int main(int argc,
char **argv)
{
l_int32 i, w, h, d, rotflag;
PIX *pixs, *pixt, *pixd;
l_float32 angle, deg2rad, pops, ang;
char *filein, *fileout;
static char mainName[] = "rotatetest1";
if (argc != 4)
return ERROR_INT(" Syntax: rotatetest1 filein angle fileout",
mainName, 1);
filein = argv[1];
angle = atof(argv[2]);
fileout = argv[3];
deg2rad = 3.1415926535 / 180.;
lept_mkdir("lept/rotate");
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("pix not made", mainName, 1);
if (pixGetDepth(pixs) == 1) {
pixt = pixScaleToGray3(pixs);
pixDestroy(&pixs);
pixs = pixAddBorderGeneral(pixt, 1, 0, 1, 0, 255);
pixDestroy(&pixt);
}
pixGetDimensions(pixs, &w, &h, &d);
fprintf(stderr, "w = %d, h = %d\n", w, h);
#if 0
/* repertory of rotation operations to choose from */
pixd = pixRotateAM(pixs, deg2rad * angle, L_BRING_IN_WHITE);
pixd = pixRotateAMColor(pixs, deg2rad * angle, 0xffffff00);
pixd = pixRotateAMColorFast(pixs, deg2rad * angle, 255);
pixd = pixRotateAMCorner(pixs, deg2rad * angle, L_BRING_IN_WHITE);
pixd = pixRotateShear(pixs, w /2, h / 2, deg2rad * angle,
L_BRING_IN_WHITE);
pixd = pixRotate3Shear(pixs, w /2, h / 2, deg2rad * angle,
L_BRING_IN_WHITE);
pixRotateShearIP(pixs, w / 2, h / 2, deg2rad * angle); pixd = pixs;
#endif
#if 0
/* timing of shear rotation */
for (i = 0; i < NITERS; i++) {
pixd = pixRotateShear(pixs, (i * w) / NITERS,
(i * h) / NITERS, deg2rad * angle,
L_BRING_IN_WHITE);
pixDisplay(pixd, 100 + 20 * i, 100 + 20 * i);
pixDestroy(&pixd);
}
#endif
#if 0
/* timing of in-place shear rotation */
for (i = 0; i < NITERS; i++) {
pixRotateShearIP(pixs, w/2, h/2, deg2rad * angle, L_BRING_IN_WHITE);
/* pixRotateShearCenterIP(pixs, deg2rad * angle, L_BRING_IN_WHITE); */
pixDisplay(pixs, 100 + 20 * i, 100 + 20 * i);
}
pixd = pixs;
if (pixGetDepth(pixd) == 1)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
pixDestroy(&pixs);
#endif
#if 0
/* timing of various rotation operations (choose) */
startTimer();
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
for (i = 0; i < NTIMES; i++) {
pixd = pixRotateShearCenter(pixs, deg2rad * angle, L_BRING_IN_WHITE);
pixDestroy(&pixd);
}
pops = (l_float32)(w * h * NTIMES / 1000000.) / stopTimer();
fprintf(stderr, "vers. 1, mpops: %f\n", pops);
startTimer();
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
for (i = 0; i < NTIMES; i++) {
pixRotateShearIP(pixs, w/2, h/2, deg2rad * angle, L_BRING_IN_WHITE);
}
pops = (l_float32)(w * h * NTIMES / 1000000.) / stopTimer();
fprintf(stderr, "shear, mpops: %f\n", pops);
pixWrite(fileout, pixs, IFF_PNG);
for (i = 0; i < NTIMES; i++) {
pixRotateShearIP(pixs, w/2, h/2, -deg2rad * angle, L_BRING_IN_WHITE);
}
pixWrite("/usr/tmp/junkout", pixs, IFF_PNG);
#endif
#if 0
/* area-mapping rotation operations */
pixd = pixRotateAM(pixs, deg2rad * angle, L_BRING_IN_WHITE);
/* pixd = pixRotateAMColorFast(pixs, deg2rad * angle, 255); */
if (pixGetDepth(pixd) == 1)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
#endif
#if 0
/* compare the standard area-map color rotation with
* the fast area-map color rotation, on a pixel basis */
{
PIX *pix1, *pix2;
NUMA *nar, *nag, *nab, *naseq;
GPLOT *gplot;
startTimer();
pix1 = pixRotateAMColor(pixs, 0.12, 0xffffff00);
fprintf(stderr, " standard color rotate: %7.2f sec\n", stopTimer());
pixWrite("/tmp/lept/rotate/color1.jpg", pix1, IFF_JFIF_JPEG);
startTimer();
pix2 = pixRotateAMColorFast(pixs, 0.12, 0xffffff00);
fprintf(stderr, " fast color rotate: %7.2f sec\n", stopTimer());
pixWrite("/tmp/lept/rotate/color2.jpg", pix2, IFF_JFIF_JPEG);
pixd = pixAbsDifference(pix1, pix2);
pixGetColorHistogram(pixd, 1, &nar, &nag, &nab);
naseq = numaMakeSequence(0., 1., 256);
gplot = gplotCreate("/tmp/lept/rotate/absdiff", GPLOT_PNG,
"Number vs diff", "diff", "number");
gplotAddPlot(gplot, naseq, nar, GPLOT_POINTS, "red");
gplotAddPlot(gplot, naseq, nag, GPLOT_POINTS, "green");
gplotAddPlot(gplot, naseq, nab, GPLOT_POINTS, "blue");
gplotMakeOutput(gplot);
l_fileDisplay("/tmp/lept/rotate/absdiff.png", 100, 100, 1.0);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pixd);
numaDestroy(&nar);
numaDestroy(&nag);
numaDestroy(&nab);
numaDestroy(&naseq);
gplotDestroy(&gplot);
}
#endif
/* Do a succession of 180 7-degree rotations in a cw
* direction, and unwind the result with another set in
* a ccw direction. Although there is a considerable amount
* of distortion after successive rotations, after all
* 360 rotations, the resulting image is restored to
* its original pristine condition! */
#if 1
rotflag = L_ROTATE_AREA_MAP;
/* rotflag = L_ROTATE_SHEAR; */
/* rotflag = L_ROTATE_SAMPLING; */
ang = 7.0 * deg2rad;
pixGetDimensions(pixs, &w, &h, NULL);
pixd = pixRotate(pixs, ang, rotflag, L_BRING_IN_WHITE, w, h);
pixWrite("/tmp/lept/rotate/rot7.png", pixd, IFF_PNG);
for (i = 1; i < 180; i++) {
pixs = pixd;
pixd = pixRotate(pixs, ang, rotflag, L_BRING_IN_WHITE, w, h);
if ((i % 30) == 0) pixDisplay(pixd, 600, 0);
pixDestroy(&pixs);
}
pixWrite("/tmp/lept/rotate/spin.png", pixd, IFF_PNG);
pixDisplay(pixd, 0, 0);
for (i = 0; i < 180; i++) {
pixs = pixd;
pixd = pixRotate(pixs, -ang, rotflag, L_BRING_IN_WHITE, w, h);
if (i && (i % 30) == 0) pixDisplay(pixd, 600, 500);
pixDestroy(&pixs);
}
pixWrite("/tmp/lept/rotate/unspin.png", pixd, IFF_PNG);
pixDisplay(pixd, 0, 500);
pixDestroy(&pixd);
#endif
return 0;
}
/*!
* pixFMorphopGen_3()
*
* Input: pixd (usual 3 choices: null, == pixs, != pixs)
* pixs (1 bpp)
* operation (L_MORPH_DILATE, L_MORPH_ERODE,
* L_MORPH_OPEN, L_MORPH_CLOSE)
* sel name
* Return: pixd
*
* Notes:
* (1) This is a dwa operation, and the Sels must be limited in
* size to not more than 31 pixels about the origin.
* (2) A border of appropriate size (32 pixels, or 64 pixels
* for safe closing with asymmetric b.c.) must be added before
* this function is called.
* (3) This handles all required setting of the border pixels
* before erosion and dilation.
* (4) The closing operation is safe; no pixels can be removed
* near the boundary.
*/
PIX *
pixFMorphopGen_3(PIX *pixd,
PIX *pixs,
l_int32 operation,
char *selname)
{
l_int32 i, index, found, w, h, wpls, wpld, bordercolor, erodeop, borderop;
l_uint32 *datad, *datas, *datat;
PIX *pixt;
PROCNAME("pixFMorphopGen_3");
if (!pixs)
return (PIX *)ERROR_PTR("pixs not defined", procName, pixd);
if (pixGetDepth(pixs) != 1)
return (PIX *)ERROR_PTR("pixs must be 1 bpp", procName, pixd);
/* Get boundary colors to use */
bordercolor = getMorphBorderPixelColor(L_MORPH_ERODE, 1);
if (bordercolor == 1)
erodeop = PIX_SET;
else
erodeop = PIX_CLR;
found = FALSE;
for (i = 0; i < NUM_SELS_GENERATED; i++) {
if (strcmp(selname, SEL_NAMES[i]) == 0) {
found = TRUE;
index = 2 * i;
break;
}
}
if (found == FALSE)
return (PIX *)ERROR_PTR("sel index not found", procName, pixd);
if (!pixd) {
if ((pixd = pixCreateTemplate(pixs)) == NULL)
return (PIX *)ERROR_PTR("pixd not made", procName, NULL);
}
else /* for in-place or pre-allocated */
pixResizeImageData(pixd, pixs);
wpls = pixGetWpl(pixs);
wpld = pixGetWpl(pixd);
/* The images must be surrounded, in advance, with a border of
* size 32 pixels (or 64, for closing), that we'll read from.
* Fabricate a "proper" image as the subimage within the 32
* pixel border, having the following parameters: */
w = pixGetWidth(pixs) - 64;
h = pixGetHeight(pixs) - 64;
datas = pixGetData(pixs) + 32 * wpls + 1;
datad = pixGetData(pixd) + 32 * wpld + 1;
if (operation == L_MORPH_DILATE || operation == L_MORPH_ERODE) {
borderop = PIX_CLR;
if (operation == L_MORPH_ERODE) {
borderop = erodeop;
index++;
}
if (pixd == pixs) { /* in-place; generate a temp image */
if ((pixt = pixCopy(NULL, pixs)) == NULL)
return (PIX *)ERROR_PTR("pixt not made", procName, pixd);
datat = pixGetData(pixt) + 32 * wpls + 1;
pixSetOrClearBorder(pixt, 32, 32, 32, 32, borderop);
fmorphopgen_low_3(datad, w, h, wpld, datat, wpls, index);
pixDestroy(&pixt);
}
else { /* not in-place */
pixSetOrClearBorder(pixs, 32, 32, 32, 32, borderop);
fmorphopgen_low_3(datad, w, h, wpld, datas, wpls, index);
}
}
else { /* opening or closing; generate a temp image */
if ((pixt = pixCreateTemplate(pixs)) == NULL)
return (PIX *)ERROR_PTR("pixt not made", procName, pixd);
datat = pixGetData(pixt) + 32 * wpls + 1;
if (operation == L_MORPH_OPEN) {
pixSetOrClearBorder(pixs, 32, 32, 32, 32, erodeop);
fmorphopgen_low_3(datat, w, h, wpls, datas, wpls, index+1);
pixSetOrClearBorder(pixt, 32, 32, 32, 32, PIX_CLR);
fmorphopgen_low_3(datad, w, h, wpld, datat, wpls, index);
}
else { /* closing */
pixSetOrClearBorder(pixs, 32, 32, 32, 32, PIX_CLR);
fmorphopgen_low_3(datat, w, h, wpls, datas, wpls, index);
pixSetOrClearBorder(pixt, 32, 32, 32, 32, erodeop);
fmorphopgen_low_3(datad, w, h, wpld, datat, wpls, index+1);
}
pixDestroy(&pixt);
}
return pixd;
}
/*!
* pixColorSegmentTryCluster()
*
* Input: pixd
* pixs
* maxdist
* maxcolors
* Return: 0 if OK, 1 on error
*
* Note: This function should only be called from pixColorSegCluster()
*/
static l_int32
pixColorSegmentTryCluster(PIX *pixd,
PIX *pixs,
l_int32 maxdist,
l_int32 maxcolors) {
l_int32 rmap[256], gmap[256], bmap[256];
l_int32 w, h, wpls, wpld, i, j, k, found, ret, index, ncolors;
l_int32 rval, gval, bval, dist2, maxdist2;
l_int32 countarray[256];
l_int32 rsum[256], gsum[256], bsum[256];
l_uint32 *ppixel;
l_uint32 *datas, *datad, *lines, *lined;
PIXCMAP *cmap;
PROCNAME("pixColorSegmentTryCluster");
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (!pixd)
return ERROR_INT("pixd not defined", procName, 1);
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
maxdist2 = maxdist * maxdist;
cmap = pixGetColormap(pixd);
pixcmapClear(cmap);
for (k = 0; k < 256; k++) {
rsum[k] = gsum[k] = bsum[k] = 0;
rmap[k] = gmap[k] = bmap[k] = 0;
}
datas = pixGetData(pixs);
datad = pixGetData(pixd);
wpls = pixGetWpl(pixs);
wpld = pixGetWpl(pixd);
for (i = 0; i < h; i++) {
lines = datas + i * wpls;
lined = datad + i * wpld;
for (j = 0; j < w; j++) {
ppixel = lines + j;
rval = GET_DATA_BYTE(ppixel, COLOR_RED);
gval = GET_DATA_BYTE(ppixel, COLOR_GREEN);
bval = GET_DATA_BYTE(ppixel, COLOR_BLUE);
ncolors = pixcmapGetCount(cmap);
found = FALSE;
for (k = 0; k < ncolors; k++) {
dist2 = (rval - rmap[k]) * (rval - rmap[k]) +
(gval - gmap[k]) * (gval - gmap[k]) +
(bval - bmap[k]) * (bval - bmap[k]);
if (dist2 <= maxdist2) { /* take it; greedy */
found = TRUE;
SET_DATA_BYTE(lined, j, k);
countarray[k]++;
rsum[k] += rval;
gsum[k] += gval;
bsum[k] += bval;
break;
}
}
if (!found) { /* Add a new color */
ret = pixcmapAddNewColor(cmap, rval, gval, bval, &index);
/* fprintf(stderr,
"index = %d, (i,j) = (%d,%d), rgb = (%d, %d, %d)\n",
index, i, j, rval, gval, bval); */
if (ret == 0 && index < maxcolors) {
countarray[index] = 1;
SET_DATA_BYTE(lined, j, index);
rmap[index] = rval;
gmap[index] = gval;
bmap[index] = bval;
rsum[index] = rval;
gsum[index] = gval;
bsum[index] = bval;
} else {
L_INFO("maxcolors exceeded for maxdist = %d\n",
procName, maxdist);
return 1;
}
}
}
}
/* Replace the colors in the colormap by the averages */
for (k = 0; k < ncolors; k++) {
rval = rsum[k] / countarray[k];
gval = gsum[k] / countarray[k];
bval = bsum[k] / countarray[k];
pixcmapResetColor(cmap, k, rval, gval, bval);
}
return 0;
}
main(int argc,
char **argv)
{
char *text;
l_int32 w, h, d, wpl, count, npages, color, format;
FILE *fp;
PIX *pix;
PIXCMAP *cmap;
char *filein;
static char mainName[] = "fileinfo";
if (argc != 2)
exit(ERROR_INT(" Syntax: fileinfo filein", mainName, 1));
filein = argv[1];
l_pngSetStrip16To8(0); /* to preserve 16 bpp if format is png */
if ((pix = pixRead(filein)) == NULL)
exit(ERROR_INT("image not returned from file", mainName, 1));
format = pixGetInputFormat(pix);
fprintf(stderr, "Input image format type: %s\n",
ImageFileFormatExtensions[format]);
pixGetDimensions(pix, &w, &h, &d);
wpl = pixGetWpl(pix);
fprintf(stderr, "w = %d, h = %d, d = %d, wpl = %d\n", w, h, d, wpl);
fprintf(stderr, "xres = %d, yres = %d\n", pixGetXRes(pix), pixGetYRes(pix));
text = pixGetText(pix);
if (text) /* not null */
fprintf(stderr, "Text: %s\n", text);
cmap = pixGetColormap(pix);
if (cmap) {
pixcmapHasColor(cmap, &color);
if (color)
fprintf(stderr, "Colormap exists and has color values:");
else
fprintf(stderr, "Colormap exists and has only gray values:");
pixcmapWriteStream(stderr, pixGetColormap(pix));
}
else
fprintf(stderr, "Colormap does not exist.\n");
if (format == IFF_TIFF || format == IFF_TIFF_G4 ||
format == IFF_TIFF_G3 || format == IFF_TIFF_PACKBITS) {
fprintf(stderr, "Tiff header information:\n");
fp = fopen(filein, "r");
tiffGetCount(fp, &npages);
fclose(fp);
if (npages == 1)
fprintf(stderr, "One page in file\n", npages);
else
fprintf(stderr, "%d pages in file\n", npages);
fprintTiffInfo(stderr, filein);
}
if (d == 1) {
pixCountPixels(pix, &count, NULL);
fprintf(stderr, "pixel ratio ON/OFF = %6.3f\n",
(l_float32)count / (l_float32)(pixGetWidth(pix) * pixGetHeight(pix)));
}
pixDestroy(&pix);
exit(0);
}
bool MakeIndividualGlyphs(Pix* pix,
const vector<BoxChar*>& vbox,
const int input_tiff_page) {
// If checks fail, return false without exiting text2image
if (!pix) {
tprintf("ERROR: MakeIndividualGlyphs(): Input Pix* is nullptr\n");
return false;
} else if (FLAGS_glyph_resized_size <= 0) {
tprintf("ERROR: --glyph_resized_size must be positive\n");
return false;
} else if (FLAGS_glyph_num_border_pixels_to_pad < 0) {
tprintf("ERROR: --glyph_num_border_pixels_to_pad must be 0 or positive\n");
return false;
}
const int n_boxes = vbox.size();
int n_boxes_saved = 0;
int current_tiff_page = 0;
int y_previous = 0;
static int glyph_count = 0;
for (int i = 0; i < n_boxes; i++) {
// Get one bounding box
Box* b = vbox[i]->mutable_box();
if (!b) continue;
const int x = b->x;
const int y = b->y;
const int w = b->w;
const int h = b->h;
// Check present tiff page (for multipage tiff)
if (y < y_previous-pixGetHeight(pix)/10) {
tprintf("ERROR: Wrap-around encountered, at i=%d\n", i);
current_tiff_page++;
}
if (current_tiff_page < input_tiff_page) continue;
else if (current_tiff_page > input_tiff_page) break;
// Check box validity
if (x < 0 || y < 0 ||
(x+w-1) >= pixGetWidth(pix) ||
(y+h-1) >= pixGetHeight(pix)) {
tprintf("ERROR: MakeIndividualGlyphs(): Index out of range, at i=%d"
" (x=%d, y=%d, w=%d, h=%d\n)", i, x, y, w, h);
continue;
} else if (w < FLAGS_glyph_num_border_pixels_to_pad &&
h < FLAGS_glyph_num_border_pixels_to_pad) {
tprintf("ERROR: Input image too small to be a character, at i=%d\n", i);
continue;
}
// Crop the boxed character
Pix* pix_glyph = pixClipRectangle(pix, b, nullptr);
if (!pix_glyph) {
tprintf("ERROR: MakeIndividualGlyphs(): Failed to clip, at i=%d\n", i);
continue;
}
// Resize to square
Pix* pix_glyph_sq = pixScaleToSize(pix_glyph,
FLAGS_glyph_resized_size,
FLAGS_glyph_resized_size);
if (!pix_glyph_sq) {
tprintf("ERROR: MakeIndividualGlyphs(): Failed to resize, at i=%d\n", i);
continue;
}
// Zero-pad
Pix* pix_glyph_sq_pad = pixAddBorder(pix_glyph_sq,
FLAGS_glyph_num_border_pixels_to_pad,
0);
if (!pix_glyph_sq_pad) {
tprintf("ERROR: MakeIndividualGlyphs(): Failed to zero-pad, at i=%d\n",
i);
continue;
}
// Write out
Pix* pix_glyph_sq_pad_8 = pixConvertTo8(pix_glyph_sq_pad, false);
char filename[1024];
snprintf(filename, 1024, "%s_%d.jpg", FLAGS_outputbase.c_str(),
glyph_count++);
if (pixWriteJpeg(filename, pix_glyph_sq_pad_8, 100, 0)) {
tprintf("ERROR: MakeIndividualGlyphs(): Failed to write JPEG to %s,"
" at i=%d\n", filename, i);
continue;
}
pixDestroy(&pix_glyph);
pixDestroy(&pix_glyph_sq);
pixDestroy(&pix_glyph_sq_pad);
pixDestroy(&pix_glyph_sq_pad_8);
n_boxes_saved++;
y_previous = y;
}
if (n_boxes_saved == 0) {
return false;
} else {
tprintf("Total number of characters saved = %d\n", n_boxes_saved);
return true;
}
}
int main(int argc,
char **argv)
{
l_int32 i, j, w, h, same, width, height, cx, cy;
l_uint32 val;
BOX *box;
PIX *pix0, *pixs, *pixse, *pixd1, *pixd2;
SEL *sel;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
pix0 = pixRead("feyn-fract.tif");
box = boxCreate(293, 37, pixGetWidth(pix0) - 691, pixGetHeight(pix0) -145);
pixs = pixClipRectangle(pix0, box, NULL);
boxDestroy(&box);
if (rp->display) pixDisplay(pixs, 100, 100);
/* Test 63 different sizes */
for (width = 1; width <= 25; width += 3) { /* 9 values */
for (height = 1; height <= 25; height += 4) { /* 7 values */
cx = width / 2;
cy = height / 2;
/* Dilate using an actual sel */
sel = selCreateBrick(height, width, cy, cx, SEL_HIT);
pixd1 = pixDilate(NULL, pixs, sel);
/* Dilate using a pix as a sel */
pixse = pixCreate(width, height, 1);
pixSetAll(pixse);
pixd2 = pixCopy(NULL, pixs);
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
pixGetPixel(pixs, j, i, &val);
if (val)
pixRasterop(pixd2, j - cx, i - cy, width, height,
PIX_SRC | PIX_DST, pixse, 0, 0);
}
}
pixEqual(pixd1, pixd2, &same);
regTestCompareValues(rp, 1, same, 0.0); /* 0 - 62 */
if (same == 0) {
fprintf(stderr,
"Results differ for SE (width,height) = (%d,%d)\n",
width, height);
}
pixDestroy(&pixse);
pixDestroy(&pixd1);
pixDestroy(&pixd2);
selDestroy(&sel);
}
}
pixDestroy(&pix0);
pixDestroy(&pixs);
return regTestCleanup(rp);
}
bool TessPDFRenderer::AddImageHandler(TessBaseAPI* api) {
size_t n;
char buf[kBasicBufSize];
char buf2[kBasicBufSize];
Pix *pix = api->GetInputImage();
char *filename = (char *)api->GetInputName();
int ppi = api->GetSourceYResolution();
if (!pix || ppi <= 0)
return false;
double width = pixGetWidth(pix) * 72.0 / ppi;
double height = pixGetHeight(pix) * 72.0 / ppi;
snprintf(buf2, sizeof(buf2), "/XObject << /Im1 %ld 0 R >>\n", obj_ + 2);
const char *xobject = (textonly_) ? "" : buf2;
// PAGE
n = snprintf(buf, sizeof(buf),
"%ld 0 obj\n"
"<<\n"
" /Type /Page\n"
" /Parent %ld 0 R\n"
" /MediaBox [0 0 %.2f %.2f]\n"
" /Contents %ld 0 R\n"
" /Resources\n"
" <<\n"
" %s"
" /ProcSet [ /PDF /Text /ImageB /ImageI /ImageC ]\n"
" /Font << /f-0-0 %ld 0 R >>\n"
" >>\n"
">>\n"
"endobj\n",
obj_,
2L, // Pages object
width, height,
obj_ + 1, // Contents object
xobject, // Image object
3L); // Type0 Font
if (n >= sizeof(buf)) return false;
pages_.push_back(obj_);
AppendPDFObject(buf);
// CONTENTS
const std::unique_ptr<char[]> pdftext(GetPDFTextObjects(api, width, height));
const size_t pdftext_len = strlen(pdftext.get());
size_t len;
unsigned char *comp_pdftext = zlibCompress(
reinterpret_cast<unsigned char *>(pdftext.get()), pdftext_len, &len);
long comp_pdftext_len = len;
n = snprintf(buf, sizeof(buf),
"%ld 0 obj\n"
"<<\n"
" /Length %ld /Filter /FlateDecode\n"
">>\n"
"stream\n", obj_, comp_pdftext_len);
if (n >= sizeof(buf)) {
lept_free(comp_pdftext);
return false;
}
AppendString(buf);
long objsize = strlen(buf);
AppendData(reinterpret_cast<char *>(comp_pdftext), comp_pdftext_len);
objsize += comp_pdftext_len;
lept_free(comp_pdftext);
const char *b2 =
"endstream\n"
"endobj\n";
AppendString(b2);
objsize += strlen(b2);
AppendPDFObjectDIY(objsize);
if (!textonly_) {
char *pdf_object = nullptr;
if (!imageToPDFObj(pix, filename, obj_, &pdf_object, &objsize)) {
return false;
}
AppendData(pdf_object, objsize);
AppendPDFObjectDIY(objsize);
delete[] pdf_object;
}
return true;
}
/*!
* \brief pixGenerateSelWithRuns()
*
* \param[in] pixs 1 bpp, typically small, to be used as a pattern
* \param[in] nhlines number of hor lines along which elements are found
* \param[in] nvlines number of vert lines along which elements are found
* \param[in] distance min distance from boundary pixel; use 0 for default
* \param[in] minlength min runlength to set hit or miss; use 0 for default
* \param[in] toppix number of extra pixels of bg added above
* \param[in] botpix number of extra pixels of bg added below
* \param[in] leftpix number of extra pixels of bg added to left
* \param[in] rightpix number of extra pixels of bg added to right
* \param[out] ppixe [optional] input pix expanded by extra pixels
* \return sel hit-miss for input pattern, or NULL on error
*
* <pre>
* Notes:
* (1) The horizontal and vertical lines along which elements are
* selected are roughly equally spaced. The actual locations of
* the hits and misses are the centers of respective run-lengths.
* (2) No elements are selected that are less than 'distance' pixels away
* from a boundary pixel of the same color. This makes the
* match much more robust to edge noise. Valid inputs of
* 'distance' are 0, 1, 2, 3 and 4. If distance is either 0 or
* greater than 4, we reset it to the default value.
* (3) The 4 numbers for adding rectangles of pixels outside the fg
* can be use if the pattern is expected to be surrounded by bg
* (white) pixels. On the other hand, if the pattern may be near
* other fg (black) components on some sides, use 0 for those sides.
* (4) The pixels added to a side allow you to have miss elements there.
* There is a constraint between distance, minlength, and
* the added pixels for this to work. We illustrate using the
* default values. If you add 5 pixels to the top, and use a
* distance of 1, then you end up with a vertical run of at least
* 4 bg pixels along the top edge of the image. If you use a
* minimum runlength of 3, each vertical line will always find
* a miss near the center of its run. However, if you use a
* minimum runlength of 5, you will not get a miss on every vertical
* line. As another example, if you have 7 added pixels and a
* distance of 2, you can use a runlength up to 5 to guarantee
* that the miss element is recorded. We give a warning if the
* contraint does not guarantee a miss element outside the
* image proper.
* (5) The input pix, as extended by the extra pixels on selected sides,
* can optionally be returned. For debugging, call
* pixDisplayHitMissSel() to visualize the hit-miss sel superimposed
* on the generating bitmap.
* </pre>
*/
SEL *
pixGenerateSelWithRuns(PIX *pixs,
l_int32 nhlines,
l_int32 nvlines,
l_int32 distance,
l_int32 minlength,
l_int32 toppix,
l_int32 botpix,
l_int32 leftpix,
l_int32 rightpix,
PIX **ppixe)
{
l_int32 ws, hs, w, h, x, y, xval, yval, i, j, nh, nm;
l_float32 delh, delw;
NUMA *nah, *nam;
PIX *pixt1, *pixt2, *pixfg, *pixbg;
PTA *ptah, *ptam;
SEL *seld, *sel;
PROCNAME("pixGenerateSelWithRuns");
if (ppixe) *ppixe = NULL;
if (!pixs)
return (SEL *)ERROR_PTR("pixs not defined", procName, NULL);
if (pixGetDepth(pixs) != 1)
return (SEL *)ERROR_PTR("pixs not 1 bpp", procName, NULL);
if (nhlines < 1 && nvlines < 1)
return (SEL *)ERROR_PTR("nvlines and nhlines both < 1", procName, NULL);
if (distance <= 0)
distance = DEFAULT_DISTANCE_TO_BOUNDARY;
if (minlength <= 0)
minlength = DEFAULT_MIN_RUNLENGTH;
if (distance > MAX_DISTANCE_TO_BOUNDARY) {
L_WARNING("distance too large; setting to max value\n", procName);
distance = MAX_DISTANCE_TO_BOUNDARY;
}
/* Locate the foreground */
pixClipToForeground(pixs, &pixt1, NULL);
if (!pixt1)
return (SEL *)ERROR_PTR("pixt1 not made", procName, NULL);
ws = pixGetWidth(pixt1);
hs = pixGetHeight(pixt1);
w = ws;
h = hs;
/* Crop out a region including the foreground, and add pixels
* on sides depending on the side flags */
if (toppix || botpix || leftpix || rightpix) {
x = y = 0;
if (toppix) {
h += toppix;
y = toppix;
if (toppix < distance + minlength)
L_WARNING("no miss elements in added top pixels\n", procName);
}
if (botpix) {
h += botpix;
if (botpix < distance + minlength)
L_WARNING("no miss elements in added bot pixels\n", procName);
}
if (leftpix) {
w += leftpix;
x = leftpix;
if (leftpix < distance + minlength)
L_WARNING("no miss elements in added left pixels\n", procName);
}
if (rightpix) {
w += rightpix;
if (rightpix < distance + minlength)
L_WARNING("no miss elements in added right pixels\n", procName);
}
pixt2 = pixCreate(w, h, 1);
pixRasterop(pixt2, x, y, ws, hs, PIX_SRC, pixt1, 0, 0);
} else {
pixt2 = pixClone(pixt1);
}
if (ppixe)
*ppixe = pixClone(pixt2);
pixDestroy(&pixt1);
/* Identify fg and bg pixels that are at least 'distance' pixels
* away from the boundary pixels in their set */
seld = selCreateBrick(2 * distance + 1, 2 * distance + 1,
distance, distance, SEL_HIT);
pixfg = pixErode(NULL, pixt2, seld);
pixbg = pixDilate(NULL, pixt2, seld);
pixInvert(pixbg, pixbg);
selDestroy(&seld);
pixDestroy(&pixt2);
/* Accumulate hit and miss points */
ptah = ptaCreate(0);
ptam = ptaCreate(0);
if (nhlines >= 1) {
delh = (l_float32)h / (l_float32)(nhlines + 1);
for (i = 0, y = 0; i < nhlines; i++) {
y += (l_int32)(delh + 0.5);
nah = pixGetRunCentersOnLine(pixfg, -1, y, minlength);
nam = pixGetRunCentersOnLine(pixbg, -1, y, minlength);
nh = numaGetCount(nah);
nm = numaGetCount(nam);
for (j = 0; j < nh; j++) {
numaGetIValue(nah, j, &xval);
ptaAddPt(ptah, xval, y);
}
for (j = 0; j < nm; j++) {
numaGetIValue(nam, j, &xval);
ptaAddPt(ptam, xval, y);
}
numaDestroy(&nah);
numaDestroy(&nam);
}
}
if (nvlines >= 1) {
delw = (l_float32)w / (l_float32)(nvlines + 1);
for (i = 0, x = 0; i < nvlines; i++) {
x += (l_int32)(delw + 0.5);
nah = pixGetRunCentersOnLine(pixfg, x, -1, minlength);
nam = pixGetRunCentersOnLine(pixbg, x, -1, minlength);
nh = numaGetCount(nah);
nm = numaGetCount(nam);
for (j = 0; j < nh; j++) {
numaGetIValue(nah, j, &yval);
ptaAddPt(ptah, x, yval);
}
for (j = 0; j < nm; j++) {
numaGetIValue(nam, j, &yval);
ptaAddPt(ptam, x, yval);
}
numaDestroy(&nah);
numaDestroy(&nam);
}
}
/* Make the Sel with those points */
sel = selCreateBrick(h, w, h / 2, w / 2, SEL_DONT_CARE);
nh = ptaGetCount(ptah);
for (i = 0; i < nh; i++) {
ptaGetIPt(ptah, i, &x, &y);
selSetElement(sel, y, x, SEL_HIT);
}
nm = ptaGetCount(ptam);
for (i = 0; i < nm; i++) {
ptaGetIPt(ptam, i, &x, &y);
selSetElement(sel, y, x, SEL_MISS);
}
pixDestroy(&pixfg);
pixDestroy(&pixbg);
ptaDestroy(&ptah);
ptaDestroy(&ptam);
return sel;
}
/*!
* \brief pixGenerateSelRandom()
*
* \param[in] pixs 1 bpp, typically small, to be used as a pattern
* \param[in] hitfract fraction of allowable fg pixels that are hits
* \param[in] missfract fraction of allowable bg pixels that are misses
* \param[in] distance min distance from boundary pixel; use 0 for default
* \param[in] toppix number of extra pixels of bg added above
* \param[in] botpix number of extra pixels of bg added below
* \param[in] leftpix number of extra pixels of bg added to left
* \param[in] rightpix number of extra pixels of bg added to right
* \param[out] ppixe [optional] input pix expanded by extra pixels
* \return sel hit-miss for input pattern, or NULL on error
*
* <pre>
* Notes:
* (1) Either of hitfract and missfract can be zero. If both are zero,
* the sel would be empty, and NULL is returned.
* (2) No elements are selected that are less than 'distance' pixels away
* from a boundary pixel of the same color. This makes the
* match much more robust to edge noise. Valid inputs of
* 'distance' are 0, 1, 2, 3 and 4. If distance is either 0 or
* greater than 4, we reset it to the default value.
* (3) The 4 numbers for adding rectangles of pixels outside the fg
* can be use if the pattern is expected to be surrounded by bg
* (white) pixels. On the other hand, if the pattern may be near
* other fg (black) components on some sides, use 0 for those sides.
* (4) The input pix, as extended by the extra pixels on selected sides,
* can optionally be returned. For debugging, call
* pixDisplayHitMissSel() to visualize the hit-miss sel superimposed
* on the generating bitmap.
* </pre>
*/
SEL *
pixGenerateSelRandom(PIX *pixs,
l_float32 hitfract,
l_float32 missfract,
l_int32 distance,
l_int32 toppix,
l_int32 botpix,
l_int32 leftpix,
l_int32 rightpix,
PIX **ppixe)
{
l_int32 ws, hs, w, h, x, y, i, j, thresh;
l_uint32 val;
PIX *pixt1, *pixt2, *pixfg, *pixbg;
SEL *seld, *sel;
PROCNAME("pixGenerateSelRandom");
if (ppixe) *ppixe = NULL;
if (!pixs)
return (SEL *)ERROR_PTR("pixs not defined", procName, NULL);
if (pixGetDepth(pixs) != 1)
return (SEL *)ERROR_PTR("pixs not 1 bpp", procName, NULL);
if (hitfract <= 0.0 && missfract <= 0.0)
return (SEL *)ERROR_PTR("no hits or misses", procName, NULL);
if (hitfract > 1.0 || missfract > 1.0)
return (SEL *)ERROR_PTR("fraction can't be > 1.0", procName, NULL);
if (distance <= 0)
distance = DEFAULT_DISTANCE_TO_BOUNDARY;
if (distance > MAX_DISTANCE_TO_BOUNDARY) {
L_WARNING("distance too large; setting to max value\n", procName);
distance = MAX_DISTANCE_TO_BOUNDARY;
}
/* Locate the foreground */
pixClipToForeground(pixs, &pixt1, NULL);
if (!pixt1)
return (SEL *)ERROR_PTR("pixt1 not made", procName, NULL);
ws = pixGetWidth(pixt1);
hs = pixGetHeight(pixt1);
w = ws;
h = hs;
/* Crop out a region including the foreground, and add pixels
* on sides depending on the side flags */
if (toppix || botpix || leftpix || rightpix) {
x = y = 0;
if (toppix) {
h += toppix;
y = toppix;
}
if (botpix)
h += botpix;
if (leftpix) {
w += leftpix;
x = leftpix;
}
if (rightpix)
w += rightpix;
pixt2 = pixCreate(w, h, 1);
pixRasterop(pixt2, x, y, ws, hs, PIX_SRC, pixt1, 0, 0);
} else {
pixt2 = pixClone(pixt1);
}
if (ppixe)
*ppixe = pixClone(pixt2);
pixDestroy(&pixt1);
/* Identify fg and bg pixels that are at least 'distance' pixels
* away from the boundary pixels in their set */
seld = selCreateBrick(2 * distance + 1, 2 * distance + 1,
distance, distance, SEL_HIT);
pixfg = pixErode(NULL, pixt2, seld);
pixbg = pixDilate(NULL, pixt2, seld);
pixInvert(pixbg, pixbg);
selDestroy(&seld);
pixDestroy(&pixt2);
/* Generate the sel from a random selection of these points */
sel = selCreateBrick(h, w, h / 2, w / 2, SEL_DONT_CARE);
if (hitfract > 0.0) {
thresh = (l_int32)(hitfract * (l_float64)RAND_MAX);
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
pixGetPixel(pixfg, j, i, &val);
if (val) {
if (rand() < thresh)
selSetElement(sel, i, j, SEL_HIT);
}
}
}
}
if (missfract > 0.0) {
thresh = (l_int32)(missfract * (l_float64)RAND_MAX);
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
pixGetPixel(pixbg, j, i, &val);
if (val) {
if (rand() < thresh)
selSetElement(sel, i, j, SEL_MISS);
}
}
}
}
pixDestroy(&pixfg);
pixDestroy(&pixbg);
return sel;
}
/*!
* pixFindSkewSweepAndSearchScorePivot()
*
* Input: pixs (1 bpp)
* &angle (<return> angle required to deskew; in degrees)
* &conf (<return> confidence given by ratio of max/min score)
* &endscore (<optional return> max score; use NULL to ignore)
* redsweep (sweep reduction factor = 1, 2, 4 or 8)
* redsearch (binary search reduction factor = 1, 2, 4 or 8;
* and must not exceed redsweep)
* sweepcenter (angle about which sweep is performed; in degrees)
* sweeprange (half the full range, taken about sweepcenter;
* in degrees)
* sweepdelta (angle increment of sweep; in degrees)
* minbsdelta (min binary search increment angle; in degrees)
* pivot (L_SHEAR_ABOUT_CORNER, L_SHEAR_ABOUT_CENTER)
* Return: 0 if OK, 1 on error or if angle measurment not valid
*
* Notes:
* (1) See notes in pixFindSkewSweepAndSearchScore().
* (2) This allows choice of shear pivoting from either the UL corner
* or the center. For small angles, the ability to discriminate
* angles is better with shearing from the UL corner. However,
* for large angles (say, greater than 20 degrees), it is better
* to shear about the center because a shear from the UL corner
* loses too much of the image.
*/
l_int32
pixFindSkewSweepAndSearchScorePivot(PIX *pixs,
l_float32 *pangle,
l_float32 *pconf,
l_float32 *pendscore,
l_int32 redsweep,
l_int32 redsearch,
l_float32 sweepcenter,
l_float32 sweeprange,
l_float32 sweepdelta,
l_float32 minbsdelta,
l_int32 pivot)
{
l_int32 ret, bzero, i, nangles, n, ratio, maxindex, minloc;
l_int32 width, height;
l_float32 deg2rad, theta, delta;
l_float32 sum, maxscore, maxangle;
l_float32 centerangle, leftcenterangle, rightcenterangle;
l_float32 lefttemp, righttemp;
l_float32 bsearchscore[5];
l_float32 minscore, minthresh;
l_float32 rangeleft;
NUMA *natheta, *nascore;
PIX *pixsw, *pixsch, *pixt1, *pixt2;
PROCNAME("pixFindSkewSweepAndSearchScorePivot");
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (pixGetDepth(pixs) != 1)
return ERROR_INT("pixs not 1 bpp", procName, 1);
if (!pangle)
return ERROR_INT("&angle not defined", procName, 1);
if (!pconf)
return ERROR_INT("&conf not defined", procName, 1);
if (redsweep != 1 && redsweep != 2 && redsweep != 4 && redsweep != 8)
return ERROR_INT("redsweep must be in {1,2,4,8}", procName, 1);
if (redsearch != 1 && redsearch != 2 && redsearch != 4 && redsearch != 8)
return ERROR_INT("redsearch must be in {1,2,4,8}", procName, 1);
if (redsearch > redsweep)
return ERROR_INT("redsearch must not exceed redsweep", procName, 1);
if (pivot != L_SHEAR_ABOUT_CORNER && pivot != L_SHEAR_ABOUT_CENTER)
return ERROR_INT("invalid pivot", procName, 1);
*pangle = 0.0;
*pconf = 0.0;
deg2rad = 3.1415926535 / 180.;
ret = 0;
/* Generate reduced image for binary search, if requested */
if (redsearch == 1)
pixsch = pixClone(pixs);
else if (redsearch == 2)
pixsch = pixReduceRankBinaryCascade(pixs, 1, 0, 0, 0);
else if (redsearch == 4)
pixsch = pixReduceRankBinaryCascade(pixs, 1, 1, 0, 0);
else /* redsearch == 8 */
pixsch = pixReduceRankBinaryCascade(pixs, 1, 1, 2, 0);
pixZero(pixsch, &bzero);
if (bzero) {
pixDestroy(&pixsch);
return 1;
}
/* Generate reduced image for sweep, if requested */
ratio = redsweep / redsearch;
if (ratio == 1) {
pixsw = pixClone(pixsch);
} else { /* ratio > 1 */
if (ratio == 2)
pixsw = pixReduceRankBinaryCascade(pixsch, 1, 0, 0, 0);
else if (ratio == 4)
pixsw = pixReduceRankBinaryCascade(pixsch, 1, 2, 0, 0);
else /* ratio == 8 */
pixsw = pixReduceRankBinaryCascade(pixsch, 1, 2, 2, 0);
}
pixt1 = pixCreateTemplate(pixsw);
if (ratio == 1)
pixt2 = pixClone(pixt1);
else
pixt2 = pixCreateTemplate(pixsch);
nangles = (l_int32)((2. * sweeprange) / sweepdelta + 1);
natheta = numaCreate(nangles);
nascore = numaCreate(nangles);
if (!pixsch || !pixsw) {
ret = ERROR_INT("pixsch and pixsw not both made", procName, 1);
goto cleanup;
}
if (!pixt1 || !pixt2) {
ret = ERROR_INT("pixt1 and pixt2 not both made", procName, 1);
goto cleanup;
}
if (!natheta || !nascore) {
ret = ERROR_INT("natheta and nascore not both made", procName, 1);
goto cleanup;
}
/* Do sweep */
rangeleft = sweepcenter - sweeprange;
for (i = 0; i < nangles; i++) {
theta = rangeleft + i * sweepdelta; /* degrees */
/* Shear pix and put the result in pixt1 */
if (pivot == L_SHEAR_ABOUT_CORNER)
pixVShearCorner(pixt1, pixsw, deg2rad * theta, L_BRING_IN_WHITE);
else
pixVShearCenter(pixt1, pixsw, deg2rad * theta, L_BRING_IN_WHITE);
/* Get score */
pixFindDifferentialSquareSum(pixt1, &sum);
#if DEBUG_PRINT_SCORES
L_INFO("sum(%7.2f) = %7.0f\n", procName, theta, sum);
#endif /* DEBUG_PRINT_SCORES */
/* Save the result in the output arrays */
numaAddNumber(nascore, sum);
numaAddNumber(natheta, theta);
}
/* Find the largest of the set (maxscore at maxangle) */
numaGetMax(nascore, &maxscore, &maxindex);
numaGetFValue(natheta, maxindex, &maxangle);
#if DEBUG_PRINT_SWEEP
L_INFO(" From sweep: angle = %7.3f, score = %7.3f\n", procName,
maxangle, maxscore);
#endif /* DEBUG_PRINT_SWEEP */
#if DEBUG_PLOT_SCORES
/* Plot the sweep result -- the scores versus rotation angle --
* using gnuplot with GPLOT_LINES (lines connecting data points). */
{GPLOT *gplot;
gplot = gplotCreate("sweep_output", GPLOT_PNG,
"Sweep. Variance of difference of ON pixels vs. angle",
"angle (deg)", "score");
gplotAddPlot(gplot, natheta, nascore, GPLOT_LINES, "plot1");
gplotAddPlot(gplot, natheta, nascore, GPLOT_POINTS, "plot2");
gplotMakeOutput(gplot);
gplotDestroy(&gplot);
}
#endif /* DEBUG_PLOT_SCORES */
/* Check if the max is at the end of the sweep. */
n = numaGetCount(natheta);
if (maxindex == 0 || maxindex == n - 1) {
L_WARNING("max found at sweep edge\n", procName);
goto cleanup;
}
/* Empty the numas for re-use */
numaEmpty(nascore);
numaEmpty(natheta);
/* Do binary search to find skew angle.
* First, set up initial three points. */
centerangle = maxangle;
if (pivot == L_SHEAR_ABOUT_CORNER) {
pixVShearCorner(pixt2, pixsch, deg2rad * centerangle, L_BRING_IN_WHITE);
pixFindDifferentialSquareSum(pixt2, &bsearchscore[2]);
pixVShearCorner(pixt2, pixsch, deg2rad * (centerangle - sweepdelta),
L_BRING_IN_WHITE);
pixFindDifferentialSquareSum(pixt2, &bsearchscore[0]);
pixVShearCorner(pixt2, pixsch, deg2rad * (centerangle + sweepdelta),
L_BRING_IN_WHITE);
pixFindDifferentialSquareSum(pixt2, &bsearchscore[4]);
} else {
pixVShearCenter(pixt2, pixsch, deg2rad * centerangle, L_BRING_IN_WHITE);
pixFindDifferentialSquareSum(pixt2, &bsearchscore[2]);
pixVShearCenter(pixt2, pixsch, deg2rad * (centerangle - sweepdelta),
L_BRING_IN_WHITE);
pixFindDifferentialSquareSum(pixt2, &bsearchscore[0]);
pixVShearCenter(pixt2, pixsch, deg2rad * (centerangle + sweepdelta),
L_BRING_IN_WHITE);
pixFindDifferentialSquareSum(pixt2, &bsearchscore[4]);
}
numaAddNumber(nascore, bsearchscore[2]);
numaAddNumber(natheta, centerangle);
numaAddNumber(nascore, bsearchscore[0]);
numaAddNumber(natheta, centerangle - sweepdelta);
numaAddNumber(nascore, bsearchscore[4]);
numaAddNumber(natheta, centerangle + sweepdelta);
/* Start the search */
delta = 0.5 * sweepdelta;
while (delta >= minbsdelta)
{
/* Get the left intermediate score */
leftcenterangle = centerangle - delta;
if (pivot == L_SHEAR_ABOUT_CORNER)
pixVShearCorner(pixt2, pixsch, deg2rad * leftcenterangle,
L_BRING_IN_WHITE);
else
pixVShearCenter(pixt2, pixsch, deg2rad * leftcenterangle,
L_BRING_IN_WHITE);
pixFindDifferentialSquareSum(pixt2, &bsearchscore[1]);
numaAddNumber(nascore, bsearchscore[1]);
numaAddNumber(natheta, leftcenterangle);
/* Get the right intermediate score */
rightcenterangle = centerangle + delta;
if (pivot == L_SHEAR_ABOUT_CORNER)
pixVShearCorner(pixt2, pixsch, deg2rad * rightcenterangle,
L_BRING_IN_WHITE);
else
pixVShearCenter(pixt2, pixsch, deg2rad * rightcenterangle,
L_BRING_IN_WHITE);
pixFindDifferentialSquareSum(pixt2, &bsearchscore[3]);
numaAddNumber(nascore, bsearchscore[3]);
numaAddNumber(natheta, rightcenterangle);
/* Find the maximum of the five scores and its location.
* Note that the maximum must be in the center
* three values, not in the end two. */
maxscore = bsearchscore[1];
maxindex = 1;
for (i = 2; i < 4; i++) {
if (bsearchscore[i] > maxscore) {
maxscore = bsearchscore[i];
maxindex = i;
}
}
/* Set up score array to interpolate for the next iteration */
lefttemp = bsearchscore[maxindex - 1];
righttemp = bsearchscore[maxindex + 1];
bsearchscore[2] = maxscore;
bsearchscore[0] = lefttemp;
bsearchscore[4] = righttemp;
/* Get new center angle and delta for next iteration */
centerangle = centerangle + delta * (maxindex - 2);
delta = 0.5 * delta;
}
*pangle = centerangle;
#if DEBUG_PRINT_SCORES
L_INFO(" Binary search score = %7.3f\n", procName, bsearchscore[2]);
#endif /* DEBUG_PRINT_SCORES */
if (pendscore) /* save if requested */
*pendscore = bsearchscore[2];
/* Return the ratio of Max score over Min score
* as a confidence value. Don't trust if the Min score
* is too small, which can happen if the image is all black
* with only a few white pixels interspersed. In that case,
* we get a contribution from the top and bottom edges when
* vertically sheared, but this contribution becomes zero when
* the shear angle is zero. For zero shear angle, the only
* contribution will be from the white pixels. We expect that
* the signal goes as the product of the (height * width^2),
* so we compute a (hopefully) normalized minimum threshold as
* a function of these dimensions. */
numaGetMin(nascore, &minscore, &minloc);
width = pixGetWidth(pixsch);
height = pixGetHeight(pixsch);
minthresh = MINSCORE_THRESHOLD_CONSTANT * width * width * height;
#if DEBUG_THRESHOLD
L_INFO(" minthresh = %10.2f, minscore = %10.2f\n", procName,
minthresh, minscore);
L_INFO(" maxscore = %10.2f\n", procName, maxscore);
#endif /* DEBUG_THRESHOLD */
if (minscore > minthresh)
*pconf = maxscore / minscore;
else
*pconf = 0.0;
/* Don't trust it if too close to the edge of the sweep
* range or if maxscore is small */
if ((centerangle > rangeleft + 2 * sweeprange - sweepdelta) ||
(centerangle < rangeleft + sweepdelta) ||
(maxscore < MIN_VALID_MAXSCORE))
*pconf = 0.0;
#if DEBUG_PRINT_BINARY
fprintf(stderr, "Binary search: angle = %7.3f, score ratio = %6.2f\n",
*pangle, *pconf);
fprintf(stderr, " max score = %8.0f\n", maxscore);
#endif /* DEBUG_PRINT_BINARY */
#if DEBUG_PLOT_SCORES
/* Plot the result -- the scores versus rotation angle --
* using gnuplot with GPLOT_POINTS. Because the data
* points are not ordered by theta (increasing or decreasing),
* using GPLOT_LINES would be confusing! */
{GPLOT *gplot;
gplot = gplotCreate("search_output", GPLOT_PNG,
"Binary search. Variance of difference of ON pixels vs. angle",
"angle (deg)", "score");
gplotAddPlot(gplot, natheta, nascore, GPLOT_POINTS, "plot1");
gplotMakeOutput(gplot);
gplotDestroy(&gplot);
}
#endif /* DEBUG_PLOT_SCORES */
cleanup:
pixDestroy(&pixsw);
pixDestroy(&pixsch);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
numaDestroy(&nascore);
numaDestroy(&natheta);
return ret;
}
/*!
* pixWriteStreamPng()
*
* Input: stream
* pix
* gamma (use 0.0 if gamma is not defined)
* Return: 0 if OK; 1 on error
*
* Notes:
* (1) If called from pixWriteStream(), the stream is positioned
* at the beginning of the file.
* (2) To do sequential writes of png format images to a stream,
* use pixWriteStreamPng() directly.
* (3) gamma is an optional png chunk. If no gamma value is to be
* placed into the file, use gamma = 0.0. Otherwise, if
* gamma > 0.0, its value is written into the header.
* (4) The use of gamma in png is highly problematic. For an illuminating
* discussion, see: http://hsivonen.iki.fi/png-gamma/
* (5) What is the effect/meaning of gamma in the png file? This
* gamma, which we can call the 'source' gamma, is the
* inverse of the gamma that was used in enhance.c to brighten
* or darken images. The 'source' gamma is supposed to indicate
* the intensity mapping that was done at the time the
* image was captured. Display programs typically apply a
* 'display' gamma of 2.2 to the output, which is intended
* to linearize the intensity based on the response of
* thermionic tubes (CRTs). Flat panel LCDs have typically
* been designed to give a similar response as CRTs (call it
* "backward compatibility"). The 'display' gamma is
* in some sense the inverse of the 'source' gamma.
* jpeg encoders attached to scanners and cameras will lighten
* the pixels, applying a gamma corresponding to approximately
* a square-root relation of output vs input:
* output = input^(gamma)
* where gamma is often set near 0.4545 (1/gamma is 2.2).
* This is stored in the image file. Then if the display
* program reads the gamma, it will apply a display gamma,
* typically about 2.2; the product is 1.0, and the
* display program produces a linear output. This works because
* the dark colors were appropriately boosted by the scanner,
* as described by the 'source' gamma, so they should not
* be further boosted by the display program.
* (6) As an example, with xv and display, if no gamma is stored,
* the program acts as if gamma were 0.4545, multiplies this by 2.2,
* and does a linear rendering. Taking this as a baseline
* brightness, if the stored gamma is:
* > 0.4545, the image is rendered lighter than baseline
* < 0.4545, the image is rendered darker than baseline
* In contrast, gqview seems to ignore the gamma chunk in png.
* (7) The only valid pixel depths in leptonica are 1, 2, 4, 8, 16
* and 32. However, it is possible, and in some cases desirable,
* to write out a png file using an rgb pix that has 24 bpp.
* For example, the open source xpdf SplashBitmap class generates
* 24 bpp rgb images. Consequently, we anble writing 24 bpp pix.
* To generate such a pix, you can make a 24 bpp pix without data
* and assign the data array to the pix; e.g.,
* pix = pixCreateHeader(w, h, 24);
* pixSetData(pix, rgbdata);
* See pixConvert32To24() for an example, where we get rgbdata
* from the 32 bpp pix. Caution: do not call pixSetPadBits(),
* because the alignment is wrong and you may erase part of the
* last pixel on each line.
*/
l_int32
pixWriteStreamPng(FILE *fp,
PIX *pix,
l_float32 gamma)
{
char commentstring[] = "Comment";
l_int32 i, j, k;
l_int32 wpl, d, cmflag;
l_int32 ncolors;
l_int32 *rmap, *gmap, *bmap;
l_uint32 *data, *ppixel;
png_byte bit_depth, color_type;
png_uint_32 w, h;
png_uint_32 xres, yres;
png_bytep *row_pointers;
png_bytep rowbuffer;
png_structp png_ptr;
png_infop info_ptr;
png_colorp palette;
PIX *pixt;
PIXCMAP *cmap;
char *text;
PROCNAME("pixWriteStreamPng");
if (!fp)
return ERROR_INT("stream not open", procName, 1);
if (!pix)
return ERROR_INT("pix not defined", procName, 1);
/* Allocate the 2 data structures */
if ((png_ptr = png_create_write_struct(PNG_LIBPNG_VER_STRING,
(png_voidp)NULL, NULL, NULL)) == NULL)
return ERROR_INT("png_ptr not made", procName, 1);
if ((info_ptr = png_create_info_struct(png_ptr)) == NULL) {
png_destroy_write_struct(&png_ptr, (png_infopp)NULL);
return ERROR_INT("info_ptr not made", procName, 1);
}
/* Set up png setjmp error handling */
if (setjmp(png_jmpbuf(png_ptr))) {
png_destroy_write_struct(&png_ptr, &info_ptr);
return ERROR_INT("internal png error", procName, 1);
}
png_init_io(png_ptr, fp);
/* With best zlib compression (9), get between 1 and 10% improvement
* over default (5), but the compression is 3 to 10 times slower.
* Our default compression is the zlib default (5). */
png_set_compression_level(png_ptr, var_ZLIB_COMPRESSION);
w = pixGetWidth(pix);
h = pixGetHeight(pix);
d = pixGetDepth(pix);
if ((cmap = pixGetColormap(pix)))
cmflag = 1;
else
cmflag = 0;
/* Set the color type and bit depth. */
if (d == 32 && var_PNG_WRITE_ALPHA == 1) {
bit_depth = 8;
color_type = PNG_COLOR_TYPE_RGBA; /* 6 */
cmflag = 0; /* ignore if it exists */
}
else if (d == 24 || d == 32) {
bit_depth = 8;
color_type = PNG_COLOR_TYPE_RGB; /* 2 */
cmflag = 0; /* ignore if it exists */
}
else {
bit_depth = d;
color_type = PNG_COLOR_TYPE_GRAY; /* 0 */
}
if (cmflag)
color_type = PNG_COLOR_TYPE_PALETTE; /* 3 */
#if DEBUG
fprintf(stderr, "cmflag = %d, bit_depth = %d, color_type = %d\n",
cmflag, bit_depth, color_type);
#endif /* DEBUG */
png_set_IHDR(png_ptr, info_ptr, w, h, bit_depth, color_type,
PNG_INTERLACE_NONE, PNG_COMPRESSION_TYPE_BASE,
PNG_FILTER_TYPE_BASE);
/* Store resolution in ppm, if known */
xres = (png_uint_32)(39.37 * (l_float32)pixGetXRes(pix) + 0.5);
yres = (png_uint_32)(39.37 * (l_float32)pixGetYRes(pix) + 0.5);
if ((xres == 0) || (yres == 0))
png_set_pHYs(png_ptr, info_ptr, 0, 0, PNG_RESOLUTION_UNKNOWN);
else
png_set_pHYs(png_ptr, info_ptr, xres, yres, PNG_RESOLUTION_METER);
if (cmflag) {
pixcmapToArrays(cmap, &rmap, &gmap, &bmap);
ncolors = pixcmapGetCount(cmap);
/* Make and save the palette */
if ((palette = (png_colorp)(CALLOC(ncolors, sizeof(png_color))))
== NULL)
return ERROR_INT("palette not made", procName, 1);
for (i = 0; i < ncolors; i++) {
palette[i].red = (png_byte)rmap[i];
palette[i].green = (png_byte)gmap[i];
palette[i].blue = (png_byte)bmap[i];
}
png_set_PLTE(png_ptr, info_ptr, palette, (int)ncolors);
FREE(rmap);
FREE(gmap);
FREE(bmap);
}
/* 0.4545 is treated as the default by some image
* display programs (not gqview). A value > 0.4545 will
* lighten an image as displayed by xv, display, etc. */
if (gamma > 0.0)
png_set_gAMA(png_ptr, info_ptr, (l_float64)gamma);
if ((text = pixGetText(pix))) {
png_text text_chunk;
text_chunk.compression = PNG_TEXT_COMPRESSION_NONE;
text_chunk.key = commentstring;
text_chunk.text = text;
text_chunk.text_length = strlen(text);
#ifdef PNG_ITXT_SUPPORTED
text_chunk.itxt_length = 0;
text_chunk.lang = NULL;
text_chunk.lang_key = NULL;
#endif
png_set_text(png_ptr, info_ptr, &text_chunk, 1);
}
/* Write header and palette info */
png_write_info(png_ptr, info_ptr);
if ((d != 32) && (d != 24)) { /* not rgb color */
/* Generate a temporary pix with bytes swapped.
* For a binary image, there are two conditions in
* which you must first invert the data for writing png:
* (a) no colormap
* (b) colormap with BLACK set to 0
* png writes binary with BLACK = 0, unless contradicted
* by a colormap. If the colormap has BLACK = "1"
* (typ. about 255), do not invert the data. If there
* is no colormap, you must invert the data to store
* in default BLACK = 0 state. */
if (d == 1 &&
(!cmap || (cmap && ((l_uint8 *)(cmap->array))[0] == 0x0))) {
pixt = pixInvert(NULL, pix);
pixEndianByteSwap(pixt);
}
else
pixt = pixEndianByteSwapNew(pix);
if (!pixt) {
png_destroy_write_struct(&png_ptr, &info_ptr);
return ERROR_INT("pixt not made", procName, 1);
}
/* Make and assign array of image row pointers */
if ((row_pointers = (png_bytep *)CALLOC(h, sizeof(png_bytep))) == NULL)
return ERROR_INT("row-pointers not made", procName, 1);
wpl = pixGetWpl(pixt);
data = pixGetData(pixt);
for (i = 0; i < h; i++)
row_pointers[i] = (png_bytep)(data + i * wpl);
png_set_rows(png_ptr, info_ptr, row_pointers);
/* Transfer the data */
png_write_image(png_ptr, row_pointers);
png_write_end(png_ptr, info_ptr);
if (cmflag)
FREE(palette);
FREE(row_pointers);
pixDestroy(&pixt);
png_destroy_write_struct(&png_ptr, &info_ptr);
return 0;
}
/* For rgb, compose and write a row at a time */
data = pixGetData(pix);
wpl = pixGetWpl(pix);
if (d == 24) { /* See note 7 above: special case of 24 bpp rgb */
for (i = 0; i < h; i++) {
ppixel = data + i * wpl;
png_write_rows(png_ptr, (png_bytepp)&ppixel, 1);
}
}
else { /* 32 bpp rgb and rgba */
if ((rowbuffer = (png_bytep)CALLOC(w, 4)) == NULL)
return ERROR_INT("rowbuffer not made", procName, 1);
for (i = 0; i < h; i++) {
ppixel = data + i * wpl;
for (j = k = 0; j < w; j++) {
rowbuffer[k++] = GET_DATA_BYTE(ppixel, COLOR_RED);
rowbuffer[k++] = GET_DATA_BYTE(ppixel, COLOR_GREEN);
rowbuffer[k++] = GET_DATA_BYTE(ppixel, COLOR_BLUE);
if (var_PNG_WRITE_ALPHA == 1)
rowbuffer[k++] = GET_DATA_BYTE(ppixel, L_ALPHA_CHANNEL);
ppixel++;
}
png_write_rows(png_ptr, &rowbuffer, 1);
}
FREE(rowbuffer);
}
png_write_end(png_ptr, info_ptr);
if (cmflag)
FREE(palette);
png_destroy_write_struct(&png_ptr, &info_ptr);
return 0;
}
/*!
* pixGetLocalSkewAngles()
*
* Input: pixs
* nslices (the number of horizontal overlapping slices; must
* be larger than 1 and not exceed 20; use 0 for default)
* redsweep (sweep reduction factor: 1, 2, 4 or 8;
* use 0 for default value)
* redsearch (search reduction factor: 1, 2, 4 or 8, and
* not larger than redsweep; use 0 for default value)
* sweeprange (half the full range, assumed about 0; in degrees;
* use 0.0 for default value)
* sweepdelta (angle increment of sweep; in degrees;
* use 0.0 for default value)
* minbsdelta (min binary search increment angle; in degrees;
* use 0.0 for default value)
* &a (<optional return> slope of skew as fctn of y)
* &b (<optional return> intercept at y=0 of skew as fctn of y)
* Return: naskew, or null on error
*
* Notes:
* (1) The local skew is measured in a set of overlapping strips.
* We then do a least square linear fit parameters to get
* the slope and intercept parameters a and b in
* skew-angle = a * y + b (degrees)
* for the local skew as a function of raster line y.
* This is then used to make naskew, which can be interpreted
* as the computed skew angle (in degrees) at the left edge
* of each raster line.
* (2) naskew can then be used to find the baselines of text, because
* each text line has a baseline that should intersect
* the left edge of the image with the angle given by this
* array, evaluated at the raster line of intersection.
*/
NUMA *
pixGetLocalSkewAngles(PIX *pixs,
l_int32 nslices,
l_int32 redsweep,
l_int32 redsearch,
l_float32 sweeprange,
l_float32 sweepdelta,
l_float32 minbsdelta,
l_float32 *pa,
l_float32 *pb)
{
l_int32 w, h, hs, i, ystart, yend, ovlap, npts;
l_float32 angle, conf, ycenter, a, b;
BOX *box;
NUMA *naskew;
PIX *pix;
PTA *pta;
PROCNAME("pixGetLocalSkewAngles");
if (!pixs)
return (NUMA *)ERROR_PTR("pixs not defined", procName, NULL);
if (nslices < 2 || nslices > 20)
nslices = DEFAULT_SLICES;
if (redsweep < 1 || redsweep > 8)
redsweep = DEFAULT_SWEEP_REDUCTION;
if (redsearch < 1 || redsearch > redsweep)
redsearch = DEFAULT_BS_REDUCTION;
if (sweeprange == 0.0)
sweeprange = DEFAULT_SWEEP_RANGE;
if (sweepdelta == 0.0)
sweepdelta = DEFAULT_SWEEP_DELTA;
if (minbsdelta == 0.0)
minbsdelta = DEFAULT_MINBS_DELTA;
h = pixGetHeight(pixs);
w = pixGetWidth(pixs);
hs = h / nslices;
ovlap = (l_int32)(OVERLAP_FRACTION * hs);
pta = ptaCreate(nslices);
for (i = 0; i < nslices; i++) {
ystart = L_MAX(0, hs * i - ovlap);
yend = L_MIN(h - 1, hs * (i + 1) + ovlap);
ycenter = (ystart + yend) / 2;
box = boxCreate(0, ystart, w, yend - ystart + 1);
pix = pixClipRectangle(pixs, box, NULL);
pixFindSkewSweepAndSearch(pix, &angle, &conf, redsweep, redsearch,
sweeprange, sweepdelta, minbsdelta);
if (conf > MIN_ALLOWED_CONFIDENCE)
ptaAddPt(pta, ycenter, angle);
pixDestroy(&pix);
boxDestroy(&box);
}
/* ptaWriteStream(stderr, pta, 0); */
/* Do linear least squares fit */
if ((npts = ptaGetCount(pta)) < 2) {
ptaDestroy(&pta);
return (NUMA *)ERROR_PTR("can't fit skew", procName, NULL);
}
ptaGetLinearLSF(pta, &a, &b, NULL);
if (pa) *pa = a;
if (pb) *pb = b;
/* Make skew angle array as function of raster line */
naskew = numaCreate(h);
for (i = 0; i < h; i++) {
angle = a * i + b;
numaAddNumber(naskew, angle);
}
#if DEBUG_PLOT
{ NUMA *nax, *nay;
GPLOT *gplot;
ptaGetArrays(pta, &nax, &nay);
gplot = gplotCreate("junkskew", GPLOT_X11, "skew as fctn of y",
"y (in raster lines from top)", "angle (in degrees)");
gplotAddPlot(gplot, NULL, naskew, GPLOT_POINTS, "linear lsf");
gplotAddPlot(gplot, nax, nay, GPLOT_POINTS, "actual data pts");
gplotMakeOutput(gplot);
gplotDestroy(&gplot);
numaDestroy(&nax);
numaDestroy(&nay);
}
#endif /* DEBUG_PLOT */
ptaDestroy(&pta);
return naskew;
}
/*!
* pixGetLocalSkewTransform()
*
* Input: pixs
* nslices (the number of horizontal overlapping slices; must
* be larger than 1 and not exceed 20; use 0 for default)
* redsweep (sweep reduction factor: 1, 2, 4 or 8;
* use 0 for default value)
* redsearch (search reduction factor: 1, 2, 4 or 8, and
* not larger than redsweep; use 0 for default value)
* sweeprange (half the full range, assumed about 0; in degrees;
* use 0.0 for default value)
* sweepdelta (angle increment of sweep; in degrees;
* use 0.0 for default value)
* minbsdelta (min binary search increment angle; in degrees;
* use 0.0 for default value)
* &ptas (<return> 4 points in the source)
* &ptad (<return> the corresponding 4 pts in the dest)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) This generates two pairs of points in the src, each pair
* corresponding to a pair of points that would lie along
* the same raster line in a transformed (dewarped) image.
* (2) The sets of 4 src and 4 dest points returned by this function
* can then be used, in a projective or bilinear transform,
* to remove keystoning in the src.
*/
l_int32
pixGetLocalSkewTransform(PIX *pixs,
l_int32 nslices,
l_int32 redsweep,
l_int32 redsearch,
l_float32 sweeprange,
l_float32 sweepdelta,
l_float32 minbsdelta,
PTA **pptas,
PTA **pptad)
{
l_int32 w, h, i;
l_float32 deg2rad, angr, angd, dely;
NUMA *naskew;
PTA *ptas, *ptad;
PROCNAME("pixGetLocalSkewTransform");
if (!pptas || !pptad)
return ERROR_INT("&ptas and &ptad not defined", procName, 1);
*pptas = *pptad = NULL;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (nslices < 2 || nslices > 20)
nslices = DEFAULT_SLICES;
if (redsweep < 1 || redsweep > 8)
redsweep = DEFAULT_SWEEP_REDUCTION;
if (redsearch < 1 || redsearch > redsweep)
redsearch = DEFAULT_BS_REDUCTION;
if (sweeprange == 0.0)
sweeprange = DEFAULT_SWEEP_RANGE;
if (sweepdelta == 0.0)
sweepdelta = DEFAULT_SWEEP_DELTA;
if (minbsdelta == 0.0)
minbsdelta = DEFAULT_MINBS_DELTA;
naskew = pixGetLocalSkewAngles(pixs, nslices, redsweep, redsearch,
sweeprange, sweepdelta, minbsdelta,
NULL, NULL);
if (!naskew)
return ERROR_INT("naskew not made", procName, 1);
deg2rad = 3.14159265 / 180.;
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
ptas = ptaCreate(4);
ptad = ptaCreate(4);
*pptas = ptas;
*pptad = ptad;
/* Find i for skew line that intersects LHS at i and RHS at h / 20 */
for (i = 0; i < h; i++) {
numaGetFValue(naskew, i, &angd);
angr = angd * deg2rad;
dely = w * tan(angr);
if (i - dely > 0.05 * h)
break;
}
ptaAddPt(ptas, 0, i);
ptaAddPt(ptas, w - 1, i - dely);
ptaAddPt(ptad, 0, i);
ptaAddPt(ptad, w - 1, i);
/* Find i for skew line that intersects LHS at i and RHS at 19h / 20 */
for (i = h - 1; i > 0; i--) {
numaGetFValue(naskew, i, &angd);
angr = angd * deg2rad;
dely = w * tan(angr);
if (i - dely < 0.95 * h)
break;
}
ptaAddPt(ptas, 0, i);
ptaAddPt(ptas, w - 1, i - dely);
ptaAddPt(ptad, 0, i);
ptaAddPt(ptad, w - 1, i);
numaDestroy(&naskew);
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 w, 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;
}
/* 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);
w = pixGetWidth(pixs);
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_X11, "junkdiff", "difference");
/* 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);
}
FREE(array);
if (debug) { /* show the raster locations for the peaks */
gplot = gplotCreate("junkloc", GPLOT_X11, "Peak locations",
"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("junkbaselines", pixd, IFF_PNG);
pixDestroy(&pixd);
}
}
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
numaDestroy(&nasum);
numaDestroy(&nadiff);
numaDestroy(&naval);
return naloc;
}
/*!
* pixaDisplayTiled()
*
* Input: pixa
* maxwidth (of output image)
* background (0 for white, 1 for black)
* spacing
* Return: pix of tiled images, or null on error
*
* Notes:
* (1) This saves a pixa to a single image file of width not to
* exceed maxwidth, with background color either white or black,
* and with each subimage spaced on a regular lattice.
* (2) The lattice size is determined from the largest width and height,
* separately, of all pix in the pixa.
* (3) All pix in the pixa must be of equal depth.
* (4) If any pix has a colormap, all pix are rendered in rgb.
* (5) Careful: because no components are omitted, this is
* dangerous if there are thousands of small components and
* one or more very large one, because the size of the
* resulting pix can be huge!
*/
PIX *
pixaDisplayTiled(PIXA *pixa,
l_int32 maxwidth,
l_int32 background,
l_int32 spacing)
{
l_int32 w, h, wmax, hmax, wd, hd, d, hascmap;
l_int32 i, j, n, ni, ncols, nrows;
l_int32 ystart, xstart, wt, ht;
PIX *pix, *pixt, *pixd;
PIXA *pixat;
PROCNAME("pixaDisplayTiled");
if (!pixa)
return (PIX *)ERROR_PTR("pixa not defined", procName, NULL);
/* If any pix have colormaps, generate rgb */
if ((n = pixaGetCount(pixa)) == 0)
return (PIX *)ERROR_PTR("no components", procName, NULL);
pixaAnyColormaps(pixa, &hascmap);
if (hascmap) {
pixat = pixaCreate(n);
for (i = 0; i < n; i++) {
pixt = pixaGetPix(pixa, i, L_CLONE);
pix = pixConvertTo32(pixt);
pixaAddPix(pixat, pix, L_INSERT);
pixDestroy(&pixt);
}
}
else
pixat = pixaCopy(pixa, L_CLONE);
/* Find the largest width and height of the subimages */
wmax = hmax = 0;
for (i = 0; i < n; i++) {
pix = pixaGetPix(pixat, i, L_CLONE);
pixGetDimensions(pix, &w, &h, NULL);
if (i == 0)
d = pixGetDepth(pix);
else if (d != pixGetDepth(pix)) {
pixDestroy(&pix);
pixaDestroy(&pixat);
return (PIX *)ERROR_PTR("depths not equal", procName, NULL);
}
if (w > wmax)
wmax = w;
if (h > hmax)
hmax = h;
pixDestroy(&pix);
}
/* Get the number of rows and columns and the output image size */
spacing = L_MAX(spacing, 0);
ncols = (l_int32)((l_float32)(maxwidth - spacing) /
(l_float32)(wmax + spacing));
nrows = (n + ncols - 1) / ncols;
wd = wmax * ncols + spacing * (ncols + 1);
hd = hmax * nrows + spacing * (nrows + 1);
if ((pixd = pixCreate(wd, hd, d)) == NULL) {
pixaDestroy(&pixat);
return (PIX *)ERROR_PTR("pixd not made", procName, NULL);
}
#if 0
fprintf(stderr, " nrows = %d, ncols = %d, wmax = %d, hmax = %d\n",
nrows, ncols, wmax, hmax);
fprintf(stderr, " space = %d, wd = %d, hd = %d, n = %d\n",
space, wd, hd, n);
#endif
/* Reset the background color if necessary */
if ((background == 1 && d == 1) || (background == 0 && d != 1))
pixSetAll(pixd);
/* Blit the images to the dest */
for (i = 0, ni = 0; i < nrows; i++) {
ystart = spacing + i * (hmax + spacing);
for (j = 0; j < ncols && ni < n; j++, ni++) {
xstart = spacing + j * (wmax + spacing);
pix = pixaGetPix(pixat, ni, L_CLONE);
wt = pixGetWidth(pix);
ht = pixGetHeight(pix);
pixRasterop(pixd, xstart, ystart, wt, ht, PIX_SRC, pix, 0, 0);
pixDestroy(&pix);
}
}
pixaDestroy(&pixat);
return pixd;
}
/*!
* \brief pixGenerateSelBoundary()
*
* \param[in] pixs 1 bpp, typically small, to be used as a pattern
* \param[in] hitdist min distance from fg boundary pixel
* \param[in] missdist min distance from bg boundary pixel
* \param[in] hitskip number of boundary pixels skipped between hits
* \param[in] missskip number of boundary pixels skipped between misses
* \param[in] topflag flag for extra pixels of bg added above
* \param[in] botflag flag for extra pixels of bg added below
* \param[in] leftflag flag for extra pixels of bg added to left
* \param[in] rightflag flag for extra pixels of bg added to right
* \param[out] ppixe [optional] input pix expanded by extra pixels
* \return sel hit-miss for input pattern, or NULL on error
*
* <pre>
* Notes:
* (1) All fg elements selected are exactly hitdist pixels away from
* the nearest fg boundary pixel, and ditto for bg elements.
* Valid inputs of hitdist and missdist are 0, 1, 2, 3 and 4.
* For example, a hitdist of 0 puts the hits at the fg boundary.
* Usually, the distances should be > 0 avoid the effect of
* noise at the boundary.
* (2) Set hitskip < 0 if no hits are to be used. Ditto for missskip.
* If both hitskip and missskip are < 0, the sel would be empty,
* and NULL is returned.
* (3) The 4 flags determine whether the sel is increased on that side
* to allow bg misses to be placed all along that boundary.
* The increase in sel size on that side is the minimum necessary
* to allow the misses to be placed at mindist. For text characters,
* the topflag and botflag are typically set to 1, and the leftflag
* and rightflag to 0.
* (4) The input pix, as extended by the extra pixels on selected sides,
* can optionally be returned. For debugging, call
* pixDisplayHitMissSel() to visualize the hit-miss sel superimposed
* on the generating bitmap.
* (5) This is probably the best of the three sel generators, in the
* sense that you have the most flexibility with the smallest number
* of hits and misses.
* </pre>
*/
SEL *
pixGenerateSelBoundary(PIX *pixs,
l_int32 hitdist,
l_int32 missdist,
l_int32 hitskip,
l_int32 missskip,
l_int32 topflag,
l_int32 botflag,
l_int32 leftflag,
l_int32 rightflag,
PIX **ppixe)
{
l_int32 ws, hs, w, h, x, y, ix, iy, i, npt;
PIX *pixt1, *pixt2, *pixt3, *pixfg, *pixbg;
SEL *selh, *selm, *sel_3, *sel;
PTA *ptah, *ptam;
PROCNAME("pixGenerateSelBoundary");
if (ppixe) *ppixe = NULL;
if (!pixs)
return (SEL *)ERROR_PTR("pixs not defined", procName, NULL);
if (pixGetDepth(pixs) != 1)
return (SEL *)ERROR_PTR("pixs not 1 bpp", procName, NULL);
if (hitdist < 0 || hitdist > 4 || missdist < 0 || missdist > 4)
return (SEL *)ERROR_PTR("dist not in {0 .. 4}", procName, NULL);
if (hitskip < 0 && missskip < 0)
return (SEL *)ERROR_PTR("no hits or misses", procName, NULL);
/* Locate the foreground */
pixClipToForeground(pixs, &pixt1, NULL);
if (!pixt1)
return (SEL *)ERROR_PTR("pixt1 not made", procName, NULL);
ws = pixGetWidth(pixt1);
hs = pixGetHeight(pixt1);
w = ws;
h = hs;
/* Crop out a region including the foreground, and add pixels
* on sides depending on the side flags */
if (topflag || botflag || leftflag || rightflag) {
x = y = 0;
if (topflag) {
h += missdist + 1;
y = missdist + 1;
}
if (botflag)
h += missdist + 1;
if (leftflag) {
w += missdist + 1;
x = missdist + 1;
}
if (rightflag)
w += missdist + 1;
pixt2 = pixCreate(w, h, 1);
pixRasterop(pixt2, x, y, ws, hs, PIX_SRC, pixt1, 0, 0);
} else {
pixt2 = pixClone(pixt1);
}
if (ppixe)
*ppixe = pixClone(pixt2);
pixDestroy(&pixt1);
/* Identify fg and bg pixels that are exactly hitdist and
* missdist (rsp) away from the boundary pixels in their set.
* Then get a subsampled set of these points. */
sel_3 = selCreateBrick(3, 3, 1, 1, SEL_HIT);
if (hitskip >= 0) {
selh = selCreateBrick(2 * hitdist + 1, 2 * hitdist + 1,
hitdist, hitdist, SEL_HIT);
pixt3 = pixErode(NULL, pixt2, selh);
pixfg = pixErode(NULL, pixt3, sel_3);
pixXor(pixfg, pixfg, pixt3);
ptah = pixSubsampleBoundaryPixels(pixfg, hitskip);
pixDestroy(&pixt3);
pixDestroy(&pixfg);
selDestroy(&selh);
}
if (missskip >= 0) {
selm = selCreateBrick(2 * missdist + 1, 2 * missdist + 1,
missdist, missdist, SEL_HIT);
pixt3 = pixDilate(NULL, pixt2, selm);
pixbg = pixDilate(NULL, pixt3, sel_3);
pixXor(pixbg, pixbg, pixt3);
ptam = pixSubsampleBoundaryPixels(pixbg, missskip);
pixDestroy(&pixt3);
pixDestroy(&pixbg);
selDestroy(&selm);
}
selDestroy(&sel_3);
pixDestroy(&pixt2);
/* Generate the hit-miss sel from these point */
sel = selCreateBrick(h, w, h / 2, w / 2, SEL_DONT_CARE);
if (hitskip >= 0) {
npt = ptaGetCount(ptah);
for (i = 0; i < npt; i++) {
ptaGetIPt(ptah, i, &ix, &iy);
selSetElement(sel, iy, ix, SEL_HIT);
}
}
if (missskip >= 0) {
npt = ptaGetCount(ptam);
for (i = 0; i < npt; i++) {
ptaGetIPt(ptam, i, &ix, &iy);
selSetElement(sel, iy, ix, SEL_MISS);
}
}
ptaDestroy(&ptah);
ptaDestroy(&ptam);
return sel;
}
/*!
* pixaaDisplayByPixa()
*
* Input: pixaa
* xspace between pix in pixa
* yspace between pixa
* max width of output pix
* Return: pix, or null on error
*
* Notes:
* (1) Displays each pixa on a line (or set of lines),
* in order from top to bottom. Within each pixa,
* the pix are displayed in order from left to right.
* (2) The size of each pix in each pixa is assumed to be
* approximately equal to the size of the first pix in
* the pixa. If this assumption is not correct, this
* function will not work properly.
* (3) This ignores the boxa of the pixaa.
*/
PIX *
pixaaDisplayByPixa(PIXAA *pixaa,
l_int32 xspace,
l_int32 yspace,
l_int32 maxw)
{
l_int32 i, j, npixa, npix;
l_int32 width, height, depth, nlines, lwidth;
l_int32 x, y, w, h, w0, h0;
PIX *pixt, *pixd;
PIXA *pixa;
PROCNAME("pixaaDisplayByPixa");
if (!pixaa)
return (PIX *)ERROR_PTR("pixaa not defined", procName, NULL);
if ((npixa = pixaaGetCount(pixaa)) == 0)
return (PIX *)ERROR_PTR("no components", procName, NULL);
/* Get size of output pix. The width is the minimum of the
* maxw and the largest pixa line width. The height is whatever
* it needs to be to accommodate all pixa. */
height = 2 * yspace;
width = 0;
for (i = 0; i < npixa; i++) {
pixa = pixaaGetPixa(pixaa, i, L_CLONE);
npix = pixaGetCount(pixa);
pixt = pixaGetPix(pixa, 0, L_CLONE);
if (i == 0)
depth = pixGetDepth(pixt);
w = pixGetWidth(pixt);
lwidth = npix * (w + xspace);
nlines = (lwidth + maxw - 1) / maxw;
if (nlines > 1)
width = maxw;
else
width = L_MAX(lwidth, width);
height += nlines * (pixGetHeight(pixt) + yspace);
pixDestroy(&pixt);
pixaDestroy(&pixa);
}
if ((pixd = pixCreate(width, height, depth)) == NULL)
return (PIX *)ERROR_PTR("pixd not made", procName, NULL);
/* Now layout the pix by pixa */
y = yspace;
for (i = 0; i < npixa; i++) {
x = 0;
pixa = pixaaGetPixa(pixaa, i, L_CLONE);
npix = pixaGetCount(pixa);
for (j = 0; j < npix; j++) {
pixt = pixaGetPix(pixa, j, L_CLONE);
if (j == 0) {
w0 = pixGetWidth(pixt);
h0 = pixGetHeight(pixt);
}
w = pixGetWidth(pixt);
if (width == maxw && x + w >= maxw) {
x = 0;
y += h0 + yspace;
}
h = pixGetHeight(pixt);
pixRasterop(pixd, x, y, w, h, PIX_PAINT, pixt, 0, 0);
pixDestroy(&pixt);
x += w0 + xspace;
}
y += h0 + yspace;
pixaDestroy(&pixa);
}
return pixd;
}
/*!
* \brief pixGetRunsOnLine()
*
* \param[in] pixs 1 bpp
* \param[in] x1, y1, x2, y2
* \return numa, or NULL on error
*
* <pre>
* Notes:
* (1) Action: this function uses the bresenham algorithm to compute
* the pixels along the specified line. It returns a Numa of the
* runlengths of the fg (black) and bg (white) runs, always
* starting with a white run.
* (2) If the first pixel on the line is black, the length of the
* first returned run (which is white) is 0.
* </pre>
*/
NUMA *
pixGetRunsOnLine(PIX *pixs,
l_int32 x1,
l_int32 y1,
l_int32 x2,
l_int32 y2)
{
l_int32 w, h, x, y, npts;
l_int32 i, runlen, preval;
l_uint32 val;
NUMA *numa;
PTA *pta;
PROCNAME("pixGetRunsOnLine");
if (!pixs)
return (NUMA *)ERROR_PTR("pixs not defined", procName, NULL);
if (pixGetDepth(pixs) != 1)
return (NUMA *)ERROR_PTR("pixs not 1 bpp", procName, NULL);
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
if (x1 < 0 || x1 >= w)
return (NUMA *)ERROR_PTR("x1 not valid", procName, NULL);
if (x2 < 0 || x2 >= w)
return (NUMA *)ERROR_PTR("x2 not valid", procName, NULL);
if (y1 < 0 || y1 >= h)
return (NUMA *)ERROR_PTR("y1 not valid", procName, NULL);
if (y2 < 0 || y2 >= h)
return (NUMA *)ERROR_PTR("y2 not valid", procName, NULL);
if ((pta = generatePtaLine(x1, y1, x2, y2)) == NULL)
return (NUMA *)ERROR_PTR("pta not made", procName, NULL);
if ((npts = ptaGetCount(pta)) == 0) {
ptaDestroy(&pta);
return (NUMA *)ERROR_PTR("pta has no pts", procName, NULL);
}
if ((numa = numaCreate(0)) == NULL) {
ptaDestroy(&pta);
return (NUMA *)ERROR_PTR("numa not made", procName, NULL);
}
for (i = 0; i < npts; i++) {
ptaGetIPt(pta, i, &x, &y);
pixGetPixel(pixs, x, y, &val);
if (i == 0) {
if (val == 1) { /* black pixel; append white run of size 0 */
numaAddNumber(numa, 0);
}
preval = val;
runlen = 1;
continue;
}
if (val == preval) { /* extend current run */
preval = val;
runlen++;
} else { /* end previous run */
numaAddNumber(numa, runlen);
preval = val;
runlen = 1;
}
}
numaAddNumber(numa, runlen); /* append last run */
ptaDestroy(&pta);
return numa;
}
main(int argc,
char **argv)
{
l_int32 i, j, w, h, same, width, height, cx, cy;
l_uint32 val;
PIX *pixs, *pixse, *pixd1, *pixd2;
SEL *sel;
static char mainName[] = "rasterop_reg";
if (argc != 1)
exit(ERROR_INT(" Syntax: rasterop_reg", mainName, 1));
if ((pixs = pixRead("feyn.tif")) == NULL)
exit(ERROR_INT("pix not made", mainName, 1));
for (width = MINW; width <= MAXW; width++) {
for (height = MINH; height <= MAXH; height++) {
cx = width / 2;
cy = height / 2;
/* dilate using an actual sel */
sel = selCreateBrick(height, width, cy, cx, SEL_HIT);
pixd1 = pixDilate(NULL, pixs, sel);
/* dilate using a pix as a sel */
pixse = pixCreate(width, height, 1);
pixSetAll(pixse);
pixd2 = pixCopy(NULL, pixs);
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
pixGetPixel(pixs, j, i, &val);
if (val)
pixRasterop(pixd2, j - cx, i - cy, width, height,
PIX_SRC | PIX_DST, pixse, 0, 0);
}
}
pixEqual(pixd1, pixd2, &same);
if (same == 1)
fprintf(stderr, "Correct: results for (%d,%d) are identical!\n",
width, height);
else {
fprintf(stderr, "Error: results are different!\n");
fprintf(stderr, "SE: width = %d, height = %d\n", width, height);
pixWrite("/tmp/junkout1.png", pixd1, IFF_PNG);
pixWrite("/tmp/junkout2.png", pixd2, IFF_PNG);
exit(1);
}
pixDestroy(&pixse);
pixDestroy(&pixd1);
pixDestroy(&pixd2);
selDestroy(&sel);
}
}
pixDestroy(&pixs);
exit(0);
}
// Adds sub-pixel resolution EdgeOffsets for the outline if the supplied
// pix is 8-bit. Does nothing otherwise.
// Operation: Consider the following near-horizontal line:
// _________
// |________
// |________
// At *every* position along this line, the gradient direction will be close
// to vertical. Extrapoaltion/interpolation of the position of the threshold
// that was used to binarize the image gives a more precise vertical position
// for each horizontal step, and the conflict in step direction and gradient
// direction can be used to ignore the vertical steps.
void C_OUTLINE::ComputeEdgeOffsets(int threshold, Pix* pix) {
if (pixGetDepth(pix) != 8) return;
const l_uint32* data = pixGetData(pix);
int wpl = pixGetWpl(pix);
int width = pixGetWidth(pix);
int height = pixGetHeight(pix);
bool negative = flag(COUT_INVERSE);
delete [] offsets;
offsets = new EdgeOffset[stepcount];
ICOORD pos = start;
ICOORD prev_gradient;
ComputeGradient(data, wpl, pos.x(), height - pos.y(), width, height,
&prev_gradient);
for (int s = 0; s < stepcount; ++s) {
ICOORD step_vec = step(s);
TPOINT pt1(pos);
pos += step_vec;
TPOINT pt2(pos);
ICOORD next_gradient;
ComputeGradient(data, wpl, pos.x(), height - pos.y(), width, height,
&next_gradient);
// Use the sum of the prev and next as the working gradient.
ICOORD gradient = prev_gradient + next_gradient;
// best_diff will be manipulated to be always positive.
int best_diff = 0;
// offset will be the extrapolation of the location of the greyscale
// threshold from the edge with the largest difference, relative to the
// location of the binary edge.
int offset = 0;
if (pt1.y == pt2.y && abs(gradient.y()) * 2 >= abs(gradient.x())) {
// Horizontal step. diff_sign == 1 indicates black above.
int diff_sign = (pt1.x > pt2.x) == negative ? 1 : -1;
int x = MIN(pt1.x, pt2.x);
int y = height - pt1.y;
int best_sum = 0;
int best_y = y;
EvaluateVerticalDiff(data, wpl, diff_sign, x, y, height,
&best_diff, &best_sum, &best_y);
// Find the strongest edge.
int test_y = y;
do {
++test_y;
} while (EvaluateVerticalDiff(data, wpl, diff_sign, x, test_y, height,
&best_diff, &best_sum, &best_y));
test_y = y;
do {
--test_y;
} while (EvaluateVerticalDiff(data, wpl, diff_sign, x, test_y, height,
&best_diff, &best_sum, &best_y));
offset = diff_sign * (best_sum / 2 - threshold) +
(y - best_y) * best_diff;
} else if (pt1.x == pt2.x && abs(gradient.x()) * 2 >= abs(gradient.y())) {
// Vertical step. diff_sign == 1 indicates black on the left.
int diff_sign = (pt1.y > pt2.y) == negative ? 1 : -1;
int x = pt1.x;
int y = height - MAX(pt1.y, pt2.y);
const l_uint32* line = pixGetData(pix) + y * wpl;
int best_sum = 0;
int best_x = x;
EvaluateHorizontalDiff(line, diff_sign, x, width,
&best_diff, &best_sum, &best_x);
// Find the strongest edge.
int test_x = x;
do {
++test_x;
} while (EvaluateHorizontalDiff(line, diff_sign, test_x, width,
&best_diff, &best_sum, &best_x));
test_x = x;
do {
--test_x;
} while (EvaluateHorizontalDiff(line, diff_sign, test_x, width,
&best_diff, &best_sum, &best_x));
offset = diff_sign * (threshold - best_sum / 2) +
(best_x - x) * best_diff;
}
offsets[s].offset_numerator =
static_cast<inT8>(ClipToRange(offset, -MAX_INT8, MAX_INT8));
offsets[s].pixel_diff = static_cast<uinT8>(ClipToRange(best_diff, 0 ,
MAX_UINT8));
if (negative) gradient = -gradient;
// Compute gradient angle quantized to 256 directions, rotated by 64 (pi/2)
// to convert from gradient direction to edge direction.
offsets[s].direction =
Modulo(FCOORD::binary_angle_plus_pi(gradient.angle()) + 64, 256);
prev_gradient = next_gradient;
}
}
/**
* Segment the page according to the current value of tessedit_pageseg_mode.
* If the pix_binary_ member is not NULL, it is used as the source image,
* and copied to image, otherwise it just uses image as the input.
* On return the blocks list owns all the constructed page layout.
*/
int Tesseract::SegmentPage(const STRING* input_file,
IMAGE* image, BLOCK_LIST* blocks) {
int width = image->get_xsize();
int height = image->get_ysize();
int resolution = image->get_res();
#ifdef HAVE_LIBLEPT
if (pix_binary_ != NULL) {
width = pixGetWidth(pix_binary_);
height = pixGetHeight(pix_binary_);
resolution = pixGetXRes(pix_binary_);
}
#endif
// Zero resolution messes up the algorithms, so make sure it is credible.
if (resolution < kMinCredibleResolution)
resolution = kDefaultResolution;
// Get page segmentation mode.
PageSegMode pageseg_mode = static_cast<PageSegMode>(
static_cast<int>(tessedit_pageseg_mode));
// If a UNLV zone file can be found, use that instead of segmentation.
if (pageseg_mode != tesseract::PSM_AUTO &&
input_file != NULL && input_file->length() > 0) {
STRING name = *input_file;
const char* lastdot = strrchr(name.string(), '.');
if (lastdot != NULL)
name[lastdot - name.string()] = '\0';
read_unlv_file(name, width, height, blocks);
}
bool single_column = pageseg_mode > PSM_AUTO;
if (blocks->empty()) {
// No UNLV file present. Work according to the PageSegMode.
// First make a single block covering the whole image.
BLOCK_IT block_it(blocks);
BLOCK* block = new BLOCK("", TRUE, 0, 0, 0, 0, width, height);
block_it.add_to_end(block);
} else {
// UNLV file present. Use PSM_SINGLE_COLUMN.
pageseg_mode = PSM_SINGLE_COLUMN;
}
TO_BLOCK_LIST land_blocks, port_blocks;
TBOX page_box;
if (pageseg_mode <= PSM_SINGLE_COLUMN) {
if (AutoPageSeg(width, height, resolution, single_column,
image, blocks, &port_blocks) < 0) {
return -1;
}
// To create blobs from the image region bounds uncomment this line:
// port_blocks.clear(); // Uncomment to go back to the old mode.
} else {
#if HAVE_LIBLEPT
image->FromPix(pix_binary_);
#endif
deskew_ = FCOORD(1.0f, 0.0f);
reskew_ = FCOORD(1.0f, 0.0f);
}
if (blocks->empty()) {
tprintf("Empty page\n");
return 0; // AutoPageSeg found an empty page.
}
if (port_blocks.empty()) {
// AutoPageSeg was not used, so we need to find_components first.
find_components(blocks, &land_blocks, &port_blocks, &page_box);
} else {
// AutoPageSeg does not need to find_components as it did that already.
page_box.set_left(0);
page_box.set_bottom(0);
page_box.set_right(width);
page_box.set_top(height);
// Filter_blobs sets up the TO_BLOCKs the same as find_components does.
filter_blobs(page_box.topright(), &port_blocks, true);
}
TO_BLOCK_IT to_block_it(&port_blocks);
ASSERT_HOST(!port_blocks.empty());
TO_BLOCK* to_block = to_block_it.data();
if (pageseg_mode <= PSM_SINGLE_BLOCK ||
to_block->line_size < 2) {
// For now, AUTO, SINGLE_COLUMN and SINGLE_BLOCK all map to the old
// textord. The difference is the number of blocks and how the are made.
textord_page(page_box.topright(), blocks, &land_blocks, &port_blocks,
this);
} else {
// SINGLE_LINE, SINGLE_WORD and SINGLE_CHAR all need a single row.
float gradient = make_single_row(page_box.topright(),
to_block, &port_blocks, this);
if (pageseg_mode == PSM_SINGLE_LINE) {
// SINGLE_LINE uses the old word maker on the single line.
make_words(page_box.topright(), gradient, blocks,
&land_blocks, &port_blocks, this);
} else {
// SINGLE_WORD and SINGLE_CHAR cram all the blobs into a
// single word, and in SINGLE_CHAR mode, all the outlines
// go in a single blob.
make_single_word(pageseg_mode == PSM_SINGLE_CHAR,
to_block->get_rows(), to_block->block->row_list());
}
}
return 0;
}
int main(int argc,
char **argv)
{
char textstr[256];
l_int32 w, h, d, i;
l_uint32 srcval, dstval;
l_float32 scalefact, sat, fract;
L_BMF *bmf8;
L_KERNEL *kel;
NUMA *na;
PIX *pix, *pixs, *pixs1, *pixs2, *pixd;
PIX *pixt0, *pixt1, *pixt2, *pixt3, *pixt4;
PIXA *pixa, *pixaf;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
pix = pixRead(filein);
pixGetDimensions(pix, &w, &h, &d);
if (d != 32)
return ERROR_INT("file not 32 bpp", argv[0], 1);
scalefact = (l_float32)WIDTH / (l_float32)w;
pixs = pixScale(pix, scalefact, scalefact);
w = pixGetWidth(pixs);
pixaf = pixaCreate(5);
/* TRC: vary gamma */
pixa = pixaCreate(20);
for (i = 0; i < 20; i++) {
pixt0 = pixGammaTRC(NULL, pixs, 0.3 + 0.15 * i, 0, 255);
pixaAddPix(pixa, pixt0, L_INSERT);
}
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, w, 5, 0, 10, 2);
pixSaveTiled(pixt1, pixaf, 1.0, 1, 20, 32);
regTestWritePixAndCheck(rp, pixt1, IFF_PNG); /* 0 */
pixDisplayWithTitle(pixt1, 0, 100, "TRC Gamma", rp->display);
pixDestroy(&pixt1);
pixaDestroy(&pixa);
/* TRC: vary black point */
pixa = pixaCreate(20);
for (i = 0; i < 20; i++) {
pixt0 = pixGammaTRC(NULL, pixs, 1.0, 5 * i, 255);
pixaAddPix(pixa, pixt0, L_INSERT);
}
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, w, 5, 0, 10, 2);
pixSaveTiled(pixt1, pixaf, 1.0, 1, 20, 0);
regTestWritePixAndCheck(rp, pixt1, IFF_PNG); /* 1 */
pixDisplayWithTitle(pixt1, 300, 100, "TRC", rp->display);
pixDestroy(&pixt1);
pixaDestroy(&pixa);
/* Vary hue */
pixa = pixaCreate(20);
for (i = 0; i < 20; i++) {
pixt0 = pixModifyHue(NULL, pixs, 0.01 + 0.05 * i);
pixaAddPix(pixa, pixt0, L_INSERT);
}
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, w, 5, 0, 10, 2);
pixSaveTiled(pixt1, pixaf, 1.0, 1, 20, 0);
regTestWritePixAndCheck(rp, pixt1, IFF_PNG); /* 2 */
pixDisplayWithTitle(pixt1, 600, 100, "Hue", rp->display);
pixDestroy(&pixt1);
pixaDestroy(&pixa);
/* Vary saturation */
pixa = pixaCreate(20);
na = numaCreate(20);
for (i = 0; i < 20; i++) {
pixt0 = pixModifySaturation(NULL, pixs, -0.9 + 0.1 * i);
pixMeasureSaturation(pixt0, 1, &sat);
pixaAddPix(pixa, pixt0, L_INSERT);
numaAddNumber(na, sat);
}
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, w, 5, 0, 10, 2);
pixSaveTiled(pixt1, pixaf, 1.0, 1, 20, 0);
gplotSimple1(na, GPLOT_PNG, "/tmp/regout/enhance.7", "Average Saturation");
regTestWritePixAndCheck(rp, pixt1, IFF_PNG); /* 3 */
pixDisplayWithTitle(pixt1, 900, 100, "Saturation", rp->display);
numaDestroy(&na);
pixDestroy(&pixt1);
pixaDestroy(&pixa);
/* Vary contrast */
pixa = pixaCreate(20);
for (i = 0; i < 20; i++) {
pixt0 = pixContrastTRC(NULL, pixs, 0.1 * i);
pixaAddPix(pixa, pixt0, L_INSERT);
}
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, w, 5, 0, 10, 2);
pixSaveTiled(pixt1, pixaf, 1.0, 1, 20, 0);
regTestWritePixAndCheck(rp, pixt1, IFF_PNG); /* 4 */
pixDisplayWithTitle(pixt1, 0, 400, "Contrast", rp->display);
pixDestroy(&pixt1);
pixaDestroy(&pixa);
/* Vary sharpening */
pixa = pixaCreate(20);
for (i = 0; i < 20; i++) {
pixt0 = pixUnsharpMasking(pixs, 3, 0.01 + 0.15 * i);
pixaAddPix(pixa, pixt0, L_INSERT);
}
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, w, 5, 0, 10, 2);
pixSaveTiled(pixt1, pixaf, 1.0, 1, 20, 0);
regTestWritePixAndCheck(rp, pixt1, IFF_PNG); /* 5 */
pixDisplayWithTitle(pixt1, 300, 400, "Sharp", rp->display);
pixDestroy(&pixt1);
pixaDestroy(&pixa);
/* Hue constant mapping to lighter background */
pixa = pixaCreate(11);
bmf8 = bmfCreate("fonts", 8);
pixt0 = pixRead("candelabrum-11.jpg");
composeRGBPixel(230, 185, 144, &srcval); /* select typical bg pixel */
for (i = 0; i <= 10; i++) {
fract = 0.10 * i;
pixelFractionalShift(230, 185, 144, fract, &dstval);
pixt1 = pixLinearMapToTargetColor(NULL, pixt0, srcval, dstval);
snprintf(textstr, 50, "Fract = %5.1f", fract);
pixt2 = pixAddSingleTextblock(pixt1, bmf8, textstr, 0xff000000,
L_ADD_BELOW, NULL);
pixSaveTiledOutline(pixt2, pixa, 1.0, (i % 4 == 0) ? 1 : 0, 30, 2, 32);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
}
pixDestroy(&pixt0);
pixd = pixaDisplay(pixa, 0, 0);
regTestWritePixAndCheck(rp, pixd, IFF_JFIF_JPEG); /* 6 */
pixDisplayWithTitle(pixd, 600, 400, "Constant hue", rp->display);
bmfDestroy(&bmf8);
pixaDestroy(&pixa);
pixDestroy(&pixd);
/* Delayed testing of saturation plot */
regTestCheckFile(rp, "/tmp/regout/enhance.7.png"); /* 7 */
/* Display results */
pixd = pixaDisplay(pixaf, 0, 0);
regTestWritePixAndCheck(rp, pixd, IFF_JFIF_JPEG); /* 8 */
pixDisplayWithTitle(pixd, 100, 100, "All", rp->display);
pixDestroy(&pixd);
pixaDestroy(&pixaf);
pixDestroy(&pix);
pixDestroy(&pixs);
/* -----------------------------------------------*
* Test global color transforms *
* -----------------------------------------------*/
/* Make identical cmap and rgb images */
pix = pixRead("wet-day.jpg");
pixs1 = pixOctreeColorQuant(pix, 200, 0);
pixs2 = pixRemoveColormap(pixs1, REMOVE_CMAP_TO_FULL_COLOR);
regTestComparePix(rp, pixs1, pixs2); /* 9 */
/* Make a diagonal color transform matrix */
kel = kernelCreate(3, 3);
kernelSetElement(kel, 0, 0, 0.7);
kernelSetElement(kel, 1, 1, 0.4);
kernelSetElement(kel, 2, 2, 1.3);
/* Apply to both cmap and rgb images. */
pixt1 = pixMultMatrixColor(pixs1, kel);
pixt2 = pixMultMatrixColor(pixs2, kel);
regTestComparePix(rp, pixt1, pixt2); /* 10 */
kernelDestroy(&kel);
/* Apply the same transform in the simpler interface */
pixt3 = pixMultConstantColor(pixs1, 0.7, 0.4, 1.3);
pixt4 = pixMultConstantColor(pixs2, 0.7, 0.4, 1.3);
regTestComparePix(rp, pixt3, pixt4); /* 11 */
regTestComparePix(rp, pixt1, pixt3); /* 12 */
regTestWritePixAndCheck(rp, pixt1, IFF_JFIF_JPEG); /* 13 */
pixDestroy(&pix);
pixDestroy(&pixs1);
pixDestroy(&pixs2);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
pixDestroy(&pixt4);
return regTestCleanup(rp);
}
bool TessPDFRenderer::AddImageHandler(TessBaseAPI* api) {
char buf[kBasicBufSize];
Pix *pix = api->GetInputImage();
char *filename = (char *)api->GetInputName();
int ppi = api->GetSourceYResolution();
if (!pix || ppi <= 0)
return false;
double width = pixGetWidth(pix) * 72.0 / ppi;
double height = pixGetHeight(pix) * 72.0 / ppi;
// PAGE
snprintf(buf, sizeof(buf),
"%ld 0 obj\n"
"<<\n"
" /Type /Page\n"
" /Parent %ld 0 R\n"
" /MediaBox [0 0 %.2f %.2f]\n"
" /Contents %ld 0 R\n"
" /Resources\n"
" <<\n"
" /XObject << /Im1 %ld 0 R >>\n"
" /ProcSet [ /PDF /Text /ImageB /ImageI /ImageC ]\n"
" /Font << /f-0-0 %ld 0 R >>\n"
" >>\n"
">>\n"
"endobj\n",
obj_,
2L, // Pages object
width,
height,
obj_ + 1, // Contents object
obj_ + 2, // Image object
3L); // Type0 Font
pages_.push_back(obj_);
AppendPDFObject(buf);
// CONTENTS
char* pdftext = GetPDFTextObjects(api, width, height, imagenum());
long pdftext_len = strlen(pdftext);
unsigned char *pdftext_casted = reinterpret_cast<unsigned char *>(pdftext);
size_t len;
unsigned char *comp_pdftext =
zlibCompress(pdftext_casted,
pdftext_len,
&len);
long comp_pdftext_len = len;
snprintf(buf, sizeof(buf),
"%ld 0 obj\n"
"<<\n"
" /Length %ld /Filter /FlateDecode\n"
">>\n"
"stream\n", obj_, comp_pdftext_len);
AppendString(buf);
long objsize = strlen(buf);
AppendData(reinterpret_cast<char *>(comp_pdftext), comp_pdftext_len);
objsize += comp_pdftext_len;
lept_free(comp_pdftext);
delete[] pdftext;
snprintf(buf, sizeof(buf),
"endstream\n"
"endobj\n");
AppendString(buf);
objsize += strlen(buf);
AppendPDFObjectDIY(objsize);
char *pdf_object;
if (!fileToPDFObj(filename, obj_, &pdf_object, &objsize)) {
if (!pixToPDFObj(pix, obj_, &pdf_object, &objsize)) {
return false;
}
}
AppendData(pdf_object, objsize);
AppendPDFObjectDIY(objsize);
delete[] pdf_object;
return true;
}
main(int argc,
char **argv)
{
l_int32 i, w, h, dir;
PIX *pixs, *pixd, *pixt;
l_float32 pops;
char *filein, *fileout;
static char mainName[] = "rotateorthtest1";
if (argc != 3 && argc != 4)
exit(ERROR_INT(" Syntax: rotateorthtest1 filein fileout [direction]",
mainName, 1));
filein = argv[1];
fileout = argv[2];
if (argc == 4)
dir = atoi(argv[3]);
else
dir = 1;
if ((pixs = pixRead(filein)) == NULL)
exit(ERROR_INT("pix not made", mainName, 1));
/* Do a single operation */
#if 1
pixd = pixRotate90(pixs, dir);
#elif 0
pixd = pixRotate180(NULL, pixs);
#elif 0
pixd = pixRotateLR(NULL, pixs);
#elif 0
pixd = pixRotateTB(NULL, pixs);
#endif
/* Time rotate 90, allocating & destroying each time */
#if 0
startTimer();
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
for (i = 0; i < NTIMES; i++) {
pixd = pixRotate90(pixs, dir);
pixDestroy(&pixd);
}
pops = (l_float32)(w * h * NTIMES) / stopTimer();
fprintf(stderr, "MPops for 90 rotation: %7.3f\n", pops / 1000000.);
pixd = pixRotate90(pixs, dir);
#endif
/* Time rotate 180, with no alloc/destroy */
#if 0
startTimer();
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
pixd = pixCreateTemplate(pixs);
for (i = 0; i < NTIMES; i++)
pixRotate180(pixd, pixs);
pops = (l_float32)(w * h * NTIMES) / stopTimer();
fprintf(stderr, "MPops for 180 rotation: %7.3f\n", pops / 1000000.);
#endif
/* Test rotate 180 not in-place */
#if 0
pixt = pixRotate180(NULL, pixs);
pixd = pixRotate180(NULL, pixt);
pixEqual(pixs, pixd, &eq);
if (eq) fprintf(stderr, "2 rots gives I\n");
else fprintf(stderr, "2 rots fail to give I\n");
pixDestroy(&pixt);
#endif
/* Test rotate 180 in-place */
#if 0
pixd = pixCopy(NULL, pixs);
pixRotate180(pixd, pixd);
pixRotate180(pixd, pixd);
pixEqual(pixs, pixd, &eq);
if (eq) fprintf(stderr, "2 rots gives I\n");
else fprintf(stderr, "2 rots fail to give I\n");
#endif
/* Mix rotate 180 with LR/TB */
#if 0
pixd = pixRotate180(NULL, pixs);
pixRotateLR(pixd, pixd);
pixRotateTB(pixd, pixd);
pixEqual(pixs, pixd, &eq);
if (eq) fprintf(stderr, "180 rot OK\n");
else fprintf(stderr, "180 rot error\n");
#endif
if (pixGetDepth(pixd) < 8)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
pixDestroy(&pixs);
pixDestroy(&pixd);
return 0;
}
/*!
* \brief pixItalicWords()
*
* \param[in] pixs 1 bpp
* \param[in] boxaw [optional] word bounding boxes; can be NULL
* \param[in] pixw [optional] word box mask; can be NULL
* \param[out] pboxa boxa of italic words
* \param[in] debugflag 1 for debug output; 0 otherwise
* \return 0 if OK, 1 on error
*
* <pre>
* Notes:
* (1) You can input the bounding boxes for the words in one of
* two forms: as bounding boxes (%boxaw) or as a word mask with
* the word bounding boxes filled (%pixw). For example,
* to compute %pixw, you can use pixWordMaskByDilation().
* (2) Alternatively, you can set both of these inputs to NULL,
* in which case the word mask is generated here. This is
* done by dilating and closing the input image to connect
* letters within a word, while leaving the words separated.
* The parameters are chosen under the assumption that the
* input is 10 to 12 pt text, scanned at about 300 ppi.
* (3) sel_ital1 and sel_ital2 detect the right edges that are
* nearly vertical, at approximately the angle of italic
* strokes. We use the right edge to avoid getting seeds
* from lower-case 'y'. The typical italic slant has a smaller
* angle with the vertical than the 'W', so in most cases we
* will not trigger on the slanted lines in the 'W'.
* (4) Note that sel_ital2 is shorter than sel_ital1. It is
* more appropriate for a typical font scanned at 200 ppi.
* </pre>
*/
l_int32
pixItalicWords(PIX *pixs,
BOXA *boxaw,
PIX *pixw,
BOXA **pboxa,
l_int32 debugflag)
{
char opstring[32];
l_int32 size;
BOXA *boxa;
PIX *pixsd, *pixm, *pixd;
SEL *sel_ital1, *sel_ital2, *sel_ital3;
PROCNAME("pixItalicWords");
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (!pboxa)
return ERROR_INT("&boxa not defined", procName, 1);
if (boxaw && pixw)
return ERROR_INT("both boxaw and pixw are defined", procName, 1);
sel_ital1 = selCreateFromString(str_ital1, 13, 6, NULL);
sel_ital2 = selCreateFromString(str_ital2, 10, 6, NULL);
sel_ital3 = selCreateFromString(str_ital3, 4, 2, NULL);
/* Make the italic seed: extract with HMT; remove noise.
* The noise removal close/open is important to exclude
* situations where a small slanted line accidentally
* matches sel_ital1. */
pixsd = pixHMT(NULL, pixs, sel_ital1);
pixClose(pixsd, pixsd, sel_ital3);
pixOpen(pixsd, pixsd, sel_ital3);
/* Make the word mask. Use input boxes or mask if given. */
size = 0; /* init */
if (boxaw) {
pixm = pixCreateTemplate(pixs);
pixMaskBoxa(pixm, pixm, boxaw, L_SET_PIXELS);
} else if (pixw) {
pixm = pixClone(pixw);
} else {
pixWordMaskByDilation(pixs, NULL, &size, NULL);
L_INFO("dilation size = %d\n", procName, size);
snprintf(opstring, sizeof(opstring), "d1.5 + c%d.1", size);
pixm = pixMorphSequence(pixs, opstring, 0);
}
/* Binary reconstruction to fill in those word mask
* components for which there is at least one seed pixel. */
pixd = pixSeedfillBinary(NULL, pixsd, pixm, 8);
boxa = pixConnComp(pixd, NULL, 8);
*pboxa = boxa;
if (debugflag) {
/* Save results at at 2x reduction */
lept_mkdir("lept/ital");
l_int32 res, upper;
BOXA *boxat;
GPLOT *gplot;
NUMA *na;
PIXA *pad;
PIX *pix1, *pix2, *pix3;
pad = pixaCreate(0);
boxat = pixConnComp(pixm, NULL, 8);
boxaWrite("/tmp/lept/ital/ital.ba", boxat);
pixSaveTiledOutline(pixs, pad, 0.5, 1, 20, 2, 32); /* orig */
pixSaveTiledOutline(pixsd, pad, 0.5, 1, 20, 2, 0); /* seed */
pix1 = pixConvertTo32(pixm);
pixRenderBoxaArb(pix1, boxat, 3, 255, 0, 0);
pixSaveTiledOutline(pix1, pad, 0.5, 1, 20, 2, 0); /* mask + outline */
pixDestroy(&pix1);
pixSaveTiledOutline(pixd, pad, 0.5, 1, 20, 2, 0); /* ital mask */
pix1 = pixConvertTo32(pixs);
pixRenderBoxaArb(pix1, boxa, 3, 255, 0, 0);
pixSaveTiledOutline(pix1, pad, 0.5, 1, 20, 2, 0); /* orig + outline */
pixDestroy(&pix1);
pix1 = pixCreateTemplate(pixs);
pix2 = pixSetBlackOrWhiteBoxa(pix1, boxa, L_SET_BLACK);
pixCopy(pix1, pixs);
pix3 = pixDilateBrick(NULL, pixs, 3, 3);
pixCombineMasked(pix1, pix3, pix2);
pixSaveTiledOutline(pix1, pad, 0.5, 1, 20, 2, 0); /* ital bolded */
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pix2 = pixaDisplay(pad, 0, 0);
pixWrite("/tmp/lept/ital/ital.png", pix2, IFF_PNG);
pixDestroy(&pix2);
/* Assuming the image represents 6 inches of actual page width,
* the pixs resolution is approximately
* (width of pixs in pixels) / 6
* and the images have been saved at half this resolution. */
res = pixGetWidth(pixs) / 12;
L_INFO("resolution = %d\n", procName, res);
l_pdfSetDateAndVersion(0);
pixaConvertToPdf(pad, res, 1.0, L_FLATE_ENCODE, 75, "Italic Finder",
"/tmp/lept/ital/ital.pdf");
l_pdfSetDateAndVersion(1);
pixaDestroy(&pad);
boxaDestroy(&boxat);
/* Plot histogram of horizontal white run sizes. A small
* initial vertical dilation removes most runs that are neither
* inter-character nor inter-word. The larger first peak is
* from inter-character runs, and the smaller second peak is
* from inter-word runs. */
pix1 = pixDilateBrick(NULL, pixs, 1, 15);
upper = L_MAX(30, 3 * size);
na = pixRunHistogramMorph(pix1, L_RUN_OFF, L_HORIZ, upper);
pixDestroy(&pix1);
gplot = gplotCreate("/tmp/lept/ital/runhisto", GPLOT_PNG,
"Histogram of horizontal runs of white pixels, vs length",
"run length", "number of runs");
gplotAddPlot(gplot, NULL, na, GPLOT_LINES, "plot1");
gplotMakeOutput(gplot);
gplotDestroy(&gplot);
numaDestroy(&na);
}
selDestroy(&sel_ital1);
selDestroy(&sel_ital2);
selDestroy(&sel_ital3);
pixDestroy(&pixsd);
pixDestroy(&pixm);
pixDestroy(&pixd);
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);
}
int main(int argc,
char **argv) {
char *filein, *fileout;
l_int32 i;
l_uint32 val;
l_float32 size;
PIX *pixs, *pixd, *pixm, *pixmi, *pixt1, *pixt2, *pixt3;
static char mainName[] = "seedfilltest";
if (argc != 3)
return ERROR_INT(" Syntax: seedfilltest filein fileout", mainName, 1);
filein = argv[1];
fileout = argv[2];
pixd = NULL;
if ((pixm = pixRead(filein)) == NULL)
return ERROR_INT("pixm not made", mainName, 1);
pixmi = pixInvert(NULL, pixm);
size = pixGetWidth(pixm) * pixGetHeight(pixm);
pixs = pixCreateTemplate(pixm);
for (i = 0; i < 100; i++) {
pixGetPixel(pixm, XS + 5 * i, YS + 5 * i, &val);
if (val == 0) break;
}
if (i == 100)
return ERROR_INT("no seed pixel found", mainName, 1);
pixSetPixel(pixs, XS + 5 * i, YS + 5 * i, 1);
#if 0
/* hole filling; use "hole-filler.png" */
pixt1 = pixHDome(pixmi, 100, 4);
pixt2 = pixThresholdToBinary(pixt1, 10);
/* pixInvert(pixt1, pixt1); */
pixDisplay(pixt1, 100, 500);
pixDisplay(pixt2, 600, 500);
pixt3 = pixHolesByFilling(pixt2, 4);
pixDilateBrick(pixt3, pixt3, 7, 7);
pixd = pixConvertTo8(pixt3, FALSE);
pixDisplay(pixd, 0, 100);
pixSeedfillGray(pixd, pixmi, CONNECTIVITY);
pixInvert(pixd, pixd);
pixDisplay(pixmi, 500, 100);
pixDisplay(pixd, 1000, 100);
pixWrite("/tmp/junkpixm.png", pixmi, IFF_PNG);
pixWrite("/tmp/junkpixd.png", pixd, IFF_PNG);
#endif
#if 0
/* hole filling; use "hole-filler.png" */
pixt1 = pixThresholdToBinary(pixm, 110);
pixInvert(pixt1, pixt1);
pixDisplay(pixt1, 100, 500);
pixt2 = pixHolesByFilling(pixt1, 4);
pixd = pixConvertTo8(pixt2, FALSE);
pixDisplay(pixd, 0, 100);
pixSeedfillGray(pixd, pixmi, CONNECTIVITY);
pixInvert(pixd, pixd);
pixDisplay(pixmi, 500, 100);
pixDisplay(pixd, 1000, 100);
pixWrite("/tmp/junkpixm.png", pixmi, IFF_PNG);
pixWrite("/tmp/junkpixd.png", pixd, IFF_PNG);
#endif
#if 0
/* hole filling; use "hole-filler.png" */
pixd = pixInvert(NULL, pixm);
pixAddConstantGray(pixd, -50);
pixDisplay(pixd, 0, 100);
/* pixt1 = pixThresholdToBinary(pixd, 20);
pixDisplayWithTitle(pixt1, 600, 600, "pixt1", DFLAG); */
pixSeedfillGray(pixd, pixmi, CONNECTIVITY);
/* pixInvert(pixd, pixd); */
pixDisplay(pixmi, 500, 100);
pixDisplay(pixd, 1000, 100);
pixWrite("/tmp/junkpixm.png", pixmi, IFF_PNG);
pixWrite("/tmp/junkpixd.png", pixd, IFF_PNG);
#endif
#if 0
/* test in-place seedfill for speed */
pixd = pixClone(pixs);
startTimer();
pixSeedfillBinary(pixs, pixs, pixmi, CONNECTIVITY);
fprintf(stderr, "Filling rate: %7.4f Mpix/sec\n",
(size/1000000.) / stopTimer());
pixWrite(fileout, pixd, IFF_PNG);
pixOr(pixd, pixd, pixm);
pixWrite("/tmp/junkout1.png", pixd, IFF_PNG);
#endif
#if 0
/* test seedfill to dest for speed */
pixd = pixCreateTemplate(pixm);
startTimer();
for (i = 0; i < NTIMES; i++) {
pixSeedfillBinary(pixd, pixs, pixmi, CONNECTIVITY);
}
fprintf(stderr, "Filling rate: %7.4f Mpix/sec\n",
(size/1000000.) * NTIMES / stopTimer());
pixWrite(fileout, pixd, IFF_PNG);
pixOr(pixd, pixd, pixm);
pixWrite("/tmp/junkout1.png", pixd, IFF_PNG);
#endif
/* use same connectivity to compare with the result of the
* slow parallel operation */
#if 1
pixDestroy(&pixd);
pixd = pixSeedfillMorph(pixs, pixmi, 100, CONNECTIVITY);
pixOr(pixd, pixd, pixm);
pixWrite("/tmp/junkout2.png", pixd, IFF_PNG);
#endif
pixDestroy(&pixs);
pixDestroy(&pixm);
pixDestroy(&pixmi);
pixDestroy(&pixd);
return 0;
}
/*!
* pixHtmlViewer()
*
* Input: dirin: directory of input image files
* dirout: directory for output files
* rootname: root name for output files
* thumbwidth: width of thumb images
* (in pixels; use 0 for default)
* viewwidth: maximum width of view images (no up-scaling)
* (in pixels; use 0 for default)
* copyorig: 1 to copy originals to dirout; 0 otherwise
* Return: 0 if OK; 1 on error
*
* Notes:
* (1) The thumb and view reduced images are generated,
* along with two html files:
* <rootname>.html and <rootname>-links.html
* (2) The thumb and view files are named
* <rootname>_thumb_xxx.jpg
* <rootname>_view_xxx.jpg
* With this naming scheme, any number of input directories
* of images can be processed into views and thumbs
* and placed in the same output directory.
*/
l_int32
pixHtmlViewer(const char *dirin,
const char *dirout,
const char *rootname,
l_int32 thumbwidth,
l_int32 viewwidth,
l_int32 copyorig)
{
char *fname, *fullname, *outname;
char *mainname, *linkname, *linknameshort;
char *viewfile, *thumbfile;
char *shtml, *slink;
char charbuf[L_BUF_SIZE];
char htmlstring[] = "<html>";
char framestring[] = "</frameset></html>";
l_int32 i, nfiles, index, w, nimages, ret;
l_float32 factor;
PIX *pix, *pixthumb, *pixview;
SARRAY *safiles, *sathumbs, *saviews, *sahtml, *salink;
PROCNAME("pixHtmlViewer");
if (!dirin)
return ERROR_INT("dirin not defined", procName, 1);
if (!dirout)
return ERROR_INT("dirout not defined", procName, 1);
if (!rootname)
return ERROR_INT("rootname not defined", procName, 1);
if (thumbwidth == 0)
thumbwidth = DEFAULT_THUMB_WIDTH;
if (thumbwidth < MIN_THUMB_WIDTH) {
L_WARNING("thumbwidth too small; using min value\n", procName);
thumbwidth = MIN_THUMB_WIDTH;
}
if (viewwidth == 0)
viewwidth = DEFAULT_VIEW_WIDTH;
if (viewwidth < MIN_VIEW_WIDTH) {
L_WARNING("viewwidth too small; using min value\n", procName);
viewwidth = MIN_VIEW_WIDTH;
}
/* Make the output directory if it doesn't already exist */
#ifndef _WIN32
sprintf(charbuf, "mkdir -p %s", dirout);
ret = system(charbuf);
#else
ret = CreateDirectory(dirout, NULL) ? 0 : 1;
#endif /* !_WIN32 */
if (ret) {
L_ERROR("output directory %s not made\n", procName, dirout);
return 1;
}
/* Capture the filenames in the input directory */
if ((safiles = getFilenamesInDirectory(dirin)) == NULL)
return ERROR_INT("safiles not made", procName, 1);
/* Generate output text file names */
sprintf(charbuf, "%s/%s.html", dirout, rootname);
mainname = stringNew(charbuf);
sprintf(charbuf, "%s/%s-links.html", dirout, rootname);
linkname = stringNew(charbuf);
linknameshort = stringJoin(rootname, "-links.html");
if ((sathumbs = sarrayCreate(0)) == NULL)
return ERROR_INT("sathumbs not made", procName, 1);
if ((saviews = sarrayCreate(0)) == NULL)
return ERROR_INT("saviews not made", procName, 1);
/* Generate the thumbs and views */
nfiles = sarrayGetCount(safiles);
index = 0;
for (i = 0; i < nfiles; i++) {
fname = sarrayGetString(safiles, i, L_NOCOPY);
fullname = genPathname(dirin, fname);
fprintf(stderr, "name: %s\n", fullname);
if ((pix = pixRead(fullname)) == NULL) {
fprintf(stderr, "file %s not a readable image\n", fullname);
FREE(fullname);
continue;
}
FREE(fullname);
if (copyorig) {
outname = genPathname(dirout, fname);
pixWrite(outname, pix, IFF_JFIF_JPEG);
FREE(outname);
}
/* Make and store the thumb */
w = pixGetWidth(pix);
factor = (l_float32)thumbwidth / (l_float32)w;
if ((pixthumb = pixScale(pix, factor, factor)) == NULL)
return ERROR_INT("pixthumb not made", procName, 1);
sprintf(charbuf, "%s_thumb_%03d.jpg", rootname, index);
sarrayAddString(sathumbs, charbuf, L_COPY);
outname = genPathname(dirout, charbuf);
pixWrite(outname, pixthumb, IFF_JFIF_JPEG);
FREE(outname);
pixDestroy(&pixthumb);
/* Make and store the view */
factor = (l_float32)viewwidth / (l_float32)w;
if (factor >= 1.0) {
pixview = pixClone(pix); /* no upscaling */
} else {
if ((pixview = pixScale(pix, factor, factor)) == NULL)
return ERROR_INT("pixview not made", procName, 1);
}
sprintf(charbuf, "%s_view_%03d.jpg", rootname, index);
sarrayAddString(saviews, charbuf, L_COPY);
outname = genPathname(dirout, charbuf);
pixWrite(outname, pixview, IFF_JFIF_JPEG);
FREE(outname);
pixDestroy(&pixview);
pixDestroy(&pix);
index++;
}
/* Generate the main html file */
if ((sahtml = sarrayCreate(0)) == NULL)
return ERROR_INT("sahtml not made", procName, 1);
sarrayAddString(sahtml, htmlstring, L_COPY);
sprintf(charbuf, "<frameset cols=\"%d, *\">", thumbwidth + 30);
sarrayAddString(sahtml, charbuf, L_COPY);
sprintf(charbuf, "<frame name=\"thumbs\" src=\"%s\">", linknameshort);
sarrayAddString(sahtml, charbuf, L_COPY);
sprintf(charbuf, "<frame name=\"views\" src=\"%s\">",
sarrayGetString(saviews, 0, L_NOCOPY));
sarrayAddString(sahtml, charbuf, L_COPY);
sarrayAddString(sahtml, framestring, L_COPY);
shtml = sarrayToString(sahtml, 1);
l_binaryWrite(mainname, "w", shtml, strlen(shtml));
FREE(shtml);
FREE(mainname);
/* Generate the link html file */
nimages = sarrayGetCount(saviews);
fprintf(stderr, "num. images = %d\n", nimages);
if ((salink = sarrayCreate(0)) == NULL)
return ERROR_INT("salink not made", procName, 1);
for (i = 0; i < nimages; i++) {
viewfile = sarrayGetString(saviews, i, L_NOCOPY);
thumbfile = sarrayGetString(sathumbs, i, L_NOCOPY);
sprintf(charbuf, "<a href=\"%s\" TARGET=views><img src=\"%s\"></a>",
viewfile, thumbfile);
sarrayAddString(salink, charbuf, L_COPY);
}
slink = sarrayToString(salink, 1);
l_binaryWrite(linkname, "w", slink, strlen(slink));
FREE(slink);
FREE(linkname);
FREE(linknameshort);
sarrayDestroy(&safiles);
sarrayDestroy(&sathumbs);
sarrayDestroy(&saviews);
sarrayDestroy(&sahtml);
sarrayDestroy(&salink);
return 0;
}
/*!
* pixAssignToNearestColor()
*
* Input: pixd (8 bpp, colormapped)
* pixs (32 bpp; 24-bit color)
* pixm (<optional> 1 bpp)
* level (of octcube used for finding nearest color in cmap)
* countarray (<optional> ptr to array, in which we can store
* the number of pixels found in each color in
* the colormap in pixd)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) This is used in phase 2 of color segmentation, where pixs
* is the original input image to pixColorSegment(), and
* pixd is the colormapped image returned from
* pixColorSegmentCluster(). It is also used, with a mask,
* in phase 4.
* (2) This is an in-place operation.
* (3) The colormap in pixd is unchanged.
* (4) pixs and pixd must be the same size (w, h).
* (5) The selection mask pixm can be null. If it exists, it must
* be the same size as pixs and pixd, and only pixels
* corresponding to fg in pixm are assigned. Set to
* NULL if all pixels in pixd are to be assigned.
* (6) The countarray can be null. If it exists, it is pre-allocated
* and of a size at least equal to the size of the colormap in pixd.
* (7) This does a best-fit (non-greedy) assignment of pixels to
* existing clusters. Specifically, it assigns each pixel
* in pixd to the color index in the pixd colormap that has a
* color closest to the corresponding rgb pixel in pixs.
* (8) 'level' is the octcube level used to quickly find the nearest
* color in the colormap for each pixel. For color segmentation,
* this parameter is set to LEVEL_IN_OCTCUBE.
* (9) We build a mapping table from octcube to colormap index so
* that this function can run in a time (otherwise) independent
* of the number of colors in the colormap. This avoids a
* brute-force search for the closest colormap color to each
* pixel in the image.
*/
l_int32
pixAssignToNearestColor(PIX *pixd,
PIX *pixs,
PIX *pixm,
l_int32 level,
l_int32 *countarray) {
l_int32 w, h, wpls, wpld, wplm, i, j;
l_int32 rval, gval, bval, index;
l_int32 *cmaptab;
l_uint32 octindex;
l_uint32 *rtab, *gtab, *btab;
l_uint32 *ppixel;
l_uint32 *datas, *datad, *datam, *lines, *lined, *linem;
PIXCMAP *cmap;
PROCNAME("pixAssignToNearestColor");
if (!pixd)
return ERROR_INT("pixd not defined", procName, 1);
if ((cmap = pixGetColormap(pixd)) == NULL)
return ERROR_INT("cmap not found", procName, 1);
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (pixGetDepth(pixs) != 32)
return ERROR_INT("pixs not 32 bpp", procName, 1);
/* Set up the tables to map rgb to the nearest colormap index */
if (makeRGBToIndexTables(&rtab, >ab, &btab, level))
return ERROR_INT("index tables not made", procName, 1);
if ((cmaptab = pixcmapToOctcubeLUT(cmap, level, L_MANHATTAN_DISTANCE))
== NULL)
return ERROR_INT("cmaptab not made", procName, 1);
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
datas = pixGetData(pixs);
datad = pixGetData(pixd);
wpls = pixGetWpl(pixs);
wpld = pixGetWpl(pixd);
if (pixm) {
datam = pixGetData(pixm);
wplm = pixGetWpl(pixm);
}
for (i = 0; i < h; i++) {
lines = datas + i * wpls;
lined = datad + i * wpld;
if (pixm)
linem = datam + i * wplm;
for (j = 0; j < w; j++) {
if (pixm) {
if (!GET_DATA_BIT(linem, j))
continue;
}
ppixel = lines + j;
rval = GET_DATA_BYTE(ppixel, COLOR_RED);
gval = GET_DATA_BYTE(ppixel, COLOR_GREEN);
bval = GET_DATA_BYTE(ppixel, COLOR_BLUE);
/* Map from rgb to octcube index */
getOctcubeIndexFromRGB(rval, gval, bval, rtab, gtab, btab,
&octindex);
/* Map from octcube index to nearest colormap index */
index = cmaptab[octindex];
if (countarray)
countarray[index]++;
SET_DATA_BYTE(lined, j, index);
}
}
FREE(cmaptab);
FREE(rtab);
FREE(gtab);
FREE(btab);
return 0;
}
int main(int argc,
char **argv)
{
l_uint8 *array1, *array2;
l_int32 i, n1, n2, n3;
size_t size1, size2;
FILE *fp;
BOXA *boxa1, *boxa2;
PIX *pixs, *pix1, *pix2, *pix3;
PIXA *pixa1;
PIXCMAP *cmap;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
pixs = pixRead("feyn.tif");
/* --------------------------------------------------------------- *
* Test pixConnComp() and pixCountConnComp(), *
* with output to both boxa and pixa *
* --------------------------------------------------------------- */
/* First, test with 4-cc */
boxa1= pixConnComp(pixs, &pixa1, 4);
n1 = boxaGetCount(boxa1);
boxa2= pixConnComp(pixs, NULL, 4);
n2 = boxaGetCount(boxa2);
pixCountConnComp(pixs, 4, &n3);
fprintf(stderr, "Number of 4 c.c.: n1 = %d; n2 = %d, n3 = %d\n",
n1, n2, n3);
regTestCompareValues(rp, n1, n2, 0); /* 0 */
regTestCompareValues(rp, n1, n3, 0); /* 1 */
regTestCompareValues(rp, n1, 4452, 0); /* 2 */
pix1 = pixaDisplay(pixa1, pixGetWidth(pixs), pixGetHeight(pixs));
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 3 */
regTestComparePix(rp, pixs, pix1); /* 4 */
pixaDestroy(&pixa1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
pixDestroy(&pix1);
/* Test with 8-cc */
boxa1= pixConnComp(pixs, &pixa1, 8);
n1 = boxaGetCount(boxa1);
boxa2= pixConnComp(pixs, NULL, 8);
n2 = boxaGetCount(boxa2);
pixCountConnComp(pixs, 8, &n3);
fprintf(stderr, "Number of 8 c.c.: n1 = %d; n2 = %d, n3 = %d\n",
n1, n2, n3);
regTestCompareValues(rp, n1, n2, 0); /* 5 */
regTestCompareValues(rp, n1, n3, 0); /* 6 */
regTestCompareValues(rp, n1, 4305, 0); /* 7 */
pix1 = pixaDisplay(pixa1, pixGetWidth(pixs), pixGetHeight(pixs));
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 8 */
regTestComparePix(rp, pixs, pix1); /* 9 */
pixaDestroy(&pixa1);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
pixDestroy(&pix1);
/* --------------------------------------------------------------- *
* Test boxa I/O *
* --------------------------------------------------------------- */
lept_mkdir("lept/conn");
boxa1 = pixConnComp(pixs, NULL, 4);
fp = lept_fopen("/tmp/lept/conn/boxa1.ba", "wb+");
boxaWriteStream(fp, boxa1);
lept_fclose(fp);
fp = lept_fopen("/tmp/lept/conn/boxa1.ba", "rb");
boxa2 = boxaReadStream(fp);
lept_fclose(fp);
fp = lept_fopen("/tmp/lept/conn/boxa2.ba", "wb+");
boxaWriteStream(fp, boxa2);
lept_fclose(fp);
array1 = l_binaryRead("/tmp/lept/conn/boxa1.ba", &size1);
array2 = l_binaryRead("/tmp/lept/conn/boxa2.ba", &size2);
regTestCompareStrings(rp, array1, size1, array2, size2); /* 10 */
lept_free(array1);
lept_free(array2);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
/* --------------------------------------------------------------- *
* Just for fun, display each component as a random color in *
* cmapped 8 bpp. Background is color 0; it is set to white. *
* --------------------------------------------------------------- */
boxa1 = pixConnComp(pixs, &pixa1, 4);
pix1 = pixaDisplayRandomCmap(pixa1, pixGetWidth(pixs), pixGetHeight(pixs));
cmap = pixGetColormap(pix1);
pixcmapResetColor(cmap, 0, 255, 255, 255); /* reset background to white */
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 11 */
if (rp->display) pixDisplay(pix1, 100, 0);
boxaDestroy(&boxa1);
pixDestroy(&pix1);
pixaDestroy(&pixa1);
pixDestroy(&pixs);
/* --------------------------------------------------------------- *
* Test iterative covering of connected components by rectangles *
* --------------------------------------------------------------- */
pixa1 = pixaCreate(0);
pix1 = pixRead("rabi.png");
pix2 = pixReduceRankBinaryCascade(pix1, 1, 1, 1, 0);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 12 - */
pixaAddPix(pixa1, pix2, L_INSERT);
for (i = 1; i < 6; i++) {
pix3 = pixMakeCoveringOfRectangles(pix2, i);
regTestWritePixAndCheck(rp, pix3, IFF_PNG); /* 13 - 17 */
pixaAddPix(pixa1, pix3, L_INSERT);
}
pix3 = pixaDisplayTiledInRows(pixa1, 1, 2500, 1.0, 0, 30, 0);
regTestWritePixAndCheck(rp, pix3, IFF_PNG); /* 18 */
pixDisplayWithTitle(pix3, 100, 900, NULL, rp->display);
pixDestroy(&pix1);
pixDestroy(&pix3);
pixaDestroy(&pixa1);
return regTestCleanup(rp);
}
