void
MakeWordBoxes2(PIX *pixs,
l_int32 reduction,
L_REGPARAMS *rp)
{
l_int32 default_minwidth = 10;
l_int32 default_minheight = 10;
l_int32 default_maxwidth = 400;
l_int32 default_maxheight = 70;
l_int32 minwidth, minheight, maxwidth, maxheight;
BOXA *boxa1, *boxa2;
NUMA *na;
PIX *pixd1, *pixd2;
PIXA *pixa;
minwidth = default_minwidth / reduction;
minheight = default_minheight / reduction;
maxwidth = default_maxwidth / reduction;
maxheight = default_maxheight / reduction;
/* Get the word boxes */
pixGetWordsInTextlines(pixs, reduction, minwidth, minheight,
maxwidth, maxheight, &boxa1, &pixa, &na);
pixaDestroy(&pixa);
numaDestroy(&na);
if (reduction == 1)
boxa2 = boxaCopy(boxa1, L_CLONE);
else
boxa2 = boxaTransform(boxa1, 0, 0, 2.0, 2.0);
pixd1 = pixConvertTo8(pixs, 1);
pixRenderBoxaArb(pixd1, boxa2, 2, 255, 0, 0);
regTestWritePixAndCheck(rp, pixd1, IFF_PNG);
pixDisplayWithTitle(pixd1, 800, 100, NULL, rp->display);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
/* Do it again with this interface. The result should be the same. */
pixGetWordBoxesInTextlines(pixs, reduction, minwidth, minheight,
maxwidth, maxheight, &boxa1, NULL);
if (reduction == 1)
boxa2 = boxaCopy(boxa1, L_CLONE);
else
boxa2 = boxaTransform(boxa1, 0, 0, 2.0, 2.0);
pixd2 = pixConvertTo8(pixs, 1);
pixRenderBoxaArb(pixd2, boxa2, 2, 255, 0, 0);
if (regTestComparePix(rp, pixd1, pixd2)) {
L_ERROR("pix not the same", "MakeWordBoxes2");
pixDisplayWithTitle(pixd2, 800, 100, NULL, rp->display);
}
pixDestroy(&pixd1);
pixDestroy(&pixd2);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
return;
}
int main(int argc,
char **argv)
{
char *filein, *fileout, *fontdir;
l_int32 w, h, width, sep;
l_float32 scalefact;
L_BMF *bmf;
BOXA *boxa1, *boxa2;
PIX *pixd;
static char mainName[] = "displayboxa";
if (argc != 4 && argc != 5) {
fprintf(stderr, "Syntax error in displayboxa:\n"
" displayboxa filein width fileout [fontdir]\n");
return 1;
}
filein = argv[1];
fileout = argv[3];
fontdir = (argc == 4) ? NULL : argv[4];
width = atoi(argv[2]);
if (width < 30) {
L_ERROR("width too small; setting to 100\n", mainName);
width = 100;
}
if ((boxa1 = boxaRead(filein)) == NULL)
return ERROR_INT("boxa not made", mainName, 1);
boxaGetExtent(boxa1, &w, &h, NULL);
scalefact = (l_float32)width / (l_float32)w;
boxa2 = boxaTransform(boxa1, 0, 0, scalefact, scalefact);
sep = L_MIN(width / 5, 20);
pixd = boxaDisplayTiled(boxa2, NULL, 1500, 2, 1.0, 0, sep, 2, fontdir);
pixWrite(fileout, pixd, IFF_PNG);
pixDisplay(pixd, 100, 100);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
pixDestroy(&pixd);
return 0;
}
/*!
* \brief pixFindBaselines()
*
* \param[in] pixs 1 bpp, 300 ppi
* \param[out] ppta [optional] pairs of pts corresponding to
* approx. ends of each text line
* \param[in] pixadb for debug output; use NULL to skip
* \return na of baseline y values, or NULL on error
*
* <pre>
* 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.
* </pre>
*/
NUMA *
pixFindBaselines(PIX *pixs,
PTA **ppta,
PIXA *pixadb)
{
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 *pix1, *pix2;
PTA *pta;
PROCNAME("pixFindBaselines");
if (ppta) *ppta = NULL;
if (!pixs || pixGetDepth(pixs) != 1)
return (NUMA *)ERROR_PTR("pixs undefined or not 1 bpp", procName, NULL);
/* Close up the text characters, removing noise */
pix1 = pixMorphSequence(pixs, "c25.1 + e15.1", 0);
/* Estimate the resolution */
if (pixadb) pixaAddPix(pixadb, pixScale(pix1, 0.25, 0.25), L_INSERT);
/* Save the difference of adjacent row sums.
* The high positive-going peaks are the baselines */
if ((nasum = pixCountPixelsByRow(pix1, NULL)) == NULL) {
pixDestroy(&pix1);
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);
}
numaDestroy(&nasum);
if (pixadb) { /* show the difference signal */
lept_mkdir("lept/baseline");
gplotSimple1(nadiff, GPLOT_PNG, "/tmp/lept/baseline/diff", "Diff Sig");
pix2 = pixRead("/tmp/lept/baseline/diff.png");
pixaAddPix(pixadb, pix2, L_INSERT);
}
/* Use the zeroes of the profile to locate each baseline. */
array = numaGetIArray(nadiff);
ndiff = numaGetCount(nadiff);
numaGetMax(nadiff, &maxval, &imaxloc);
numaDestroy(&nadiff);
/* 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);
}
}
}
LEPT_FREE(array);
/* If array[ndiff-1] is max, eg. no descenders, baseline at bottom */
if (inpeak) {
numaAddNumber(naval, max);
numaAddNumber(naloc, maxloc);
}
if (pixadb) { /* 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);
pix2 = pixRead("/tmp/lept/baseline/loc.png");
pixaAddPix(pixadb, pix2, L_INSERT);
}
numaDestroy(&naval);
/* 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. */
pix2 = pixMorphSequence(pix1, "r11 + c20.1 + o30.1 +c1.3", 0);
if (pixadb) pixaAddPix(pixadb, pix2, L_COPY);
boxa1 = pixConnComp(pix2, NULL, 4);
pixDestroy(&pix1);
pixDestroy(&pix2);
if (boxaGetCount(boxa1) == 0) {
numaDestroy(&naloc);
boxaDestroy(&boxa1);
L_INFO("no compnents after filtering\n", procName);
return NULL;
}
boxa2 = boxaTransform(boxa1, 0, 0, 4., 4.);
boxa3 = boxaSort(boxa2, L_SORT_BY_Y, L_SORT_INCREASING, NULL);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
/* Optionally, find the baseline segments */
pta = NULL;
if (ppta) {
pta = ptaCreate(0);
*ppta = pta;
}
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;
}
}
}
boxaDestroy(&boxa3);
if (pixadb && pta) { /* display baselines */
l_int32 npts, x1, y1, x2, y2;
pix1 = pixConvertTo32(pixs);
npts = ptaGetCount(pta);
for (i = 0; i < npts; i += 2) {
ptaGetIPt(pta, i, &x1, &y1);
ptaGetIPt(pta, i + 1, &x2, &y2);
pixRenderLineArb(pix1, x1, y1, x2, y2, 2, 255, 0, 0);
}
pixWrite("/tmp/lept/baseline/baselines.png", pix1, IFF_PNG);
pixaAddPix(pixadb, pixScale(pix1, 0.25, 0.25), L_INSERT);
pixDestroy(&pix1);
}
return naloc;
}
int main(int argc,
char **argv)
{
l_uint8 *data1, *data2;
l_int32 i, same, w, h, width, success, nba;
size_t size1, size2;
l_float32 diffarea, diffxor, scalefact;
BOX *box;
BOXA *boxa1, *boxa2, *boxa3;
BOXAA *baa1, *baa2, *baa3;
PIX *pix1, *pixdb;
PIXA *pixa1, *pixa2;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
lept_mkdir("lept/boxa");
/* Make a boxa and display its contents */
boxa1 = boxaCreate(6);
box = boxCreate(60, 60, 40, 20);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(120, 50, 20, 50);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(50, 140, 46, 60);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(166, 130, 64, 28);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(64, 224, 44, 34);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(117, 206, 26, 74);
boxaAddBox(boxa1, box, L_INSERT);
pix1 = DisplayBoxa(boxa1);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 0 */
pixDisplayWithTitle(pix1, 0, 0, NULL, rp->display);
pixDestroy(&pix1);
boxaCompareRegions(boxa1, boxa1, 100, &same, &diffarea, &diffxor, NULL);
regTestCompareValues(rp, 1, same, 0.0); /* 1 */
regTestCompareValues(rp, 0.0, diffarea, 0.0); /* 2 */
regTestCompareValues(rp, 0.0, diffxor, 0.0); /* 3 */
boxa2 = boxaTransform(boxa1, -13, -13, 1.0, 1.0);
boxaCompareRegions(boxa1, boxa2, 10, &same, &diffarea, &diffxor, NULL);
regTestCompareValues(rp, 1, same, 0.0); /* 4 */
regTestCompareValues(rp, 0.0, diffarea, 0.0); /* 5 */
regTestCompareValues(rp, 0.0, diffxor, 0.0); /* 6 */
boxaDestroy(&boxa2);
boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP_AND_BOT, 6,
L_ADJUST_CHOOSE_MIN, 1.0, 0);
pix1 = DisplayBoxa(boxa2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 7 */
pixDisplayWithTitle(pix1, 200, 0, NULL, rp->display);
pixDestroy(&pix1);
boxaCompareRegions(boxa1, boxa2, 10, &same, &diffarea, &diffxor, &pixdb);
regTestCompareValues(rp, 1, same, 0.0); /* 8 */
regTestCompareValues(rp, 0.053, diffarea, 0.002); /* 9 */
regTestCompareValues(rp, 0.240, diffxor, 0.002); /* 10 */
regTestWritePixAndCheck(rp, pixdb, IFF_PNG); /* 11 */
pixDisplayWithTitle(pixdb, 400, 0, NULL, rp->display);
pixDestroy(&pixdb);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
/* Input is a fairly clean boxa */
boxa1 = boxaRead("boxa1.ba");
boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
L_ADJUST_CHOOSE_MIN, 1.05, 1);
width = 100;
boxaGetExtent(boxa2, &w, &h, NULL);
scalefact = (l_float32)width / (l_float32)w;
boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 12 */
pixDisplayWithTitle(pix1, 600, 0, NULL, rp->display);
pixDestroy(&pix1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
/* Input is an unsmoothed and noisy boxa */
boxa1 = boxaRead("boxa2.ba");
boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
L_ADJUST_CHOOSE_MIN, 1.05, 1);
width = 100;
boxaGetExtent(boxa2, &w, &h, NULL);
scalefact = (l_float32)width / (l_float32)w;
boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 13 */
pixDisplayWithTitle(pix1, 800, 0, NULL, rp->display);
pixDestroy(&pix1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
/* Input is a boxa smoothed with a median window filter */
boxa1 = boxaRead("boxa3.ba");
boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
L_ADJUST_CHOOSE_MIN, 1.05, 1);
width = 100;
boxaGetExtent(boxa2, &w, &h, NULL);
scalefact = (l_float32)width / (l_float32)w;
boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 14 */
pixDisplayWithTitle(pix1, 1000, 0, NULL, rp->display);
pixDestroy(&pix1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
/* Test serialized boxa I/O to and from memory */
data1 = l_binaryRead("boxa2.ba", &size1);
boxa1 = boxaReadMem(data1, size1);
boxaWriteMem(&data2, &size2, boxa1);
boxa2 = boxaReadMem(data2, size2);
boxaWrite("/tmp/lept/boxa/boxa1.ba", boxa1);
boxaWrite("/tmp/lept/boxa/boxa2.ba", boxa2);
filesAreIdentical("/tmp/lept/boxa/boxa1.ba", "/tmp/lept/boxa/boxa2.ba",
&same);
regTestCompareValues(rp, 1, same, 0.0); /* 15 */
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
lept_free(data1);
lept_free(data2);
/* ----------- Test pixaDisplayBoxaa() ------------ */
pixa1 = pixaReadBoth("showboxes.pac");
baa1 = boxaaRead("showboxes1.baa");
baa2 = boxaaTranspose(baa1);
baa3 = boxaaTranspose(baa2);
nba = boxaaGetCount(baa1);
success = TRUE;
for (i = 0; i < nba; i++) {
boxa1 = boxaaGetBoxa(baa1, i, L_CLONE);
boxa2 = boxaaGetBoxa(baa3, i, L_CLONE);
boxaEqual(boxa1, boxa2, 0, NULL, &same);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
if (!same) success = FALSE;
}
/* Check that the transpose is reversible */
regTestCompareValues(rp, 1, success, 0.0); /* 16 */
pixa2 = pixaDisplayBoxaa(pixa1, baa2, L_DRAW_RGB, 2);
pix1 = pixaDisplayTiledInRows(pixa2, 32, 1400, 1.0, 0, 10, 0);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 17 */
pixDisplayWithTitle(pix1, 0, 600, NULL, rp->display);
fprintf(stderr, "Writing to: /tmp/lept/boxa/show.pdf\n");
l_pdfSetDateAndVersion(FALSE);
pixaConvertToPdf(pixa2, 75, 1.0, 0, 0, NULL, "/tmp/lept/boxa/show.pdf");
regTestCheckFile(rp, "/tmp/lept/boxa/show.pdf"); /* 18 */
pixDestroy(&pix1);
pixaDestroy(&pixa1);
pixaDestroy(&pixa2);
boxaaDestroy(&baa1);
boxaaDestroy(&baa2);
boxaaDestroy(&baa3);
return regTestCleanup(rp);
}
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;
}
/*!
* pixSplitIntoCharacters()
*
* Input: pixs (1 bpp, contains only deskewed text)
* minw (minimum component width for initial filtering; typ. 4)
* minh (minimum component height for initial filtering; typ. 4)
* &boxa (<optional return> character bounding boxes)
* &pixa (<optional return> character images)
* &pixdebug (<optional return> showing splittings)
*
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) This is a simple function that attempts to find split points
* based on vertical pixel profiles.
* (2) It should be given an image that has an arbitrary number
* of text characters.
* (3) The returned pixa includes the boxes from which the
* (possibly split) components are extracted.
*/
l_int32
pixSplitIntoCharacters(PIX *pixs,
l_int32 minw,
l_int32 minh,
BOXA **pboxa,
PIXA **ppixa,
PIX **ppixdebug)
{
l_int32 ncomp, i, xoff, yoff;
BOXA *boxa1, *boxa2, *boxat1, *boxat2, *boxad;
BOXAA *baa;
PIX *pix, *pix1, *pix2, *pixdb;
PIXA *pixa1, *pixadb;
PROCNAME("pixSplitIntoCharacters");
if (pboxa) *pboxa = NULL;
if (ppixa) *ppixa = NULL;
if (ppixdebug) *ppixdebug = NULL;
if (!pixs || pixGetDepth(pixs) != 1)
return ERROR_INT("pixs not defined or not 1 bpp", procName, 1);
/* Remove the small stuff */
pix1 = pixSelectBySize(pixs, minw, minh, 8, L_SELECT_IF_BOTH,
L_SELECT_IF_GT, NULL);
/* Small vertical close for consolidation */
pix2 = pixMorphSequence(pix1, "c1.10", 0);
pixDestroy(&pix1);
/* Get the 8-connected components */
boxa1 = pixConnComp(pix2, &pixa1, 8);
pixDestroy(&pix2);
boxaDestroy(&boxa1);
/* Split the components if obvious */
ncomp = pixaGetCount(pixa1);
boxa2 = boxaCreate(ncomp);
pixadb = (ppixdebug) ? pixaCreate(ncomp) : NULL;
for (i = 0; i < ncomp; i++) {
pix = pixaGetPix(pixa1, i, L_CLONE);
if (ppixdebug) {
boxat1 = pixSplitComponentWithProfile(pix, 10, 7, &pixdb);
if (pixdb)
pixaAddPix(pixadb, pixdb, L_INSERT);
} else {
boxat1 = pixSplitComponentWithProfile(pix, 10, 7, NULL);
}
pixaGetBoxGeometry(pixa1, i, &xoff, &yoff, NULL, NULL);
boxat2 = boxaTransform(boxat1, xoff, yoff, 1.0, 1.0);
boxaJoin(boxa2, boxat2, 0, -1);
pixDestroy(&pix);
boxaDestroy(&boxat1);
boxaDestroy(&boxat2);
}
pixaDestroy(&pixa1);
/* Generate the debug image */
if (ppixdebug) {
if (pixaGetCount(pixadb) > 0) {
*ppixdebug = pixaDisplayTiledInRows(pixadb, 32, 1500,
1.0, 0, 20, 1);
}
pixaDestroy(&pixadb);
}
/* Do a 2D sort on the bounding boxes, and flatten the result to 1D */
baa = boxaSort2d(boxa2, NULL, 0, 0, 5);
boxad = boxaaFlattenToBoxa(baa, NULL, L_CLONE);
boxaaDestroy(&baa);
boxaDestroy(&boxa2);
/* Optionally extract the pieces from the input image */
if (ppixa)
*ppixa = pixClipRectangles(pixs, boxad);
if (pboxa)
*pboxa = boxad;
else
boxaDestroy(&boxad);
return 0;
}
main(int argc,
char **argv)
{
l_int32 h;
l_float32 scalefactor;
BOX *box;
BOXA *boxa1, *boxa2;
BOXAA *baa;
PIX *pix1, *pix2, *pix3, *pix4, *pix5, *pix6, *pix7, *pix8, *pix9;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
lept_rmdir("segtest");
lept_mkdir("segtest");
baa = boxaaCreate(5);
/* Image region input. */
pix1 = pixRead("wet-day.jpg");
pix2 = pixScaleToSize(pix1, WIDTH, 0);
pixWrite("/tmp/segtest/0.jpg", pix2, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/segtest/0.jpg"); /* 0 */
box = boxCreate(105, 161, 620, 872); /* image region */
boxa1 = boxaCreate(1);
boxaAddBox(boxa1, box, L_INSERT);
boxaaAddBoxa(baa, boxa1, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
/* Compute image region at w = 2 * WIDTH */
pix1 = pixRead("candelabrum-11.jpg");
pix2 = pixScaleToSize(pix1, WIDTH, 0);
pix3 = pixConvertTo1(pix2, 100);
pix4 = pixExpandBinaryPower2(pix3, 2); /* w = 2 * WIDTH */
pix5 = pixGenHalftoneMask(pix4, NULL, NULL, 1);
pix6 = pixMorphSequence(pix5, "c20.1 + c1.20", 0);
pix7 = pixMaskConnComp(pix6, 8, &boxa1);
pix8 = pixReduceBinary2(pix7, NULL); /* back to w = WIDTH */
pix9 = pixBackgroundNormSimple(pix2, pix8, NULL);
pixWrite("/tmp/segtest/1.jpg", pix9, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/segtest/1.jpg"); /* 1 */
boxa2 = boxaTransform(boxa1, 0, 0, 0.5, 0.5); /* back to w = WIDTH */
boxaaAddBoxa(baa, boxa2, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
pixDestroy(&pix5);
pixDestroy(&pix6);
pixDestroy(&pix7);
pixDestroy(&pix8);
pixDestroy(&pix9);
boxaDestroy(&boxa1);
/* Use mask to find image region */
pix1 = pixRead("lion-page.00016.jpg");
pix2 = pixScaleToSize(pix1, WIDTH, 0);
pixWrite("/tmp/segtest/2.jpg", pix2, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/segtest/2.jpg"); /* 2 */
pix3 = pixRead("lion-mask.00016.tif");
pix4 = pixScaleToSize(pix3, WIDTH, 0);
boxa1 = pixConnComp(pix4, NULL, 8);
boxaaAddBoxa(baa, boxa1, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
/* Compute image region at full res */
pix1 = pixRead("rabi.png");
scalefactor = (l_float32)WIDTH / (l_float32)pixGetWidth(pix1);
pix2 = pixScaleToGray(pix1, scalefactor);
pixWrite("/tmp/segtest/3.jpg", pix2, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/segtest/3.jpg"); /* 3 */
pix3 = pixGenHalftoneMask(pix1, NULL, NULL, 0);
pix4 = pixMorphSequence(pix3, "c20.1 + c1.20", 0);
boxa1 = pixConnComp(pix4, NULL, 8);
boxa2 = boxaTransform(boxa1, 0, 0, scalefactor, scalefactor);
boxaaAddBoxa(baa, boxa2, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
boxaDestroy(&boxa1);
/* Page with no image regions */
pix1 = pixRead("lucasta-47.jpg");
pix2 = pixScaleToSize(pix1, WIDTH, 0);
boxa1 = boxaCreate(1);
pixWrite("/tmp/segtest/4.jpg", pix2, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/segtest/4.jpg"); /* 4 */
boxaaAddBoxa(baa, boxa1, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
/* Page that is all image */
pix1 = pixRead("map1.jpg");
pix2 = pixScaleToSize(pix1, WIDTH, 0);
pixWrite("/tmp/segtest/5.jpg", pix2, IFF_JFIF_JPEG);
regTestCheckFile(rp, "/tmp/segtest/5.jpg"); /* 5 */
h = pixGetHeight(pix2);
box = boxCreate(0, 0, WIDTH, h);
boxa1 = boxaCreate(1);
boxaAddBox(boxa1, box, L_INSERT);
boxaaAddBoxa(baa, boxa1, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
/* Save the boxaa file */
boxaaWrite("/tmp/segtest/seg.baa", baa);
regTestCheckFile(rp, "/tmp/segtest/seg.baa"); /* 6 */
/* Do the conversion */
l_pdfSetDateAndVersion(FALSE);
convertSegmentedFilesToPdf("/tmp/segtest", ".jpg", 100, L_G4_ENCODE,
140, baa, 75, 0.6, "Segmentation Test",
"/tmp/pdfseg.7.pdf");
regTestCheckFile(rp, "/tmp/pdfseg.7.pdf"); /* 7 */
boxaaDestroy(&baa);
return regTestCleanup(rp);
}
void static
TestBoxa(L_REGPARAMS *rp,
l_int32 index)
{
l_uint8 *data;
l_int32 w, h, medw, medh, isame;
size_t size;
l_float32 scalefact, devw, devh, ratiowh, fvarp, fvarm;
BOXA *boxa1, *boxa2, *boxa3;
PIX *pix1;
boxa1 = boxaRead(boxafiles[index]);
/* Read and display initial boxa */
boxaGetExtent(boxa1, &w, &h, NULL);
scalefact = 100.0 / (l_float32)w;
boxa2 = boxaTransform(boxa1, 0, 0, scalefact, scalefact);
boxaWriteMem(&data, &size, boxa2);
regTestWriteDataAndCheck(rp, data, size, "ba"); /* 0, 13, 26 */
lept_free(data);
pix1 = boxaDisplayTiled(boxa2, NULL, 0, -1, 2200, 2, 1.0, 0, 3, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 1, 14, 27 */
pixDisplayWithTitle(pix1, 0, 0, NULL, rp->display);
pixDestroy(&pix1);
/* Find the median sizes */
boxaMedianDimensions(boxa2, &medw, &medh, NULL, NULL, NULL, NULL,
NULL, NULL);
if (rp->display)
fprintf(stderr, "median width = %d, median height = %d\n", medw, medh);
/* Check for deviations from median by pairs: method 1 */
boxaSizeConsistency1(boxa2, L_CHECK_HEIGHT, 0.0, 0.0,
&fvarp, &fvarm, &isame);
regTestCompareValues(rp, varp[index], fvarp, 0.003); /* 2, 15, 28 */
regTestCompareValues(rp, varm[index], fvarm, 0.003); /* 3, 16, 29 */
regTestCompareValues(rp, same[index], isame, 0); /* 4, 17, 30 */
if (rp->display)
fprintf(stderr, "fvarp = %7.4f, fvarm = %7.4f, same = %d\n",
fvarp, fvarm, isame);
/* Check for deviations from median by pairs: method 2 */
boxaSizeConsistency2(boxa2, &devw, &devh, 0);
regTestCompareValues(rp, devwidth[index], devw, 0.001); /* 5, 18, 31 */
regTestCompareValues(rp, devheight[index], devh, 0.001); /* 6, 19, 32 */
if (rp->display)
fprintf(stderr, "dev width = %7.4f, dev height = %7.4f\n", devw, devh);
/* Reconcile widths */
boxa3 = boxaReconcileSizeByMedian(boxa2, L_CHECK_WIDTH, 0.05, 0.04, 1.03,
NULL, NULL, &ratiowh);
boxaWriteMem(&data, &size, boxa3);
regTestWriteDataAndCheck(rp, data, size, "ba"); /* 7, 20, 33 */
lept_free(data);
pix1 = boxaDisplayTiled(boxa3, NULL, 0, -1, 2200, 2, 1.0, 0, 3, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 8, 21, 34 */
pixDisplayWithTitle(pix1, 500, 0, NULL, rp->display);
if (rp->display)
fprintf(stderr, "ratio median width/height = %6.3f\n", ratiowh);
boxaDestroy(&boxa3);
pixDestroy(&pix1);
/* Reconcile heights */
boxa3 = boxaReconcileSizeByMedian(boxa2, L_CHECK_HEIGHT, 0.05, 0.04, 1.03,
NULL, NULL, NULL);
boxaWriteMem(&data, &size, boxa3);
regTestWriteDataAndCheck(rp, data, size, "ba"); /* 9, 22, 35 */
lept_free(data);
pix1 = boxaDisplayTiled(boxa3, NULL, 0, -1, 2200, 2, 1.0, 0, 3, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 10, 23, 36 */
pixDisplayWithTitle(pix1, 1000, 0, NULL, rp->display);
boxaDestroy(&boxa3);
pixDestroy(&pix1);
/* Reconcile both widths and heights */
boxa3 = boxaReconcileSizeByMedian(boxa2, L_CHECK_BOTH, 0.05, 0.04, 1.03,
NULL, NULL, NULL);
boxaWriteMem(&data, &size, boxa3);
regTestWriteDataAndCheck(rp, data, size, "ba"); /* 11, 24, 37 */
lept_free(data);
pix1 = boxaDisplayTiled(boxa3, NULL, 0, -1, 2200, 2, 1.0, 0, 3, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 12, 25, 38 */
pixDisplayWithTitle(pix1, 1500, 0, NULL, rp->display);
boxaDestroy(&boxa3);
pixDestroy(&pix1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
}
int main(int argc,
char **argv)
{
l_uint8 *data1, *data2;
l_int32 i, same, w, h, width, success, nba;
size_t size1, size2;
l_float32 diffarea, diffxor, scalefact;
BOX *box;
BOXA *boxa1, *boxa2, *boxa3;
BOXAA *baa1, *baa2, *baa3;
PIX *pix1, *pixdb;
PIXA *pixa1, *pixa2;
static char mainName[] = "boxa1_reg";
if (argc != 1)
return ERROR_INT(" Syntax: boxa1_reg", mainName, 1);
lept_mkdir("lept/boxa");
/* Make a boxa and display its contents */
boxa1 = boxaCreate(6);
box = boxCreate(60, 60, 40, 20);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(120, 50, 20, 50);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(50, 140, 46, 60);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(166, 130, 64, 28);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(64, 224, 44, 34);
boxaAddBox(boxa1, box, L_INSERT);
box = boxCreate(117, 206, 26, 74);
boxaAddBox(boxa1, box, L_INSERT);
pix1 = DisplayBoxa(boxa1);
pixDisplay(pix1, 100, 100);
pixDestroy(&pix1);
boxaCompareRegions(boxa1, boxa1, 100, &same, &diffarea, &diffxor, NULL);
fprintf(stderr, "same = %d, diffarea = %5.3f, diffxor = %5.3f\n",
same, diffarea, diffxor);
boxa2 = boxaTransform(boxa1, -13, -13, 1.0, 1.0);
boxaCompareRegions(boxa1, boxa2, 10, &same, &diffarea, &diffxor, NULL);
fprintf(stderr, "same = %d, diffarea = %5.3f, diffxor = %5.3f\n",
same, diffarea, diffxor);
boxaDestroy(&boxa2);
boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP_AND_BOT, 6,
L_ADJUST_CHOOSE_MIN, 1.0, 0);
pix1 = DisplayBoxa(boxa2);
pixDisplay(pix1, 100, 500);
pixDestroy(&pix1);
boxaCompareRegions(boxa1, boxa2, 10, &same, &diffarea, &diffxor, &pixdb);
fprintf(stderr, "same = %d, diffarea = %5.3f, diffxor = %5.3f\n",
same, diffarea, diffxor);
pixDisplay(pixdb, 700, 100);
pixDestroy(&pixdb);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
/* Input is a fairly clean boxa */
boxa1 = boxaRead("boxa1.ba");
boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
L_ADJUST_CHOOSE_MIN, 1.05, 1);
width = 100;
boxaGetExtent(boxa2, &w, &h, NULL);
scalefact = (l_float32)width / (l_float32)w;
boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
pixDisplay(pix1, 0, 100);
pixWrite("/tmp/lept/boxa/pix1.png", pix1, IFF_PNG);
pixDestroy(&pix1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
/* Input is an unsmoothed and noisy boxa */
boxa1 = boxaRead("boxa2.ba");
boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
L_ADJUST_CHOOSE_MIN, 1.05, 1);
width = 100;
boxaGetExtent(boxa2, &w, &h, NULL);
scalefact = (l_float32)width / (l_float32)w;
boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
pixDisplay(pix1, 500, 100);
pixWrite("/tmp/lept/boxa/pix2.png", pix1, IFF_PNG);
pixDestroy(&pix1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
/* Input is a boxa smoothed with a median window filter */
boxa1 = boxaRead("boxa3.ba");
boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
L_ADJUST_CHOOSE_MIN, 1.05, 1);
width = 100;
boxaGetExtent(boxa2, &w, &h, NULL);
scalefact = (l_float32)width / (l_float32)w;
boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
pixDisplay(pix1, 1000, 100);
pixWrite("/tmp/lept/boxa/pix3.png", pix1, IFF_PNG);
pixDestroy(&pix1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
/* Test serialized boxa I/O to and from memory */
data1 = l_binaryRead("boxa2.ba", &size1);
boxa1 = boxaReadMem(data1, size1);
boxaWriteMem(&data2, &size2, boxa1);
boxa2 = boxaReadMem(data2, size2);
boxaWrite("/tmp/lept/boxa/boxa1.ba", boxa1);
boxaWrite("/tmp/lept/boxa/boxa2.ba", boxa2);
filesAreIdentical("/tmp/lept/boxa/boxa1.ba", "/tmp/lept/boxa/boxa2.ba",
&same);
if (same)
fprintf(stderr, "Good: boxes files are identical\n");
else
fprintf(stderr, "Bad: boxes files differ\n");
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
lept_free(data1);
lept_free(data2);
/* Test pixaDisplayBoxaa() */
pixa1 = pixaReadBoth("showboxes.pac");
baa1 = boxaaRead("showboxes1.baa");
baa2 = boxaaTranspose(baa1);
baa3 = boxaaTranspose(baa2);
nba = boxaaGetCount(baa1);
success = TRUE;
for (i = 0; i < nba; i++) {
boxa1 = boxaaGetBoxa(baa1, i, L_CLONE);
boxa2 = boxaaGetBoxa(baa3, i, L_CLONE);
boxaEqual(boxa1, boxa2, 0, NULL, &same);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
if (!same) success = FALSE;
}
if (success)
fprintf(stderr, "Good: transpose is reversible\n");
else
fprintf(stderr, "Bad: transpose failed\n");
pixa2 = pixaDisplayBoxaa(pixa1, baa2, L_DRAW_RGB, 2);
pix1 = pixaDisplayTiledInRows(pixa2, 32, 1400, 1.0, 0, 10, 0);
pixDisplay(pix1, 0, 600);
fprintf(stderr, "Writing to: /tmp/lept/boxa/show.pdf\n");
pixaConvertToPdf(pixa2, 75, 1.0, 0, 0, NULL, "/tmp/lept/boxa/show.pdf");
pixDestroy(&pix1);
pixaDestroy(&pixa1);
pixaDestroy(&pixa2);
boxaaDestroy(&baa1);
boxaaDestroy(&baa2);
boxaaDestroy(&baa3);
return 0;
}