpredictable.
* (5) Output file format is as follows:
* format == IFF_JFIF_JPEG:
* png if d < 8 or d == 16 or if the output pix
* has a colormap. Otherwise, output is jpg.
* format == IFF_PNG:
* png (lossless) on all images.
* (6) For 16 bpp, the choice of full dynamic range with log scale
* is the best for displaying these images. Alternative outputs are
* pix8 = pixMaxDynamicRange(pixt, L_LINEAR_SCALE);
* pix8 = pixConvert16To8(pixt, 0); // low order byte
* pix8 = pixConvert16To8(pixt, 1); // high order byte
*/
l_int32
pixDisplayWriteFormat(PIX *pixs,
l_int32 reduction,
l_int32 format)
{
char buffer[L_BUF_SIZE];
l_int32 ignore;
l_float32 scale;
PIX *pixt, *pix8;
static l_int32 index = 0; /* caution: not .so or thread safe */
PROCNAME("pixDisplayWriteFormat");
if (reduction == 0) return 0;
if (reduction < 0) {
index = 0; /* reset; this will cause erasure at next call to write */
return 0;
}
if (format != IFF_JFIF_JPEG && format != IFF_PNG)
return ERROR_INT("invalid format", procName, 1);
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (index == 0) {
snprintf(buffer, L_BUF_SIZE,
"rm -f /tmp/junk_write_display.*.png /tmp/junk_write_display.*.jpg");
ignore = system(buffer);
}
index++;
if (reduction == 1)
pixt = pixClone(pixs);
else {
scale = 1. / (l_float32)reduction;
if (pixGetDepth(pixs) == 1)
pixt = pixScaleToGray(pixs, scale);
else
pixt = pixScale(pixs, scale, scale);
}
if (pixGetDepth(pixt) == 16) {
pix8 = pixMaxDynamicRange(pixt, L_LOG_SCALE);
snprintf(buffer, L_BUF_SIZE, "/tmp/junk_write_display.%03d.png", index);
pixWrite(buffer, pix8, IFF_PNG);
pixDestroy(&pix8);
}
else if (pixGetDepth(pixt) < 8 || pixGetColormap(pixt) ||
format == IFF_PNG) {
snprintf(buffer, L_BUF_SIZE, "/tmp/junk_write_display.%03d.png", index);
pixWrite(buffer, pixt, IFF_PNG);
}
else {
snprintf(buffer, L_BUF_SIZE, "/tmp/junk_write_display.%03d.jpg", index);
pixWrite(buffer, pixt, format);
}
pixDestroy(&pixt);
return 0;
}
// Creates and returns a Pix with the same resolution as the original
// in which 1 (black) pixels represent likely non text (photo, line drawing)
// areas of the page, deleting from the blob_block the blobs that were
// determined to be non-text.
// The photo_map is used to bias the decision towards non-text, rather than
// supplying definite decision.
// The blob_block is the usual result of connected component analysis,
// holding the detected blobs.
// The returned Pix should be PixDestroyed after use.
Pix* CCNonTextDetect::ComputeNonTextMask(bool debug, Pix* photo_map,
TO_BLOCK* blob_block) {
// Insert the smallest blobs into the grid.
InsertBlobList(&blob_block->small_blobs);
InsertBlobList(&blob_block->noise_blobs);
// Add the medium blobs that don't have a good strokewidth neighbour.
// Those that do go into good_grid as an antidote to spreading beyond the
// real reaches of a noise region.
BlobGrid good_grid(gridsize(), bleft(), tright());
BLOBNBOX_IT blob_it(&blob_block->blobs);
for (blob_it.mark_cycle_pt(); !blob_it.cycled_list(); blob_it.forward()) {
BLOBNBOX* blob = blob_it.data();
double perimeter_area_ratio = blob->cblob()->perimeter() / 4.0;
perimeter_area_ratio *= perimeter_area_ratio / blob->enclosed_area();
if (blob->GoodTextBlob() == 0 || perimeter_area_ratio < kMinGoodTextPARatio)
InsertBBox(true, true, blob);
else
good_grid.InsertBBox(true, true, blob);
}
noise_density_ = ComputeNoiseDensity(debug, photo_map, &good_grid);
good_grid.Clear(); // Not needed any more.
Pix* pix = noise_density_->ThresholdToPix(max_noise_count_);
if (debug) {
pixWrite("junknoisemask.png", pix, IFF_PNG);
}
ScrollView* win = NULL;
#ifndef GRAPHICS_DISABLED
if (debug) {
win = MakeWindow(0, 400, "Photo Mask Blobs");
}
#endif // GRAPHICS_DISABLED
// Large and medium blobs are not text if they overlap with "a lot" of small
// blobs.
MarkAndDeleteNonTextBlobs(&blob_block->large_blobs,
kMaxLargeOverlapsWithSmall,
win, ScrollView::DARK_GREEN, pix);
MarkAndDeleteNonTextBlobs(&blob_block->blobs, kMaxMediumOverlapsWithSmall,
win, ScrollView::WHITE, pix);
// Clear the grid of small blobs and insert the medium blobs.
Clear();
InsertBlobList(&blob_block->blobs);
MarkAndDeleteNonTextBlobs(&blob_block->large_blobs,
kMaxLargeOverlapsWithMedium,
win, ScrollView::DARK_GREEN, pix);
// Clear again before we start deleting the blobs in the grid.
Clear();
MarkAndDeleteNonTextBlobs(&blob_block->noise_blobs, -1,
win, ScrollView::CORAL, pix);
MarkAndDeleteNonTextBlobs(&blob_block->small_blobs, -1,
win, ScrollView::GOLDENROD, pix);
MarkAndDeleteNonTextBlobs(&blob_block->blobs, -1,
win, ScrollView::WHITE, pix);
if (debug) {
#ifndef GRAPHICS_DISABLED
win->Update();
#endif // GRAPHICS_DISABLED
pixWrite("junkccphotomask.png", pix, IFF_PNG);
#ifndef GRAPHICS_DISABLED
delete win->AwaitEvent(SVET_DESTROY);
delete win;
#endif // GRAPHICS_DISABLED
}
return pix;
}
main(int argc,
char **argv)
{
char label[512];
l_int32 w, h, i, j, rwhite, gwhite, bwhite, count;
GPLOT *gplot1, *gplot2;
NUMA *naseq, *na;
NUMAA *naa1, *naa2;
PIX *pixs, *pixt, *pixt0, *pixt1, *pixt2;
PIX *pixr, *pixg, *pixb; /* for color content extraction */
PIXA *pixa;
PIXCMAP *cmap;
static char mainName[] = "colorspacetest";
if (argc != 2)
exit(ERROR_INT(" Syntax: colorspacetest filein", mainName, 1));
if ((pixs = pixRead(argv[1])) == NULL)
exit(ERROR_INT("pixs not made", mainName, 1));
/* Colorspace conversion in rgb */
pixDisplayWrite(pixs, 1);
pixt = pixConvertRGBToHSV(NULL, pixs);
pixDisplayWrite(pixt, 1);
pixConvertHSVToRGB(pixt, pixt);
pixDisplayWrite(pixt, 1);
pixDestroy(&pixt);
/* Colorspace conversion on a colormap */
pixt = pixOctreeQuantNumColors(pixs, 25, 0);
pixDisplayWrite(pixt, 1);
cmap = pixGetColormap(pixt);
pixcmapWriteStream(stderr, cmap);
pixcmapConvertRGBToHSV(cmap);
pixcmapWriteStream(stderr, cmap);
pixDisplayWrite(pixt, 1);
pixcmapConvertHSVToRGB(cmap);
pixcmapWriteStream(stderr, cmap);
pixDisplayWrite(pixt, 1);
pixDestroy(&pixt);
/* Color content extraction */
pixColorContent(pixs, 0, 0, 0, 0, &pixr, &pixg, &pixb);
pixDisplayWrite(pixr, 1);
pixDisplayWrite(pixg, 1);
pixDisplayWrite(pixb, 1);
pixDestroy(&pixr);
pixDestroy(&pixg);
pixDestroy(&pixb);
/* Color content measurement */
pixa = pixaCreate(20);
naseq = numaMakeSequence(100, 5, 20);
naa1 = numaaCreate(6);
naa2 = numaaCreate(6);
for (i = 0; i < 6; i++) {
na = numaCreate(20);
numaaAddNuma(naa1, na, L_COPY);
numaaAddNuma(naa2, na, L_INSERT);
}
pixGetDimensions(pixs, &w, &h, NULL);
for (i = 0; i < 20; i++) {
rwhite = 100 + 5 * i;
gwhite = 200 - 5 * i;
bwhite = 150;
pixt0 = pixGlobalNormRGB(NULL, pixs, rwhite, gwhite, bwhite, 255);
pixaAddPix(pixa, pixt0, L_INSERT);
pixt1 = pixColorMagnitude(pixs, rwhite, gwhite, bwhite,
L_MAX_DIFF_FROM_AVERAGE_2);
for (j = 0; j < 6; j++) {
pixt2 = pixThresholdToBinary(pixt1, 30 + 10 * j);
pixInvert(pixt2, pixt2);
pixCountPixels(pixt2, &count, NULL);
na = numaaGetNuma(naa1, j, L_CLONE);
numaAddNumber(na, (l_float32)count / (l_float32)(w * h));
numaDestroy(&na);
pixDestroy(&pixt2);
}
pixDestroy(&pixt1);
pixt1 = pixColorMagnitude(pixs, rwhite, gwhite, bwhite,
L_MAX_MIN_DIFF_FROM_2);
for (j = 0; j < 6; j++) {
pixt2 = pixThresholdToBinary(pixt1, 30 + 10 * j);
pixInvert(pixt2, pixt2);
pixCountPixels(pixt2, &count, NULL);
na = numaaGetNuma(naa2, j, L_CLONE);
numaAddNumber(na, (l_float32)count / (l_float32)(w * h));
numaDestroy(&na);
pixDestroy(&pixt2);
}
pixDestroy(&pixt1);
}
gplot1 = gplotCreate("/tmp/junkplot1", GPLOT_X11,
"Fraction with given color (diff from average)",
"white point space for red", "amount of color");
gplot2 = gplotCreate("/tmp/junkplot2", GPLOT_X11,
"Fraction with given color (min diff)",
"white point space for red", "amount of color");
for (j = 0; j < 6; j++) {
na = numaaGetNuma(naa1, j, L_CLONE);
sprintf(label, "thresh %d", 30 + 10 * j);
gplotAddPlot(gplot1, naseq, na, GPLOT_LINES, label);
numaDestroy(&na);
na = numaaGetNuma(naa2, j, L_CLONE);
gplotAddPlot(gplot2, naseq, na, GPLOT_LINES, label);
numaDestroy(&na);
}
gplotMakeOutput(gplot1);
gplotMakeOutput(gplot2);
gplotDestroy(&gplot1);
gplotDestroy(&gplot2);
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, 250, 4, 0, 10, 2);
pixWrite("/tmp/junkcolormag", pixt1, IFF_PNG);
pixDisplayWithTitle(pixt1, 0, 100, "Color magnitude", 1);
pixDestroy(&pixt1);
pixaDestroy(&pixa);
numaDestroy(&naseq);
numaaDestroy(&naa1);
numaaDestroy(&naa2);
pixDisplayMultiple("/tmp/junk_write_display*");
pixDestroy(&pixs);
return 0;
}
/*!
* pixFindLargestRectangle()
*
* Input: pixs (1 bpp)
* polarity (0 within background, 1 within foreground)
* &box (<return> largest rectangle, either by area or
* by perimeter)
* debugflag (1 to output image with rectangle drawn on it)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) Why is this here? This is a simple and elegant solution to
* a problem in computational geometry that at first appears
* quite difficult: what is the largest rectangle that can
* be placed in the image, covering only pixels of one polarity
* (bg or fg)? The solution is O(n), where n is the number
* of pixels in the image, and it requires nothing more than
* using a simple recursion relation in a single sweep of the image.
* (2) In a sweep from UL to LR with left-to-right being the fast
* direction, calculate the largest white rectangle at (x, y),
* using previously calculated values at pixels #1 and #2:
* #1: (x, y - 1)
* #2: (x - 1, y)
* We also need the most recent "black" pixels that were seen
* in the current row and column.
* Consider the largest area. There are only two possibilities:
* (a) Min(w(1), horizdist) * (h(1) + 1)
* (b) Min(h(2), vertdist) * (w(2) + 1)
* where
* horizdist: the distance from the rightmost "black" pixel seen
* in the current row across to the current pixel
* vertdist: the distance from the lowest "black" pixel seen
* in the current column down to the current pixel
* and we choose the Max of (a) and (b).
* (3) To convince yourself that these recursion relations are correct,
* it helps to draw the maximum rectangles at #1 and #2.
* Then for #1, you try to extend the rectangle down one line,
* so that the height is h(1) + 1. Do you get the full
* width of #1, w(1)? It depends on where the black pixels are
* in the current row. You know the final width is bounded by w(1)
* and w(2) + 1, but the actual value depends on the distribution
* of black pixels in the current row that are at a distance
* from the current pixel that is between these limits.
* We call that value "horizdist", and the area is then given
* by the expression (a) above. Using similar reasoning for #2,
* where you attempt to extend the rectangle to the right
* by 1 pixel, you arrive at (b). The largest rectangle is
* then found by taking the Max.
*/
l_int32
pixFindLargestRectangle(PIX *pixs,
l_int32 polarity,
BOX **pbox,
const char *debugfile)
{
l_int32 i, j, w, h, d, wpls, val;
l_int32 wp, hp, w1, w2, h1, h2, wmin, hmin, area1, area2;
l_int32 xmax, ymax; /* LR corner of the largest rectangle */
l_int32 maxarea, wmax, hmax, vertdist, horizdist, prevfg;
l_int32 *lowestfg;
l_uint32 *datas, *lines;
l_uint32 **linew, **lineh;
BOX *box;
PIX *pixw, *pixh; /* keeps the width and height for the largest */
/* rectangles whose LR corner is located there. */
PROCNAME("pixFindLargestRectangle");
if (!pbox)
return ERROR_INT("&box not defined", procName, 1);
*pbox = NULL;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
pixGetDimensions(pixs, &w, &h, &d);
if (d != 1)
return ERROR_INT("pixs not 1 bpp", procName, 1);
if (polarity != 0 && polarity != 1)
return ERROR_INT("invalid polarity", procName, 1);
/* Initialize lowest "fg" seen so far for each column */
lowestfg = (l_int32 *)CALLOC(w, sizeof(l_int32));
for (i = 0; i < w; i++)
lowestfg[i] = -1;
/* The combination (val ^ polarity) is the color for which we
* are searching for the maximum rectangle. For polarity == 0,
* we search in the bg (white). */
pixw = pixCreate(w, h, 32); /* stores width */
pixh = pixCreate(w, h, 32); /* stores height */
linew = (l_uint32 **)pixGetLinePtrs(pixw, NULL);
lineh = (l_uint32 **)pixGetLinePtrs(pixh, NULL);
datas = pixGetData(pixs);
wpls = pixGetWpl(pixs);
maxarea = xmax = ymax = wmax = hmax = 0;
for (i = 0; i < h; i++) {
lines = datas + i * wpls;
prevfg = -1;
for (j = 0; j < w; j++) {
val = GET_DATA_BIT(lines, j);
if ((val ^ polarity) == 0) { /* bg (0) if polarity == 0, etc. */
if (i == 0 && j == 0) {
wp = hp = 1;
}
else if (i == 0) {
wp = linew[i][j - 1] + 1;
hp = 1;
}
else if (j == 0) {
wp = 1;
hp = lineh[i - 1][j] + 1;
}
else {
/* Expand #1 prev rectangle down */
w1 = linew[i - 1][j];
h1 = lineh[i - 1][j];
horizdist = j - prevfg;
wmin = L_MIN(w1, horizdist); /* width of new rectangle */
area1 = wmin * (h1 + 1);
/* Expand #2 prev rectangle to right */
w2 = linew[i][j - 1];
h2 = lineh[i][j - 1];
vertdist = i - lowestfg[j];
hmin = L_MIN(h2, vertdist); /* height of new rectangle */
area2 = hmin * (w2 + 1);
if (area1 > area2) {
wp = wmin;
hp = h1 + 1;
}
else {
wp = w2 + 1;
hp = hmin;
}
}
}
else { /* fg (1) if polarity == 0; bg (0) if polarity == 1 */
prevfg = j;
lowestfg[j] = i;
wp = hp = 0;
}
linew[i][j] = wp;
lineh[i][j] = hp;
if (wp * hp > maxarea) {
maxarea = wp * hp;
xmax = j;
ymax = i;
wmax = wp;
hmax = hp;
}
}
}
/* Translate from LR corner to Box coords (UL corner, w, h) */
box = boxCreate(xmax - wmax + 1, ymax - hmax + 1, wmax, hmax);
*pbox = box;
if (debugfile) {
PIX *pixdb;
pixdb = pixConvertTo8(pixs, TRUE);
pixRenderHashBoxArb(pixdb, box, 6, 2, L_NEG_SLOPE_LINE, 1, 255, 0, 0);
pixWrite(debugfile, pixdb, IFF_PNG);
pixDestroy(&pixdb);
}
FREE(linew);
FREE(lineh);
FREE(lowestfg);
pixDestroy(&pixw);
pixDestroy(&pixh);
return 0;
}
l_int32
DoPageSegmentation(PIX *pixs, /* should be at least 300 ppi */
l_int32 which) /* 1, 2, 3, 4 */
{
char buf[256];
l_int32 zero;
BOXA *boxatm, *boxahm;
PIX *pixr; /* image reduced to 150 ppi */
PIX *pixhs; /* image of halftone seed, 150 ppi */
PIX *pixm; /* image of mask of components, 150 ppi */
PIX *pixhm1; /* image of halftone mask, 150 ppi */
PIX *pixhm2; /* image of halftone mask, 300 ppi */
PIX *pixht; /* image of halftone components, 150 ppi */
PIX *pixnht; /* image without halftone components, 150 ppi */
PIX *pixi; /* inverted image, 150 ppi */
PIX *pixvws; /* image of vertical whitespace, 150 ppi */
PIX *pixm1; /* image of closed textlines, 150 ppi */
PIX *pixm2; /* image of refined text line mask, 150 ppi */
PIX *pixm3; /* image of refined text line mask, 300 ppi */
PIX *pixb1; /* image of text block mask, 150 ppi */
PIX *pixb2; /* image of text block mask, 300 ppi */
PIX *pixnon; /* image of non-text or halftone, 150 ppi */
PIX *pix1, *pix2, *pix3, *pix4;
PIXA *pixa;
PIXCMAP *cmap;
PTAA *ptaa;
l_int32 ht_flag = 0;
l_int32 ws_flag = 0;
l_int32 text_flag = 0;
l_int32 block_flag = 0;
PROCNAME("DoPageSegmentation");
if (which == 1)
ht_flag = 1;
else if (which == 2)
ws_flag = 1;
else if (which == 3)
text_flag = 1;
else if (which == 4)
block_flag = 1;
else
return ERROR_INT("invalid parameter: not in [1...4]", procName, 1);
pixa = pixaCreate(0);
lept_mkdir("lept/livre");
/* Reduce to 150 ppi */
pix1 = pixScaleToGray2(pixs);
if (ws_flag || ht_flag || block_flag) pixaAddPix(pixa, pix1, L_COPY);
if (which == 1)
pixWrite("/tmp/lept/livre/orig.gray.150.png", pix1, IFF_PNG);
pixDestroy(&pix1);
pixr = pixReduceRankBinaryCascade(pixs, 1, 0, 0, 0);
/* Get seed for halftone parts */
pix1 = pixReduceRankBinaryCascade(pixr, 4, 4, 3, 0);
pix2 = pixOpenBrick(NULL, pix1, 5, 5);
pixhs = pixExpandBinaryPower2(pix2, 8);
if (ht_flag) pixaAddPix(pixa, pixhs, L_COPY);
if (which == 1)
pixWrite("/tmp/lept/livre/htseed.150.png", pixhs, IFF_PNG);
pixDestroy(&pix1);
pixDestroy(&pix2);
/* Get mask for connected regions */
pixm = pixCloseSafeBrick(NULL, pixr, 4, 4);
if (ht_flag) pixaAddPix(pixa, pixm, L_COPY);
if (which == 1)
pixWrite("/tmp/lept/livre/ccmask.150.png", pixm, IFF_PNG);
/* Fill seed into mask to get halftone mask */
pixhm1 = pixSeedfillBinary(NULL, pixhs, pixm, 4);
if (ht_flag) pixaAddPix(pixa, pixhm1, L_COPY);
if (which == 1) pixWrite("/tmp/lept/livre/htmask.150.png", pixhm1, IFF_PNG);
pixhm2 = pixExpandBinaryPower2(pixhm1, 2);
/* Extract halftone stuff */
pixht = pixAnd(NULL, pixhm1, pixr);
if (which == 1) pixWrite("/tmp/lept/livre/ht.150.png", pixht, IFF_PNG);
/* Extract non-halftone stuff */
pixnht = pixXor(NULL, pixht, pixr);
if (text_flag) pixaAddPix(pixa, pixnht, L_COPY);
if (which == 1) pixWrite("/tmp/lept/livre/text.150.png", pixnht, IFF_PNG);
pixZero(pixht, &zero);
if (zero)
fprintf(stderr, "No halftone parts found\n");
else
fprintf(stderr, "Halftone parts found\n");
/* Get bit-inverted image */
pixi = pixInvert(NULL, pixnht);
if (ws_flag) pixaAddPix(pixa, pixi, L_COPY);
if (which == 1) pixWrite("/tmp/lept/livre/invert.150.png", pixi, IFF_PNG);
/* The whitespace mask will break textlines where there
* is a large amount of white space below or above.
* We can prevent this by identifying regions of the
* inverted image that have large horizontal (bigger than
* the separation between columns) and significant
* vertical extent (bigger than the separation between
* textlines), and subtracting this from the whitespace mask. */
pix1 = pixMorphCompSequence(pixi, "o80.60", 0);
pix2 = pixSubtract(NULL, pixi, pix1);
if (ws_flag) pixaAddPix(pixa, pix2, L_COPY);
pixDestroy(&pix1);
/* Identify vertical whitespace by opening inverted image */
pix3 = pixOpenBrick(NULL, pix2, 5, 1); /* removes thin vertical lines */
pixvws = pixOpenBrick(NULL, pix3, 1, 200); /* gets long vertical lines */
if (text_flag || ws_flag) pixaAddPix(pixa, pixvws, L_COPY);
if (which == 1) pixWrite("/tmp/lept/livre/vertws.150.png", pixvws, IFF_PNG);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* Get proto (early processed) text line mask. */
/* First close the characters and words in the textlines */
pixm1 = pixCloseSafeBrick(NULL, pixnht, 30, 1);
if (text_flag) pixaAddPix(pixa, pixm1, L_COPY);
if (which == 1)
pixWrite("/tmp/lept/livre/textmask1.150.png", pixm1, IFF_PNG);
/* Next open back up the vertical whitespace corridors */
pixm2 = pixSubtract(NULL, pixm1, pixvws);
if (which == 1)
pixWrite("/tmp/lept/livre/textmask2.150.png", pixm2, IFF_PNG);
/* Do a small opening to remove noise */
pixOpenBrick(pixm2, pixm2, 3, 3);
if (text_flag) pixaAddPix(pixa, pixm2, L_COPY);
if (which == 1)
pixWrite("/tmp/lept/livre/textmask3.150.png", pixm2, IFF_PNG);
pixm3 = pixExpandBinaryPower2(pixm2, 2);
/* Join pixels vertically to make text block mask */
pixb1 = pixMorphSequence(pixm2, "c1.10 + o4.1", 0);
if (block_flag) pixaAddPix(pixa, pixb1, L_COPY);
if (which == 1)
pixWrite("/tmp/lept/livre/textblock1.150.png", pixb1, IFF_PNG);
/* Solidify the textblock mask and remove noise:
* (1) For each c.c., close the blocks and dilate slightly
* to form a solid mask.
* (2) Small horizontal closing between components
* (3) Open the white space between columns, again
* (4) Remove small components */
pix1 = pixMorphSequenceByComponent(pixb1, "c30.30 + d3.3", 8, 0, 0, NULL);
pixCloseSafeBrick(pix1, pix1, 10, 1);
if (block_flag) pixaAddPix(pixa, pix1, L_COPY);
pix2 = pixSubtract(NULL, pix1, pixvws);
pix3 = pixSelectBySize(pix2, 25, 5, 8, L_SELECT_IF_BOTH,
L_SELECT_IF_GTE, NULL);
if (block_flag) pixaAddPix(pixa, pix3, L_COPY);
if (which == 1)
pixWrite("/tmp/lept/livre/textblock2.150.png", pix3, IFF_PNG);
pixb2 = pixExpandBinaryPower2(pix3, 2);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* Identify the outlines of each textblock */
ptaa = pixGetOuterBordersPtaa(pixb2);
pix1 = pixRenderRandomCmapPtaa(pixb2, ptaa, 1, 8, 1);
cmap = pixGetColormap(pix1);
pixcmapResetColor(cmap, 0, 130, 130, 130); /* set interior to gray */
if (which == 1)
pixWrite("/tmp/lept/livre/textblock3.300.png", pix1, IFF_PNG);
pixDisplayWithTitle(pix1, 480, 360, "textblock mask with outlines", DFLAG);
ptaaDestroy(&ptaa);
pixDestroy(&pix1);
/* Fill line mask (as seed) into the original */
pix1 = pixSeedfillBinary(NULL, pixm3, pixs, 8);
pixOr(pixm3, pixm3, pix1);
pixDestroy(&pix1);
if (which == 1)
pixWrite("/tmp/lept/livre/textmask.300.png", pixm3, IFF_PNG);
pixDisplayWithTitle(pixm3, 480, 360, "textline mask 4", DFLAG);
/* Fill halftone mask (as seed) into the original */
pix1 = pixSeedfillBinary(NULL, pixhm2, pixs, 8);
pixOr(pixhm2, pixhm2, pix1);
pixDestroy(&pix1);
if (which == 1)
pixWrite("/tmp/lept/livre/htmask.300.png", pixhm2, IFF_PNG);
pixDisplayWithTitle(pixhm2, 520, 390, "halftonemask 2", DFLAG);
/* Find objects that are neither text nor halftones */
pix1 = pixSubtract(NULL, pixs, pixm3); /* remove text pixels */
pixnon = pixSubtract(NULL, pix1, pixhm2); /* remove halftone pixels */
pixDestroy(&pix1);
if (which == 1)
pixWrite("/tmp/lept/livre/other.300.png", pixnon, IFF_PNG);
pixDisplayWithTitle(pixnon, 540, 420, "other stuff", DFLAG);
/* Write out b.b. for text line mask and halftone mask components */
boxatm = pixConnComp(pixm3, NULL, 4);
boxahm = pixConnComp(pixhm2, NULL, 8);
if (which == 1) {
boxaWrite("/tmp/lept/livre/textmask.boxa", boxatm);
boxaWrite("/tmp/lept/livre/htmask.boxa", boxahm);
}
pix1 = pixaDisplayTiledAndScaled(pixa, 8, 250, 4, 0, 25, 2);
pixDisplay(pix1, 0, 375 * (which - 1));
snprintf(buf, sizeof(buf), "/tmp/lept/livre/segout.%d.png", which);
pixWrite(buf, pix1, IFF_PNG);
pixDestroy(&pix1);
pixaDestroy(&pixa);
/* clean up to test with valgrind */
pixDestroy(&pixr);
pixDestroy(&pixhs);
pixDestroy(&pixm);
pixDestroy(&pixhm1);
pixDestroy(&pixhm2);
pixDestroy(&pixht);
pixDestroy(&pixi);
pixDestroy(&pixnht);
pixDestroy(&pixvws);
pixDestroy(&pixm1);
pixDestroy(&pixm2);
pixDestroy(&pixm3);
pixDestroy(&pixb1);
pixDestroy(&pixb2);
pixDestroy(&pixnon);
boxaDestroy(&boxatm);
boxaDestroy(&boxahm);
return 0;
}
main(int argc,
char **argv)
{
BOX *box;
PIX *pixs, *pixc, *pixp, *pixsel, *pixhmt;
PIX *pixd1, *pixd2, *pixd3;
SEL *selhm;
static char mainName[] = "findpattern3";
if (argc != 1)
exit(ERROR_INT(" Syntax: findpattern3", mainName, 1));
if ((pixs = pixRead("feyn.tif")) == NULL)
exit(ERROR_INT("pixs not made", mainName, 1));
/* -------------------------------------------- *
* Extract the pattern for a single character *
* ---------------------------------------------*/
box = boxCreate(599, 1055, 18, 23);
pixc = pixClipRectangle(pixs, box, NULL);
/* Make a hit-miss sel */
selhm = pixGenerateSelBoundary(pixc, 1, 2, 2, 2, 1, 1, 0, 0, &pixp);
/* Display the sel */
pixsel = pixDisplayHitMissSel(pixp, selhm, 7, HitColor, MissColor);
pixDisplay(pixsel, 200, 200);
pixWrite("/tmp/junkpixsel1", pixsel, IFF_PNG);
/* Use the Sel to find all instances in the page */
startTimer();
pixhmt = pixHMT(NULL, pixs, selhm);
fprintf(stderr, "Time to find patterns = %7.3f\n", stopTimer());
/* Color each instance at full res */
pixd1 = pixDisplayMatchedPattern(pixs, pixp, pixhmt, selhm->cx,
selhm->cy, 0x0000ff00, 1.0, 5);
pixWrite("/tmp/junkpixd11", pixd1, IFF_PNG);
/* Color each instance at 0.3 scale */
pixd2 = pixDisplayMatchedPattern(pixs, pixp, pixhmt, selhm->cx,
selhm->cy, 0x0000ff00, 0.5, 5);
pixWrite("/tmp/junkpixd12", pixd2, IFF_PNG);
/* Remove each instance from the input image */
pixd3 = pixCopy(NULL, pixs);
pixRemoveMatchedPattern(pixd3, pixp, pixhmt, selhm->cx,
selhm->cy, 1);
pixWrite("/tmp/junkpixr1", pixd3, IFF_PNG);
boxDestroy(&box);
selDestroy(&selhm);
pixDestroy(&pixc);
pixDestroy(&pixp);
pixDestroy(&pixsel);
pixDestroy(&pixhmt);
pixDestroy(&pixd1);
pixDestroy(&pixd2);
pixDestroy(&pixd3);
/* -------------------------------------------- *
* Extract the pattern for a word *
* ---------------------------------------------*/
box = boxCreate(208, 872, 130, 35);
pixc = pixClipRectangle(pixs, box, NULL);
/* Make a hit-miss sel */
selhm = pixGenerateSelBoundary(pixc, 2, 2, 1, 4, 1, 1, 0, 0, &pixp);
/* Display the sel */
pixsel = pixDisplayHitMissSel(pixp, selhm, 7, HitColor, MissColor);
pixDisplay(pixsel, 200, 200);
pixWrite("/tmp/junkpixsel2", pixsel, IFF_PNG);
/* Use the Sel to find all instances in the page */
startTimer();
pixhmt = pixHMT(NULL, pixs, selhm);
fprintf(stderr, "Time to find word patterns = %7.3f\n", stopTimer());
/* Color each instance at full res */
pixd1 = pixDisplayMatchedPattern(pixs, pixp, pixhmt, selhm->cx,
selhm->cy, 0x0000ff00, 1.0, 5);
pixWrite("/tmp/junkpixd21", pixd1, IFF_PNG);
/* Color each instance at 0.3 scale */
pixd2 = pixDisplayMatchedPattern(pixs, pixp, pixhmt, selhm->cx,
selhm->cy, 0x0000ff00, 0.5, 5);
pixWrite("/tmp/junkpixd22", pixd2, IFF_PNG);
/* Remove each instance from the input image */
pixd3 = pixCopy(NULL, pixs);
pixRemoveMatchedPattern(pixd3, pixp, pixhmt, selhm->cx,
selhm->cy, 1);
pixWrite("/tmp/junkpixr2", pixd3, IFF_PNG);
selDestroy(&selhm);
boxDestroy(&box);
pixDestroy(&pixc);
pixDestroy(&pixp);
pixDestroy(&pixsel);
pixDestroy(&pixhmt);
pixDestroy(&pixd1);
pixDestroy(&pixd2);
pixDestroy(&pixd3);
pixDestroy(&pixs);
return 0;
}
/*!
* pixFindBaselines()
*
* Input: pixs (1 bpp)
* &pta (<optional return> pairs of pts corresponding to
* approx. ends of each text line)
* debug (usually 0; set to 1 for debugging output)
* Return: na (of baseline y values), or null on error
*
* Notes:
* (1) Input binary image must have text lines already aligned
* horizontally. This can be done by either rotating the
* image with pixDeskew(), or, if a projective transform
* is required, by doing pixDeskewLocal() first.
* (2) Input null for &pta if you don't want this returned.
* The pta will come in pairs of points (left and right end
* of each baseline).
* (3) Caution: this will not work properly on text with multiple
* columns, where the lines are not aligned between columns.
* If there are multiple columns, they should be extracted
* separately before finding the baselines.
* (4) This function constructs different types of output
* for baselines; namely, a set of raster line values and
* a set of end points of each baseline.
* (5) This function was designed to handle short and long text lines
* without using dangerous thresholds on the peak heights. It does
* this by combining the differential signal with a morphological
* analysis of the locations of the text lines. One can also
* combine this data to normalize the peak heights, by weighting
* the differential signal in the region of each baseline
* by the inverse of the width of the text line found there.
* (6) There are various debug sections that can be turned on
* with the debug flag.
*/
NUMA *
pixFindBaselines(PIX *pixs,
PTA **ppta,
l_int32 debug)
{
l_int32 h, i, j, nbox, val1, val2, ndiff, bx, by, bw, bh;
l_int32 imaxloc, peakthresh, zerothresh, inpeak;
l_int32 mintosearch, max, maxloc, nloc, locval;
l_int32 *array;
l_float32 maxval;
BOXA *boxa1, *boxa2, *boxa3;
GPLOT *gplot;
NUMA *nasum, *nadiff, *naloc, *naval;
PIX *pixt1, *pixt2;
PTA *pta;
PROCNAME("pixFindBaselines");
if (!pixs)
return (NUMA *)ERROR_PTR("pixs not defined", procName, NULL);
pta = NULL;
if (ppta) {
pta = ptaCreate(0);
*ppta = pta;
}
lept_mkdir("/lept/baseline");
/* Close up the text characters, removing noise */
pixt1 = pixMorphSequence(pixs, "c25.1 + e3.1", 0);
/* Save the difference of adjacent row sums.
* The high positive-going peaks are the baselines */
if ((nasum = pixCountPixelsByRow(pixt1, NULL)) == NULL)
return (NUMA *)ERROR_PTR("nasum not made", procName, NULL);
h = pixGetHeight(pixs);
nadiff = numaCreate(h);
numaGetIValue(nasum, 0, &val2);
for (i = 0; i < h - 1; i++) {
val1 = val2;
numaGetIValue(nasum, i + 1, &val2);
numaAddNumber(nadiff, val1 - val2);
}
if (debug) /* show the difference signal */
gplotSimple1(nadiff, GPLOT_PNG, "/tmp/lept/baseline/diff", "Diff Sig");
/* Use the zeroes of the profile to locate each baseline. */
array = numaGetIArray(nadiff);
ndiff = numaGetCount(nadiff);
numaGetMax(nadiff, &maxval, &imaxloc);
/* Use this to begin locating a new peak: */
peakthresh = (l_int32)maxval / PEAK_THRESHOLD_RATIO;
/* Use this to begin a region between peaks: */
zerothresh = (l_int32)maxval / ZERO_THRESHOLD_RATIO;
naloc = numaCreate(0);
naval = numaCreate(0);
inpeak = FALSE;
for (i = 0; i < ndiff; i++) {
if (inpeak == FALSE) {
if (array[i] > peakthresh) { /* transition to in-peak */
inpeak = TRUE;
mintosearch = i + MIN_DIST_IN_PEAK; /* accept no zeros
* between i and mintosearch */
max = array[i];
maxloc = i;
}
} else { /* inpeak == TRUE; look for max */
if (array[i] > max) {
max = array[i];
maxloc = i;
mintosearch = i + MIN_DIST_IN_PEAK;
} else if (i > mintosearch && array[i] <= zerothresh) { /* leave */
inpeak = FALSE;
numaAddNumber(naval, max);
numaAddNumber(naloc, maxloc);
}
}
}
/* If array[ndiff-1] is max, eg. no descenders, baseline at bottom */
if (inpeak) {
numaAddNumber(naval, max);
numaAddNumber(naloc, maxloc);
}
LEPT_FREE(array);
if (debug) { /* show the raster locations for the peaks */
gplot = gplotCreate("/tmp/lept/baseline/loc", GPLOT_PNG, "Peak locs",
"rasterline", "height");
gplotAddPlot(gplot, naloc, naval, GPLOT_POINTS, "locs");
gplotMakeOutput(gplot);
gplotDestroy(&gplot);
}
/* Generate an approximate profile of text line width.
* First, filter the boxes of text, where there may be
* more than one box for a given textline. */
pixt2 = pixMorphSequence(pixt1, "r11 + c25.1 + o7.1 +c1.3", 0);
boxa1 = pixConnComp(pixt2, NULL, 4);
boxa2 = boxaTransform(boxa1, 0, 0, 4., 4.);
boxa3 = boxaSort(boxa2, L_SORT_BY_Y, L_SORT_INCREASING, NULL);
/* Then find the baseline segments */
if (pta) {
nloc = numaGetCount(naloc);
nbox = boxaGetCount(boxa3);
for (i = 0; i < nbox; i++) {
boxaGetBoxGeometry(boxa3, i, &bx, &by, &bw, &bh);
for (j = 0; j < nloc; j++) {
numaGetIValue(naloc, j, &locval);
if (L_ABS(locval - (by + bh)) > 25)
continue;
ptaAddPt(pta, bx, locval);
ptaAddPt(pta, bx + bw, locval);
break;
}
}
}
if (debug) { /* display baselines */
PIX *pixd;
l_int32 npts, x1, y1, x2, y2;
if (pta) {
pixd = pixConvertTo32(pixs);
npts = ptaGetCount(pta);
for (i = 0; i < npts; i += 2) {
ptaGetIPt(pta, i, &x1, &y1);
ptaGetIPt(pta, i + 1, &x2, &y2);
pixRenderLineArb(pixd, x1, y1, x2, y2, 1, 255, 0, 0);
}
pixDisplay(pixd, 200, 200);
pixWrite("/tmp/lept/baseline/baselines.png", pixd, IFF_PNG);
pixDestroy(&pixd);
}
}
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
numaDestroy(&nasum);
numaDestroy(&nadiff);
numaDestroy(&naval);
return naloc;
}
/*
* convertToPSEmbed()
*
* Input: filein (input image file -- any format)
* fileout (output ps file)
* level (compression: 1 (uncompressed), 2 or 3)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) This is a wrapper function that generates a PS file with
* a bounding box, from any input image file.
* (2) Do the best job of compression given the specified level.
* @level=3 does flate compression on anything that is not
* tiffg4 (1 bpp) or jpeg (8 bpp or rgb).
* (3) If @level=2 and the file is not tiffg4 or jpeg, it will
* first be written to file as jpeg with quality = 75.
* This will remove the colormap and cause some degradation
* in the image.
* (4) The bounding box is required when a program such as TeX
* (through epsf) places and rescales the image. It is
* sized for fitting the image to an 8.5 x 11.0 inch page.
*/
l_int32
convertToPSEmbed(const char *filein,
const char *fileout,
l_int32 level)
{
const char nametif[] = "/tmp/junk_convert_ps_embed.tif";
const char namejpg[] = "/tmp/junk_convert_ps_embed.jpg";
l_int32 d, format;
PIX *pix, *pixs;
PROCNAME("convertToPSEmbed");
if (!filein)
return ERROR_INT("filein not defined", procName, 1);
if (!fileout)
return ERROR_INT("fileout not defined", procName, 1);
if (level != 1 && level != 2 && level != 3) {
L_ERROR("invalid level specified; using level 2\n", procName);
level = 2;
}
if (level == 1) { /* no compression */
pixWritePSEmbed(filein, fileout);
return 0;
}
/* Find the format and write out directly if in jpeg or tiff g4 */
findFileFormat(filein, &format);
if (format == IFF_JFIF_JPEG) {
convertJpegToPSEmbed(filein, fileout);
return 0;
} else if (format == IFF_TIFF_G4) {
convertG4ToPSEmbed(filein, fileout);
return 0;
} else if (format == IFF_UNKNOWN) {
L_ERROR("format of %s not known\n", procName, filein);
return 1;
}
/* If level 3, flate encode. */
if (level == 3) {
convertFlateToPSEmbed(filein, fileout);
return 0;
}
/* OK, it's level 2, so we must convert to jpeg or tiff g4 */
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("image not read from file", procName, 1);
d = pixGetDepth(pixs);
if ((d == 2 || d == 4) && !pixGetColormap(pixs))
pix = pixConvertTo8(pixs, 0);
else if (d == 16)
pix = pixConvert16To8(pixs, 1);
else
pix = pixRemoveColormap(pixs, REMOVE_CMAP_BASED_ON_SRC);
d = pixGetDepth(pix);
if (d == 1) {
pixWrite(nametif, pix, IFF_TIFF_G4);
convertG4ToPSEmbed(nametif, fileout);
} else {
pixWrite(namejpg, pix, IFF_JFIF_JPEG);
convertJpegToPSEmbed(namejpg, fileout);
}
pixDestroy(&pix);
pixDestroy(&pixs);
return 0;
}
/*!
* pixaGenerateFont()
*
* Input: pix (of 95 characters in 3 rows)
* fontsize (4, 6, 8, ... , 20, in pts at 300 ppi)
* &bl1 (<return> baseline of row 1)
* &bl2 (<return> baseline of row 2)
* &bl3 (<return> baseline of row 3)
* Return: pixa of font bitmaps for 95 characters, or null on error
*
* Notes:
* (1) This does all the work. See pixaGenerateFontFromFile()
* for an overview.
* (2) The pix is for one of the 9 fonts. @fontsize is only
* used here for debugging.
*/
PIXA *
pixaGenerateFont(PIX *pixs,
l_int32 fontsize,
l_int32 *pbl0,
l_int32 *pbl1,
l_int32 *pbl2)
{
l_int32 i, j, nrows, nrowchars, nchars, h, yval;
l_int32 width, height;
l_int32 baseline[3];
l_int32 *tab = NULL;
BOX *box, *box1, *box2;
BOXA *boxar, *boxac, *boxacs;
PIX *pix1, *pix2, *pixr, *pixrc, *pixc;
PIXA *pixa;
l_int32 n, w, inrow, top;
l_int32 *ia;
NUMA *na;
PROCNAME("pixaGenerateFont");
if (!pbl0 || !pbl1 || !pbl2)
return (PIXA *)ERROR_PTR("&bl not all defined", procName, NULL);
*pbl0 = *pbl1 = *pbl2 = 0;
if (!pixs)
return (PIXA *)ERROR_PTR("pixs not defined", procName, NULL);
/* Locate the 3 rows of characters */
w = pixGetWidth(pixs);
na = pixCountPixelsByRow(pixs, NULL);
boxar = boxaCreate(0);
n = numaGetCount(na);
ia = numaGetIArray(na);
inrow = 0;
for (i = 0; i < n; i++) {
if (!inrow && ia[i] > 0) {
inrow = 1;
top = i;
} else if (inrow && ia[i] == 0) {
inrow = 0;
box = boxCreate(0, top, w, i - top);
boxaAddBox(boxar, box, L_INSERT);
}
}
FREE(ia);
numaDestroy(&na);
nrows = boxaGetCount(boxar);
#if DEBUG_FONT_GEN
L_INFO("For fontsize %s, have %d rows\n", procName, fontsize, nrows);
#endif /* DEBUG_FONT_GEN */
if (nrows != 3) {
L_INFO("nrows = %d; skipping fontsize %d\n", procName, nrows, fontsize);
return (PIXA *)ERROR_PTR("3 rows not generated", procName, NULL);
}
/* Grab the character images and baseline data */
#if DEBUG_BASELINE
lept_rmdir("baseline");
lept_mkdir("baseline");
#endif /* DEBUG_BASELINE */
tab = makePixelSumTab8();
pixa = pixaCreate(95);
for (i = 0; i < nrows; i++) {
box = boxaGetBox(boxar, i, L_CLONE);
pixr = pixClipRectangle(pixs, box, NULL); /* row of chars */
pixGetTextBaseline(pixr, tab, &yval);
baseline[i] = yval;
#if DEBUG_BASELINE
L_INFO("Baseline info: row %d, yval = %d, h = %d\n", procName,
i, yval, pixGetHeight(pixr));
pix1 = pixCopy(NULL, pixr);
pixRenderLine(pix1, 0, yval, pixGetWidth(pix1), yval, 1,
L_FLIP_PIXELS);
if (i == 0 )
pixWrite("/tmp/baseline/row0.png", pix1, IFF_PNG);
else if (i == 1)
pixWrite("/tmp/baseline/row1.png", pix1, IFF_PNG);
else
pixWrite("/tmp/baseline/row2.png", pix1, IFF_PNG);
pixDestroy(&pix1);
#endif /* DEBUG_BASELINE */
boxDestroy(&box);
pixrc = pixCloseSafeBrick(NULL, pixr, 1, 35);
boxac = pixConnComp(pixrc, NULL, 8);
boxacs = boxaSort(boxac, L_SORT_BY_X, L_SORT_INCREASING, NULL);
if (i == 0) { /* consolidate the two components of '"' */
box1 = boxaGetBox(boxacs, 1, L_CLONE);
box2 = boxaGetBox(boxacs, 2, L_CLONE);
box1->w = box2->x + box2->w - box1->x; /* increase width */
boxDestroy(&box1);
boxDestroy(&box2);
boxaRemoveBox(boxacs, 2);
}
h = pixGetHeight(pixr);
nrowchars = boxaGetCount(boxacs);
for (j = 0; j < nrowchars; j++) {
box = boxaGetBox(boxacs, j, L_COPY);
if (box->w <= 2 && box->h == 1) { /* skip 1x1, 2x1 components */
boxDestroy(&box);
continue;
}
box->y = 0;
box->h = h - 1;
pixc = pixClipRectangle(pixr, box, NULL);
boxDestroy(&box);
if (i == 0 && j == 0) /* add a pix for the space; change later */
pixaAddPix(pixa, pixc, L_COPY);
if (i == 2 && j == 0) /* add a pix for the '\'; change later */
pixaAddPix(pixa, pixc, L_COPY);
pixaAddPix(pixa, pixc, L_INSERT);
}
pixDestroy(&pixr);
pixDestroy(&pixrc);
boxaDestroy(&boxac);
boxaDestroy(&boxacs);
}
FREE(tab);
nchars = pixaGetCount(pixa);
if (nchars != 95)
return (PIXA *)ERROR_PTR("95 chars not generated", procName, NULL);
*pbl0 = baseline[0];
*pbl1 = baseline[1];
*pbl2 = baseline[2];
/* Fix the space character up; it should have no ON pixels,
* and be about twice as wide as the '!' character. */
pix1 = pixaGetPix(pixa, 0, L_CLONE);
width = 2 * pixGetWidth(pix1);
height = pixGetHeight(pix1);
pixDestroy(&pix1);
pix1 = pixCreate(width, height, 1);
pixaReplacePix(pixa, 0, pix1, NULL);
/* Fix up the '\' character; use a LR flip of the '/' char */
pix1 = pixaGetPix(pixa, 15, L_CLONE);
pix2 = pixFlipLR(NULL, pix1);
pixDestroy(&pix1);
pixaReplacePix(pixa, 60, pix2, NULL);
#if DEBUG_CHARS
pix1 = pixaDisplayTiled(pixa, 1500, 0, 10);
pixDisplay(pix1, 100 * i, 200);
pixDestroy(&pix1);
#endif /* DEBUG_CHARS */
boxaDestroy(&boxar);
return pixa;
}
main(int argc,
char **argv)
{
char *filein, *fileout;
l_int32 w, h, d, w2, h2, i, ncols;
l_float32 angle, conf;
BOX *box;
BOXA *boxa, *boxas, *boxad, *boxa2;
NUMA *numa;
PIX *pixs, *pixt, *pixb, *pixb2, *pixd;
PIX *pixtlm, *pixvws;
PIX *pixt1, *pixt2, *pixt3, *pixt4, *pixt5, *pixt6;
PIXA *pixam, *pixac, *pixad, *pixat;
PIXAA *pixaa, *pixaa2;
PTA *pta;
SEL *selsplit;
static char mainName[] = "textlinemask";
if (argc != 3)
exit(ERROR_INT(" Syntax: textlinemask filein fileout", mainName, 1));
filein = argv[1];
fileout = argv[2];
pixDisplayWrite(NULL, -1); /* init debug output */
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
pixGetDimensions(pixs, &w, &h, &d);
/* Binarize input */
if (d == 8)
pixt = pixThresholdToBinary(pixs, 128);
else if (d == 1)
pixt = pixClone(pixs);
else {
fprintf(stderr, "depth is %d\n", d);
exit(1);
}
/* Deskew */
pixb = pixFindSkewAndDeskew(pixt, 1, &angle, &conf);
pixDestroy(&pixt);
fprintf(stderr, "Skew angle: %7.2f degrees; %6.2f conf\n", angle, conf);
pixDisplayWrite(pixb, DEBUG_OUTPUT);
#if 1
/* Use full image morphology to find columns, at 2x reduction.
* This only works for very simple layouts where each column
* of text extends the full height of the input image.
* pixam has a pix component over each column. */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixt1 = pixMorphCompSequence(pixb2, "c5.500", 0);
boxa = pixConnComp(pixt1, &pixam, 8);
ncols = boxaGetCount(boxa);
fprintf(stderr, "Num columns: %d\n", ncols);
pixDisplayWrite(pixt1, DEBUG_OUTPUT);
/* Use selective region-based morphology to get the textline mask. */
pixad = pixaMorphSequenceByRegion(pixb2, pixam, "c100.3", 0, 0);
pixGetDimensions(pixb2, &w2, &h2, NULL);
if (DEBUG_OUTPUT) {
pixt2 = pixaDisplay(pixad, w2, h2);
pixDisplayWrite(pixt2, DEBUG_OUTPUT);
pixDestroy(&pixt2);
}
/* Some of the lines may be touching, so use a HMT to split the
* lines in each column, and use a pixaa to save the results. */
selsplit = selCreateFromString(seltext, 17, 7, "selsplit");
pixaa = pixaaCreate(ncols);
for (i = 0; i < ncols; i++) {
pixt3 = pixaGetPix(pixad, i, L_CLONE);
box = pixaGetBox(pixad, i, L_COPY);
pixt4 = pixHMT(NULL, pixt3, selsplit);
pixXor(pixt4, pixt4, pixt3);
boxa2 = pixConnComp(pixt4, &pixac, 8);
pixaaAddPixa(pixaa, pixac, L_INSERT);
pixaaAddBox(pixaa, box, L_INSERT);
if (DEBUG_OUTPUT) {
pixt5 = pixaDisplayRandomCmap(pixac, 0, 0);
pixDisplayWrite(pixt5, DEBUG_OUTPUT);
fprintf(stderr, "Num textlines in col %d: %d\n", i,
boxaGetCount(boxa2));
pixDestroy(&pixt5);
}
pixDestroy(&pixt3);
pixDestroy(&pixt4);
boxaDestroy(&boxa2);
}
/* Visual output */
if (DEBUG_OUTPUT) {
pixDisplayMultiple("/tmp/junk_write_display*");
pixat = pixaReadFiles("/tmp", "junk_write_display");
pixt5 = selDisplayInPix(selsplit, 31, 2);
pixaAddPix(pixat, pixt5, L_INSERT);
pixt6 = pixaDisplayTiledAndScaled(pixat, 32, 400, 3, 0, 35, 3);
pixWrite(fileout, pixt6, IFF_PNG);
pixaDestroy(&pixat);
pixDestroy(&pixt6);
}
/* Test pixaa I/O */
pixaaWrite("/tmp/junkpixaa", pixaa);
pixaa2 = pixaaRead("/tmp/junkpixaa");
pixaaWrite("/tmp/junkpixaa2", pixaa2);
/* Test pixaa display */
pixd = pixaaDisplay(pixaa, w2, h2);
pixWrite("/tmp/junkdisplay", pixd, IFF_PNG);
pixDestroy(&pixd);
/* Cleanup */
pixDestroy(&pixb2);
pixDestroy(&pixt1);
pixaDestroy(&pixam);
pixaDestroy(&pixad);
pixaaDestroy(&pixaa);
pixaaDestroy(&pixaa2);
boxaDestroy(&boxa);
selDestroy(&selsplit);
#endif
#if 0
/* Use the baseline finder; not really what is needed */
numa = pixFindBaselines(pixb, &pta, 1);
#endif
#if 0
/* Use the textline mask function; parameters are not quite right */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixtlm = pixGenTextlineMask(pixb2, &pixvws, NULL, 1);
pixDisplay(pixtlm, 0, 100);
pixDisplay(pixvws, 500, 100);
pixDestroy(&pixb2);
pixDestroy(&pixtlm);
pixDestroy(&pixvws);
#endif
#if 0
/* Use the Breuel whitespace partition method; slow and we would
* still need to work to extract the fg regions. */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
boxas = pixConnComp(pixb2, NULL, 8);
boxad = boxaGetWhiteblocks(boxas, NULL, L_SORT_BY_HEIGHT,
3, 0.1, 200, 0.2, 0);
pixd = pixDrawBoxa(pixb2, boxad, 7, 0xe0708000);
pixDisplay(pixd, 100, 500);
pixDestroy(&pixb2);
pixDestroy(&pixd);
boxaDestroy(&boxas);
boxaDestroy(&boxad);
#endif
#if 0
/* Use morphology to find columns and then selective
* region-based morphology to get the textline mask.
* This is for display; we really want to get a pixa of the
* specific textline masks. */
startTimer();
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixt1 = pixMorphCompSequence(pixb2, "c5.500", 0); /* column mask */
pixt2 = pixMorphSequenceByRegion(pixb2, pixt1, "c100.3", 8, 0, 0, &boxa);
fprintf(stderr, "time = %7.3f sec\n", stopTimer());
pixDisplay(pixt1, 100, 500);
pixDisplay(pixt2, 800, 500);
pixDestroy(&pixb2);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
boxaDestroy(&boxa);
#endif
pixDestroy(&pixs);
pixDestroy(&pixb);
exit(0);
}
int main(int argc,
char **argv)
{
char buffer[512];
char *tempfile1, *tempfile2;
l_uint8 *data;
l_int32 i, j, w, h, seq, ret, same;
size_t nbytes;
const char *title;
BOX *box;
BOXA *boxa1, *boxa2;
L_BYTEA *ba;
L_PDF_DATA *lpd;
PIX *pix1, *pix2, *pix3, *pix4, *pix5, *pix6;
PIX *pixs, *pixt, *pixg, *pixgc, *pixc;
static char mainName[] = "pdfiotest";
if (argc != 1)
return ERROR_INT("syntax: pdfiotest", mainName, 1);
l_pdfSetDateAndVersion(0);
lept_mkdir("lept/pdf");
#if 1
/* --------------- Single image tests ------------------- */
fprintf(stderr, "\n*** Writing single images as pdf files\n");
convertToPdf("weasel2.4c.png", L_FLATE_ENCODE, 0, "/tmp/lept/pdf/file01.pdf",
0, 0, 72, "weasel2.4c.png", NULL, 0);
convertToPdf("test24.jpg", L_JPEG_ENCODE, 0, "/tmp/lept/pdf/file02.pdf",
0, 0, 72, "test24.jpg", NULL, 0);
convertToPdf("feyn.tif", L_G4_ENCODE, 0, "/tmp/lept/pdf/file03.pdf",
0, 0, 300, "feyn.tif", NULL, 0);
pixs = pixRead("feyn.tif");
pixConvertToPdf(pixs, L_G4_ENCODE, 0, "/tmp/lept/pdf/file04.pdf", 0, 0, 300,
"feyn.tif", NULL, 0);
pixDestroy(&pixs);
pixs = pixRead("test24.jpg");
pixConvertToPdf(pixs, L_JPEG_ENCODE, 5, "/tmp/lept/pdf/file05.pdf",
0, 0, 72, "test24.jpg", NULL, 0);
pixDestroy(&pixs);
pixs = pixRead("feyn.tif");
pixt = pixScaleToGray2(pixs);
pixWrite("/tmp/lept/pdf/feyn8.png", pixt, IFF_PNG);
convertToPdf("/tmp/lept/pdf/feyn8.png", L_JPEG_ENCODE, 0,
"/tmp/lept/pdf/file06.pdf", 0, 0, 150, "feyn8.png", NULL, 0);
pixDestroy(&pixs);
pixDestroy(&pixt);
convertToPdf("weasel4.16g.png", L_FLATE_ENCODE, 0,
"/tmp/lept/pdf/file07.pdf", 0, 0, 30,
"weasel4.16g.png", NULL, 0);
pixs = pixRead("test24.jpg");
pixg = pixConvertTo8(pixs, 0);
box = boxCreate(100, 100, 100, 100);
pixc = pixClipRectangle(pixs, box, NULL);
pixgc = pixClipRectangle(pixg, box, NULL);
pixWrite("/tmp/lept/pdf/pix32.jpg", pixc, IFF_JFIF_JPEG);
pixWrite("/tmp/lept/pdf/pix8.jpg", pixgc, IFF_JFIF_JPEG);
convertToPdf("/tmp/lept/pdf/pix32.jpg", L_FLATE_ENCODE, 0,
"/tmp/lept/pdf/file08.pdf", 0, 0, 72, "pix32.jpg", NULL, 0);
convertToPdf("/tmp/lept/pdf/pix8.jpg", L_FLATE_ENCODE, 0,
"/tmp/lept/pdf/file09.pdf", 0, 0, 72, "pix8.jpg", NULL, 0);
pixDestroy(&pixs);
pixDestroy(&pixg);
pixDestroy(&pixc);
pixDestroy(&pixgc);
boxDestroy(&box);
#endif
#if 1
/* --------------- Multiple image tests ------------------- */
fprintf(stderr, "\n*** Writing multiple images as single page pdf files\n");
pix1 = pixRead("feyn-fract.tif");
pix2 = pixRead("weasel8.240c.png");
/* l_pdfSetDateAndVersion(0); */
/* First, write the 1 bpp image through the mask onto the weasels */
for (i = 0; i < 5; i++) {
for (j = 0; j < 10; j++) {
seq = (i == 0 && j == 0) ? L_FIRST_IMAGE : L_NEXT_IMAGE;
title = (i == 0 && j == 0) ? "feyn-fract.tif" : NULL;
pixConvertToPdf(pix2, L_FLATE_ENCODE, 0, NULL, 100 * j,
100 * i, 70, title, &lpd, seq);
}
}
pixConvertToPdf(pix1, L_G4_ENCODE, 0, "/tmp/lept/pdf/file10.pdf", 0, 0, 80,
NULL, &lpd, L_LAST_IMAGE);
/* Now, write the 1 bpp image over the weasels */
l_pdfSetG4ImageMask(0);
for (i = 0; i < 5; i++) {
for (j = 0; j < 10; j++) {
seq = (i == 0 && j == 0) ? L_FIRST_IMAGE : L_NEXT_IMAGE;
title = (i == 0 && j == 0) ? "feyn-fract.tif" : NULL;
pixConvertToPdf(pix2, L_FLATE_ENCODE, 0, NULL, 100 * j,
100 * i, 70, title, &lpd, seq);
}
}
pixConvertToPdf(pix1, L_G4_ENCODE, 0, "/tmp/lept/pdf/file11.pdf", 0, 0, 80,
NULL, &lpd, L_LAST_IMAGE);
l_pdfSetG4ImageMask(1);
pixDestroy(&pix1);
pixDestroy(&pix2);
#endif
#if 1
/* -------- pdf convert segmented with no image regions -------- */
fprintf(stderr, "\n*** Writing segmented images without image regions\n");
pix1 = pixRead("rabi.png");
pix2 = pixScaleToGray2(pix1);
pixWrite("/tmp/lept/pdf/rabi8.jpg", pix2, IFF_JFIF_JPEG);
pix3 = pixThresholdTo4bpp(pix2, 16, 1);
pixWrite("/tmp/lept/pdf/rabi4.png", pix3, IFF_PNG);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* 1 bpp input */
convertToPdfSegmented("rabi.png", 300, L_G4_ENCODE, 128, NULL, 0, 0,
NULL, "/tmp/lept/pdf/file12.pdf");
convertToPdfSegmented("rabi.png", 300, L_JPEG_ENCODE, 128, NULL, 0, 0,
NULL, "/tmp/lept/pdf/file13.pdf");
convertToPdfSegmented("rabi.png", 300, L_FLATE_ENCODE, 128, NULL, 0, 0,
NULL, "/tmp/lept/pdf/file14.pdf");
/* 8 bpp input, no cmap */
convertToPdfSegmented("/tmp/lept/pdf/rabi8.jpg", 150, L_G4_ENCODE, 128,
NULL, 0, 0, NULL, "/tmp/lept/pdf/file15.pdf");
convertToPdfSegmented("/tmp/lept/pdf/rabi8.jpg", 150, L_JPEG_ENCODE, 128,
NULL, 0, 0, NULL, "/tmp/lept/pdf/file16.pdf");
convertToPdfSegmented("/tmp/lept/pdf/rabi8.jpg", 150, L_FLATE_ENCODE, 128,
NULL, 0, 0, NULL, "/tmp/lept/pdf/file17.pdf");
/* 4 bpp input, cmap */
convertToPdfSegmented("/tmp/lept/pdf/rabi4.png", 150, L_G4_ENCODE, 128,
NULL, 0, 0, NULL, "/tmp/lept/pdf/file18.pdf");
convertToPdfSegmented("/tmp/lept/pdf/rabi4.png", 150, L_JPEG_ENCODE, 128,
NULL, 0, 0, NULL, "/tmp/lept/pdf/file19.pdf");
convertToPdfSegmented("/tmp/lept/pdf/rabi4.png", 150, L_FLATE_ENCODE, 128,
NULL, 0, 0, NULL, "/tmp/lept/pdf/file20.pdf");
#endif
#if 1
/* ---------- pdf convert segmented with image regions ---------- */
fprintf(stderr, "\n*** Writing segmented images with image regions\n");
/* Get the image region(s) for rabi.png. There are two
* small bogus regions at the top, but we'll keep them for
* the demonstration. */
pix1 = pixRead("rabi.png");
pixSetResolution(pix1, 300, 300);
pixGetDimensions(pix1, &w, &h, NULL);
pix2 = pixGenerateHalftoneMask(pix1, NULL, NULL, NULL);
pix3 = pixMorphSequence(pix2, "c20.1 + c1.20", 0);
boxa1 = pixConnComp(pix3, NULL, 8);
boxa2 = boxaTransform(boxa1, 0, 0, 0.5, 0.5);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* 1 bpp input */
convertToPdfSegmented("rabi.png", 300, L_G4_ENCODE, 128, boxa1,
0, 0.25, NULL, "/tmp/lept/pdf/file21.pdf");
convertToPdfSegmented("rabi.png", 300, L_JPEG_ENCODE, 128, boxa1,
0, 0.25, NULL, "/tmp/lept/pdf/file22.pdf");
convertToPdfSegmented("rabi.png", 300, L_FLATE_ENCODE, 128, boxa1,
0, 0.25, NULL, "/tmp/lept/pdf/file23.pdf");
/* 8 bpp input, no cmap */
convertToPdfSegmented("/tmp/lept/pdf/rabi8.jpg", 150, L_G4_ENCODE, 128,
boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file24.pdf");
convertToPdfSegmented("/tmp/lept/pdf/rabi8.jpg", 150, L_JPEG_ENCODE, 128,
boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file25.pdf");
convertToPdfSegmented("/tmp/lept/pdf/rabi8.jpg", 150, L_FLATE_ENCODE, 128,
boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file26.pdf");
/* 4 bpp input, cmap */
convertToPdfSegmented("/tmp/lept/pdf/rabi4.png", 150, L_G4_ENCODE, 128,
boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file27.pdf");
convertToPdfSegmented("/tmp/lept/pdf/rabi4.png", 150, L_JPEG_ENCODE, 128,
boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file28.pdf");
convertToPdfSegmented("/tmp/lept/pdf/rabi4.png", 150, L_FLATE_ENCODE, 128,
boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file29.pdf");
/* 4 bpp input, cmap, data output */
data = NULL;
convertToPdfDataSegmented("/tmp/lept/pdf/rabi4.png", 150, L_G4_ENCODE,
128, boxa2, 0, 0.5, NULL, &data, &nbytes);
l_binaryWrite("/tmp/lept/pdf/file30.pdf", "w", data, nbytes);
lept_free(data);
convertToPdfDataSegmented("/tmp/lept/pdf/rabi4.png", 150, L_JPEG_ENCODE,
128, boxa2, 0, 0.5, NULL, &data, &nbytes);
l_binaryWrite("/tmp/lept/pdf/file31.pdf", "w", data, nbytes);
lept_free(data);
convertToPdfDataSegmented("/tmp/lept/pdf/rabi4.png", 150, L_FLATE_ENCODE,
128, boxa2, 0, 0.5, NULL, &data, &nbytes);
l_binaryWrite("/tmp/lept/pdf/file32.pdf", "w", data, nbytes);
lept_free(data);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
#endif
#if 1
/* -------- pdf convert segmented from color image -------- */
fprintf(stderr, "\n*** Writing color segmented images\n");
pix1 = pixRead("candelabrum.011.jpg");
pix2 = pixScale(pix1, 3.0, 3.0);
pixWrite("/tmp/lept/pdf/candelabrum3.jpg", pix2, IFF_JFIF_JPEG);
GetImageMask(pix2, 200, &boxa1, "/tmp/lept/pdf/seg1.jpg");
convertToPdfSegmented("/tmp/lept/pdf/candelabrum3.jpg", 200, L_G4_ENCODE,
100, boxa1, 0, 0.25, NULL,
"/tmp/lept/pdf/file33.pdf");
convertToPdfSegmented("/tmp/lept/pdf/candelabrum3.jpg", 200, L_JPEG_ENCODE,
100, boxa1, 0, 0.25, NULL,
"/tmp/lept/pdf/file34.pdf");
convertToPdfSegmented("/tmp/lept/pdf/candelabrum3.jpg", 200, L_FLATE_ENCODE,
100, boxa1, 0, 0.25, NULL,
"/tmp/lept/pdf/file35.pdf");
pixDestroy(&pix1);
pixDestroy(&pix2);
boxaDestroy(&boxa1);
pix1 = pixRead("lion-page.00016.jpg");
pix2 = pixScale(pix1, 3.0, 3.0);
pixWrite("/tmp/lept/pdf/lion16.jpg", pix2, IFF_JFIF_JPEG);
pix3 = pixRead("lion-mask.00016.tif");
boxa1 = pixConnComp(pix3, NULL, 8);
boxa2 = boxaTransform(boxa1, 0, 0, 3.0, 3.0);
convertToPdfSegmented("/tmp/lept/pdf/lion16.jpg", 200, L_G4_ENCODE,
190, boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file36.pdf");
convertToPdfSegmented("/tmp/lept/pdf/lion16.jpg", 200, L_JPEG_ENCODE,
190, boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file37.pdf");
convertToPdfSegmented("/tmp/lept/pdf/lion16.jpg", 200, L_FLATE_ENCODE,
190, boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file38.pdf");
/* Quantize the non-image part and flate encode.
* This is useful because it results in a smaller file than
* when you flate-encode the un-quantized non-image regions. */
pix4 = pixScale(pix3, 3.0, 3.0); /* higher res mask, for combining */
pix5 = QuantizeNonImageRegion(pix2, pix4, 12);
pixWrite("/tmp/lept/pdf/lion16-quant.png", pix5, IFF_PNG);
convertToPdfSegmented("/tmp/lept/pdf/lion16-quant.png", 200, L_FLATE_ENCODE,
190, boxa2, 0, 0.5, NULL, "/tmp/lept/pdf/file39.pdf");
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
pixDestroy(&pix5);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
#endif
#if 1
/* ------------------ Test multipage pdf generation ----------------- */
fprintf(stderr, "\n*** Writing multipage pdfs from single page pdfs\n");
/* Generate a multi-page pdf from all these files */
startTimer();
concatenatePdf("/tmp/lept/pdf", "file", "/tmp/lept/pdf/cat_lept.pdf");
fprintf(stderr,
"All files have been concatenated: /tmp/lept/pdf/cat_lept.pdf\n"
"Concatenation time: %7.3f\n", stopTimer());
#endif
#if 1
/* ----------- Test corruption recovery by concatenation ------------ */
/* Put two good pdf files in a directory */
lept_rmdir("lept/good");
lept_mkdir("lept/good");
lept_cp("testfile1.pdf", "lept/good", NULL, NULL);
lept_cp("testfile2.pdf", "lept/good", NULL, NULL);
concatenatePdf("/tmp/lept/good", "file", "/tmp/lept/pdf/good.pdf");
/* Make a bad version with the pdf id removed, so that it is not
* recognized as a pdf */
lept_rmdir("lept/bad");
lept_mkdir("lept/bad");
ba = l_byteaInitFromFile("testfile2.pdf");
data = l_byteaGetData(ba, &nbytes);
l_binaryWrite("/tmp/lept/bad/testfile0.notpdf.pdf", "w",
data + 10, nbytes - 10);
/* Make a version with a corrupted trailer */
if (data)
data[2297] = '2'; /* munge trailer object 6: change 458 --> 428 */
l_binaryWrite("/tmp/lept/bad/testfile2.bad.pdf", "w", data, nbytes);
l_byteaDestroy(&ba);
/* Copy testfile1.pdf to the /tmp/lept/bad directory. Then
* run concat on the bad files. The "not pdf" file should be
* ignored, and the corrupted pdf file should be properly parsed,
* so the resulting concatenated pdf files should be identical. */
fprintf(stderr, "\nWe attempt to build from the bad directory\n");
lept_cp("testfile1.pdf", "lept/bad", NULL, NULL);
concatenatePdf("/tmp/lept/bad", "file", "/tmp/lept/pdf/bad.pdf");
filesAreIdentical("/tmp/lept/pdf/good.pdf", "/tmp/lept/pdf/bad.pdf", &same);
if (same)
fprintf(stderr, "Fixed: files are the same\n"
"Attempt succeeded\n");
else
fprintf(stderr, "Busted: files are different\n");
#endif
#if 0
fprintf(stderr, "\n*** pdftk writes multipage pdfs from images\n");
tempfile1 = genPathname("/tmp/lept/pdf", "file*.pdf");
tempfile2 = genPathname("/tmp/lept/pdf", "cat_pdftk.pdf");
snprintf(buffer, sizeof(buffer), "pdftk %s output %s",
tempfile1, tempfile2);
ret = system(buffer); /* pdftk */
lept_free(tempfile1);
lept_free(tempfile2);
#endif
#if 1
/* -- Test simple interface for generating multi-page pdf from images -- */
fprintf(stderr, "\n*** Writing multipage pdfs from images\n");
/* Put four image files in a directory. They will be encoded thus:
* file1.png: flate (8 bpp, only 10 colors)
* file2.jpg: dct (8 bpp, 256 colors because of the jpeg encoding)
* file3.tif: g4 (1 bpp)
* file4.jpg: dct (32 bpp) */
lept_mkdir("lept/image");
pix1 = pixRead("feyn.tif");
pix2 = pixRead("rabi.png");
pix3 = pixScaleToGray3(pix1);
pix4 = pixScaleToGray3(pix2);
pix5 = pixScale(pix1, 0.33, 0.33);
pix6 = pixRead("test24.jpg");
pixWrite("/tmp/lept/image/file1.png", pix3, IFF_PNG); /* 10 colors */
pixWrite("/tmp/lept/image/file2.jpg", pix4, IFF_JFIF_JPEG); /* 256 colors */
pixWrite("/tmp/lept/image/file3.tif", pix5, IFF_TIFF_G4);
pixWrite("/tmp/lept/image/file4.jpg", pix6, IFF_JFIF_JPEG);
startTimer();
convertFilesToPdf("/tmp/lept/image", "file", 100, 0.8, 0, 75, "4 file test",
"/tmp/lept/pdf/fourimages.pdf");
fprintf(stderr, "4-page pdf generated: /tmp/lept/pdf/fourimages.pdf\n"
"Time: %7.3f\n", stopTimer());
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
pixDestroy(&pix5);
pixDestroy(&pix6);
#endif
return 0;
}
l_int32
DoComparisonDwa2(PIX *pixs,
PIX *pixt1,
PIX *pixt2,
PIX *pixt3,
PIX *pixt4,
PIX *pixt5,
PIX *pixt6,
l_int32 size) /* exactly decomposable */
{
fprintf(stderr, "..%d..", size);
if (TIMING) startTimer();
pixDilateCompBrickExtendDwa(pixt1, pixs, size, 1);
pixDilateCompBrickExtendDwa(pixt3, pixs, 1, size);
pixDilateCompBrickExtendDwa(pixt5, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Dwa: %7.3f sec\n", stopTimer());
if (TIMING) startTimer();
pixDilateBrick(pixt2, pixs, size, 1);
pixDilateBrick(pixt4, pixs, 1, size);
pixDilateBrick(pixt6, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Rop: %7.3f sec\n", stopTimer());
PixCompareDwa(size, "dilate", pixt1, pixt2, pixt3, pixt4, pixt5, pixt6);
if (TIMING) startTimer();
pixErodeCompBrickExtendDwa(pixt1, pixs, size, 1);
pixErodeCompBrickExtendDwa(pixt3, pixs, 1, size);
pixErodeCompBrickExtendDwa(pixt5, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Dwa: %7.3f sec\n", stopTimer());
if (TIMING) startTimer();
pixErodeBrick(pixt2, pixs, size, 1);
pixErodeBrick(pixt4, pixs, 1, size);
pixErodeBrick(pixt6, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Rop: %7.3f sec\n", stopTimer());
PixCompareDwa(size, "erode", pixt1, pixt2, pixt3, pixt4, pixt5, pixt6);
if (TIMING) startTimer();
pixOpenCompBrickExtendDwa(pixt1, pixs, size, 1);
pixOpenCompBrickExtendDwa(pixt3, pixs, 1, size);
pixOpenCompBrickExtendDwa(pixt5, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Dwa: %7.3f sec\n", stopTimer());
if (TIMING) startTimer();
pixOpenBrick(pixt2, pixs, size, 1);
pixOpenBrick(pixt4, pixs, 1, size);
pixOpenBrick(pixt6, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Rop: %7.3f sec\n", stopTimer());
PixCompareDwa(size, "open", pixt1, pixt2, pixt3, pixt4, pixt5, pixt6);
if (TIMING) startTimer();
pixCloseCompBrickExtendDwa(pixt1, pixs, size, 1);
pixCloseCompBrickExtendDwa(pixt3, pixs, 1, size);
pixCloseCompBrickExtendDwa(pixt5, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Dwa: %7.3f sec\n", stopTimer());
if (TIMING) startTimer();
pixCloseSafeBrick(pixt2, pixs, size, 1);
pixCloseSafeBrick(pixt4, pixs, 1, size);
pixCloseSafeBrick(pixt6, pixs, size, size);
if (TIMING) fprintf(stderr, "Time Rop: %7.3f sec\n", stopTimer());
PixCompareDwa(size, "close", pixt1, pixt2, pixt3, pixt4, pixt5, pixt6);
#if 0
pixWrite("/tmp/junkpixt3.png", pixt3, IFF_PNG);
pixWrite("/tmp/junkpixt4.png", pixt4, IFF_PNG);
pixXor(pixt3, pixt3, pixt4);
pixWrite("/tmp/junkxor.png", pixt3, IFF_PNG);
#endif
return 0;
}
/**
* Sets up auto page segmentation, determines the orientation, and corrects it.
* Somewhat arbitrary chunk of functionality, factored out of AutoPageSeg to
* facilitate testing.
* photo_mask_pix is a pointer to a NULL pointer that will be filled on return
* with the leptonica photo mask, which must be pixDestroyed by the caller.
* to_blocks is an empty list that will be filled with (usually a single)
* block that is used during layout analysis. This ugly API is required
* because of the possibility of a unlv zone file.
* TODO(rays) clean this up.
* See AutoPageSeg for other arguments.
* The returned ColumnFinder must be deleted after use.
*/
ColumnFinder* Tesseract::SetupPageSegAndDetectOrientation(
bool single_column, bool osd, bool only_osd,
BLOCK_LIST* blocks, Tesseract* osd_tess, OSResults* osr,
TO_BLOCK_LIST* to_blocks, Pix** photo_mask_pix, Pix** music_mask_pix) {
int vertical_x = 0;
int vertical_y = 1;
TabVector_LIST v_lines;
TabVector_LIST h_lines;
ICOORD bleft(0, 0);
ASSERT_HOST(pix_binary_ != NULL);
if (tessedit_dump_pageseg_images) {
pixWrite("tessinput.png", pix_binary_, IFF_PNG);
}
// Leptonica is used to find the rule/separator lines in the input.
LineFinder::FindAndRemoveLines(source_resolution_,
textord_tabfind_show_vlines, pix_binary_,
&vertical_x, &vertical_y, music_mask_pix,
&v_lines, &h_lines);
if (tessedit_dump_pageseg_images)
pixWrite("tessnolines.png", pix_binary_, IFF_PNG);
// Leptonica is used to find a mask of the photo regions in the input.
*photo_mask_pix = ImageFind::FindImages(pix_binary_);
if (tessedit_dump_pageseg_images)
pixWrite("tessnoimages.png", pix_binary_, IFF_PNG);
if (single_column)
v_lines.clear();
// The rest of the algorithm uses the usual connected components.
textord_.find_components(pix_binary_, blocks, to_blocks);
TO_BLOCK_IT to_block_it(to_blocks);
// There must be exactly one input block.
// TODO(rays) handle new textline finding with a UNLV zone file.
ASSERT_HOST(to_blocks->singleton());
TO_BLOCK* to_block = to_block_it.data();
TBOX blkbox = to_block->block->bounding_box();
ColumnFinder* finder = NULL;
if (to_block->line_size >= 2) {
finder = new ColumnFinder(static_cast<int>(to_block->line_size),
blkbox.botleft(), blkbox.topright(),
source_resolution_,
&v_lines, &h_lines, vertical_x, vertical_y);
finder->SetupAndFilterNoise(*photo_mask_pix, to_block);
if (equ_detect_) {
equ_detect_->LabelSpecialText(to_block);
}
BLOBNBOX_CLIST osd_blobs;
// osd_orientation is the number of 90 degree rotations to make the
// characters upright. (See osdetect.h for precise definition.)
// We want the text lines horizontal, (vertical text indicates vertical
// textlines) which may conflict (eg vertically written CJK).
int osd_orientation = 0;
bool vertical_text = finder->IsVerticallyAlignedText(to_block, &osd_blobs);
if (osd && osd_tess != NULL && osr != NULL) {
os_detect_blobs(&osd_blobs, osr, osd_tess);
if (only_osd) {
delete finder;
return NULL;
}
osd_orientation = osr->best_result.orientation_id;
double osd_score = osr->orientations[osd_orientation];
double osd_margin = min_orientation_margin * 2;
for (int i = 0; i < 4; ++i) {
if (i != osd_orientation &&
osd_score - osr->orientations[i] < osd_margin) {
osd_margin = osd_score - osr->orientations[i];
}
}
if (osd_margin < min_orientation_margin) {
// The margin is weak.
int best_script_id = osr->best_result.script_id;
bool cjk = (best_script_id == osd_tess->unicharset.han_sid()) ||
(best_script_id == osd_tess->unicharset.hiragana_sid()) ||
(best_script_id == osd_tess->unicharset.katakana_sid());
if (!cjk && !vertical_text && osd_orientation == 2) {
// upside down latin text is improbable with such a weak margin.
tprintf("OSD: Weak margin (%.2f), horiz textlines, not CJK: "
"Don't rotate.\n", osd_margin);
osd_orientation = 0;
} else {
tprintf("OSD: Weak margin (%.2f) for %d blob text block, "
"but using orientation anyway: %d\n",
osd_blobs.length(), osd_margin, osd_orientation);
}
}
}
osd_blobs.shallow_clear();
finder->CorrectOrientation(to_block, vertical_text, osd_orientation);
}
return finder;
}