void
MakeWordBoxes2(PIX *pixs,
l_int32 reduction,
L_REGPARAMS *rp)
{
l_int32 default_minwidth = 10;
l_int32 default_minheight = 10;
l_int32 default_maxwidth = 400;
l_int32 default_maxheight = 70;
l_int32 minwidth, minheight, maxwidth, maxheight;
BOXA *boxa1, *boxa2;
NUMA *na;
PIX *pixd1, *pixd2;
PIXA *pixa;
minwidth = default_minwidth / reduction;
minheight = default_minheight / reduction;
maxwidth = default_maxwidth / reduction;
maxheight = default_maxheight / reduction;
/* Get the word boxes */
pixGetWordsInTextlines(pixs, reduction, minwidth, minheight,
maxwidth, maxheight, &boxa1, &pixa, &na);
pixaDestroy(&pixa);
numaDestroy(&na);
if (reduction == 1)
boxa2 = boxaCopy(boxa1, L_CLONE);
else
boxa2 = boxaTransform(boxa1, 0, 0, 2.0, 2.0);
pixd1 = pixConvertTo8(pixs, 1);
pixRenderBoxaArb(pixd1, boxa2, 2, 255, 0, 0);
regTestWritePixAndCheck(rp, pixd1, IFF_PNG);
pixDisplayWithTitle(pixd1, 800, 100, NULL, rp->display);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
/* Do it again with this interface. The result should be the same. */
pixGetWordBoxesInTextlines(pixs, reduction, minwidth, minheight,
maxwidth, maxheight, &boxa1, NULL);
if (reduction == 1)
boxa2 = boxaCopy(boxa1, L_CLONE);
else
boxa2 = boxaTransform(boxa1, 0, 0, 2.0, 2.0);
pixd2 = pixConvertTo8(pixs, 1);
pixRenderBoxaArb(pixd2, boxa2, 2, 255, 0, 0);
if (regTestComparePix(rp, pixd1, pixd2)) {
L_ERROR("pix not the same", "MakeWordBoxes2");
pixDisplayWithTitle(pixd2, 800, 100, NULL, rp->display);
}
pixDestroy(&pixd1);
pixDestroy(&pixd2);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
return;
}
int main(int argc,
char **argv)
{
char filename[BUF_SIZE];
char *dirin, *rootname, *fname;
l_int32 i, j, w, h, firstpage, npages, nfiles, ncomp;
l_int32 index, ival, rval, gval, bval;
BOX *box;
BOXA *boxa;
BOXAA *baa;
JBDATA *data;
JBCLASSER *classer;
NUMA *nai;
NUMAA *naa;
SARRAY *safiles;
PIX *pixs, *pixt1, *pixt2, *pixd;
PIXCMAP *cmap;
static char mainName[] = "wordsinorder";
if (argc != 3 && argc != 5)
return ERROR_INT(
" Syntax: wordsinorder dirin rootname [firstpage, npages]",
mainName, 1);
dirin = argv[1];
rootname = argv[2];
if (argc == 3) {
firstpage = 0;
npages = 0;
}
else {
firstpage = atoi(argv[3]);
npages = atoi(argv[4]);
}
/* Compute the word bounding boxes at 2x reduction, along with
* the textlines that they are in. */
safiles = getSortedPathnamesInDirectory(dirin, NULL, firstpage, npages);
nfiles = sarrayGetCount(safiles);
baa = boxaaCreate(nfiles);
naa = numaaCreate(nfiles);
for (i = 0; i < nfiles; i++) {
fname = sarrayGetString(safiles, i, 0);
if ((pixs = pixRead(fname)) == NULL) {
L_WARNING("image file %d not read\n", mainName, i);
continue;
}
pixGetWordBoxesInTextlines(pixs, 2, MIN_WORD_WIDTH, MIN_WORD_HEIGHT,
MAX_WORD_WIDTH, MAX_WORD_HEIGHT,
&boxa, &nai);
boxaaAddBoxa(baa, boxa, L_INSERT);
numaaAddNuma(naa, nai, L_INSERT);
#if RENDER_PAGES
/* Show the results on a 2x reduced image, where each
* word is outlined and the color of the box depends on the
* computed textline. */
pixt1 = pixReduceRankBinary2(pixs, 2, NULL);
pixGetDimensions(pixt1, &w, &h, NULL);
pixd = pixCreate(w, h, 8);
cmap = pixcmapCreateRandom(8, 1, 1); /* first color is black */
pixSetColormap(pixd, cmap);
pixt2 = pixUnpackBinary(pixt1, 8, 1);
pixRasterop(pixd, 0, 0, w, h, PIX_SRC | PIX_DST, pixt2, 0, 0);
ncomp = boxaGetCount(boxa);
for (j = 0; j < ncomp; j++) {
box = boxaGetBox(boxa, j, L_CLONE);
numaGetIValue(nai, j, &ival);
index = 1 + (ival % 254); /* omit black and white */
pixcmapGetColor(cmap, index, &rval, &gval, &bval);
pixRenderBoxArb(pixd, box, 2, rval, gval, bval);
boxDestroy(&box);
}
snprintf(filename, BUF_SIZE, "%s.%05d", rootname, i);
fprintf(stderr, "filename: %s\n", filename);
pixWrite(filename, pixd, IFF_PNG);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixs);
pixDestroy(&pixd);
#endif /* RENDER_PAGES */
}
boxaaDestroy(&baa);
numaaDestroy(&naa);
sarrayDestroy(&safiles);
return 0;
}
l_int32 main(int argc,
char **argv)
{
char buf[64];
BOXA *boxa1, *boxa2, *boxa3, *boxa4;
L_DEWARP *dew;
L_DEWARPA *dewa;
PIX *pixs, *pixn, *pixg, *pixb, *pix2, *pix3, *pix4, *pix5, *pix6;
lept_mkdir("lept");
snprintf(buf, sizeof(buf), "cat.%03d.jpg", pageno);
pixs = pixRead(buf);
dewa = dewarpaCreate(40, 30, 1, 15, 10);
dewarpaUseBothArrays(dewa, 1);
/* Normalize for varying background and binarize */
pixn = pixBackgroundNormSimple(pixs, NULL, NULL);
pixg = pixConvertRGBToGray(pixn, 0.5, 0.3, 0.2);
pixb = pixThresholdToBinary(pixg, 130);
pixDisplay(pixb, 0, 100);
/* Build the model */
dew = dewarpCreate(pixb, pageno);
dewarpaInsertDewarp(dewa, dew);
if (build_output) {
snprintf(buf, sizeof(buf), "/tmp/lept/dewarp_build_%d.pdf", pageno);
dewarpBuildPageModel(dew, buf);
} else {
dewarpBuildPageModel(dew, NULL);
}
/* Apply the model */
dewarpPopulateFullRes(dew, pixg, 0, 0);
if (apply_output) {
snprintf(buf, sizeof(buf), "/tmp/lept/dewarp_apply_%d.pdf", pageno);
dewarpaApplyDisparity(dewa, pageno, pixb, 200, 0, 0, &pix2, buf);
} else {
dewarpaApplyDisparity(dewa, pageno, pixb, 200, 0, 0, &pix2, NULL);
}
pixDisplay(pix2, 200, 100);
/* Reverse direction: get the word boxes for the dewarped pix ... */
pixGetWordBoxesInTextlines(pix2, 5, 5, 500, 100, &boxa1, NULL);
pix3 = pixConvertTo32(pix2);
pixRenderBoxaArb(pix3, boxa1, 2, 255, 0, 0);
pixDisplay(pix3, 400, 100);
/* ... and map to the word boxes for the input image */
if (map_output) {
snprintf(buf, sizeof(buf), "/tmp/lept/dewarp_map1_%d.pdf", pageno);
dewarpaApplyDisparityBoxa(dewa, pageno, pix2, boxa1, 0, 0, 0, &boxa2,
buf);
} else {
dewarpaApplyDisparityBoxa(dewa, pageno, pix2, boxa1, 0, 0, 0, &boxa2,
NULL);
}
pix4 = pixConvertTo32(pixb);
pixRenderBoxaArb(pix4, boxa2, 2, 0, 255, 0);
pixDisplay(pix4, 600, 100);
/* Forward direction: get the word boxes for the input pix ... */
pixGetWordBoxesInTextlines(pixb, 5, 5, 500, 100, &boxa3, NULL);
pix5 = pixConvertTo32(pixb);
pixRenderBoxaArb(pix5, boxa3, 2, 255, 0, 0);
pixDisplay(pix5, 800, 100);
/* ... and map to the word boxes for the dewarped image */
if (map_output) {
snprintf(buf, sizeof(buf), "/tmp/lept/dewarp_map2_%d.pdf", pageno);
dewarpaApplyDisparityBoxa(dewa, pageno, pixb, boxa3, 1, 0, 0, &boxa4,
buf);
} else {
dewarpaApplyDisparityBoxa(dewa, pageno, pixb, boxa3, 1, 0, 0, &boxa4,
NULL);
}
pix6 = pixConvertTo32(pix2);
pixRenderBoxaArb(pix6, boxa4, 2, 0, 255, 0);
pixDisplay(pix6, 1000, 100);
dewarpaDestroy(&dewa);
pixDestroy(&pixs);
pixDestroy(&pixn);
pixDestroy(&pixg);
pixDestroy(&pixb);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
pixDestroy(&pix5);
pixDestroy(&pix6);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
boxaDestroy(&boxa3);
boxaDestroy(&boxa4);
return 0;
}
main(int argc,
char **argv)
{
l_int32 i, n, w, h, success, display;
FILE *fp;
BOXA *boxa;
NUMA *naindex, *naw, *nah, *naw_med, *nah_med;
PIX *pixs, *pixt, *pixd;
if (regTestSetup(argc, argv, &fp, &display, &success, NULL))
return 1;
/* Generate arrays of word widths and heights */
pixs = pixRead("feyn.tif");
pixGetWordBoxesInTextlines(pixs, 1, 6, 6, 500, 50, &boxa, &naindex);
n = boxaGetCount(boxa);
naw = numaCreate(0);
nah = numaCreate(0);
for (i = 0; i < n; i++) {
boxaGetBoxGeometry(boxa, i, NULL, NULL, &w, &h);
numaAddNumber(naw, w);
numaAddNumber(nah, h);
}
boxaDestroy(&boxa);
numaDestroy(&naindex);
/* Make the rank bin arrays of median values, with 10 bins */
numaGetRankBinValues(naw, 10, NULL, &naw_med);
numaGetRankBinValues(nah, 10, NULL, &nah_med);
gplotSimple1(naw_med, GPLOT_PNG, "/tmp/w_10bin",
"width vs rank bins (10)");
gplotSimple1(nah_med, GPLOT_PNG, "/tmp/h_10bin",
"height vs rank bins (10)");
numaDestroy(&naw_med);
numaDestroy(&nah_med);
/* Make the rank bin arrays of median values, with 30 bins */
numaGetRankBinValues(naw, 30, NULL, &naw_med);
numaGetRankBinValues(nah, 30, NULL, &nah_med);
gplotSimple1(naw_med, GPLOT_PNG, "/tmp/w_30bin",
"width vs rank bins (30)");
gplotSimple1(nah_med, GPLOT_PNG, "/tmp/h_30bin",
"height vs rank bins (30)");
numaDestroy(&naw_med);
numaDestroy(&nah_med);
/* Give gnuplot time to write out the files */
#ifndef _WIN32
sleep(2);
#else
Sleep(2000);
#endif /* _WIN32 */
/* Save as golden files, or check against them */
regTestCheckFile(fp, argv, "/tmp/w_10bin.png", 0, &success);
regTestCheckFile(fp, argv, "/tmp/h_10bin.png", 1, &success);
regTestCheckFile(fp, argv, "/tmp/w_30bin.png", 2, &success);
regTestCheckFile(fp, argv, "/tmp/h_30bin.png", 3, &success);
/* Display results for debugging */
pixt = pixRead("/tmp/w_10bin.png");
pixDisplayWithTitle(pixt, 0, 0, NULL, display);
pixDestroy(&pixt);
pixt = pixRead("/tmp/h_10bin.png");
pixDisplayWithTitle(pixt, 650, 0, NULL, display);
pixDestroy(&pixt);
pixt = pixRead("/tmp/w_30bin.png");
pixDisplayWithTitle(pixt, 0, 550, NULL, display);
pixDestroy(&pixt);
pixt = pixRead("/tmp/h_30bin.png");
pixDisplayWithTitle(pixt, 650, 550, NULL, display);
pixDestroy(&pixt);
pixDestroy(&pixs);
numaDestroy(&naw);
numaDestroy(&nah);
regTestCleanup(argc, argv, fp, success, NULL);
return 0;
}
int main(int argc,
char **argv)
{
l_int32 i, n, w, h;
BOXA *boxa;
NUMA *naindex, *naw, *nah, *naw_med, *nah_med;
PIX *pixs, *pixt;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
/* Generate arrays of word widths and heights */
pixs = pixRead("feyn.tif");
pixGetWordBoxesInTextlines(pixs, 1, 6, 6, 500, 50, &boxa, &naindex);
n = boxaGetCount(boxa);
naw = numaCreate(0);
nah = numaCreate(0);
for (i = 0; i < n; i++) {
boxaGetBoxGeometry(boxa, i, NULL, NULL, &w, &h);
numaAddNumber(naw, w);
numaAddNumber(nah, h);
}
boxaDestroy(&boxa);
numaDestroy(&naindex);
/* Make the rank bin arrays of median values, with 10 bins */
lept_rmfile("/tmp/lept/regout/w_10bin.png"); /* remove existing ones */
lept_rmfile("/tmp/lept/regout/h_10bin.png");
lept_rmfile("/tmp/lept/regout/w_30bin.png");
lept_rmfile("/tmp/lept/regout/h_30bin.png");
numaGetRankBinValues(naw, 10, NULL, &naw_med);
numaGetRankBinValues(nah, 10, NULL, &nah_med);
gplotSimple1(naw_med, GPLOT_PNG, "/tmp/lept/regout/w_10bin",
"width vs rank bins (10)");
gplotSimple1(nah_med, GPLOT_PNG, "/tmp/lept/regout/h_10bin",
"height vs rank bins (10)");
numaDestroy(&naw_med);
numaDestroy(&nah_med);
/* Make the rank bin arrays of median values, with 30 bins */
numaGetRankBinValues(naw, 30, NULL, &naw_med);
numaGetRankBinValues(nah, 30, NULL, &nah_med);
gplotSimple1(naw_med, GPLOT_PNG, "/tmp/lept/regout/w_30bin",
"width vs rank bins (30)");
gplotSimple1(nah_med, GPLOT_PNG, "/tmp/lept/regout/h_30bin",
"height vs rank bins (30)");
numaDestroy(&naw_med);
numaDestroy(&nah_med);
/* Save as golden files, or check against them */
regTestCheckFile(rp, "/tmp/lept/regout/w_10bin.png"); /* 0 */
regTestCheckFile(rp, "/tmp/lept/regout/h_10bin.png"); /* 1 */
regTestCheckFile(rp, "/tmp/lept/regout/w_30bin.png"); /* 2 */
regTestCheckFile(rp, "/tmp/lept/regout/h_30bin.png"); /* 3 */
/* Display results for debugging */
pixt = pixRead("/tmp/lept/regout/w_10bin.png");
pixDisplayWithTitle(pixt, 0, 0, NULL, rp->display);
pixDestroy(&pixt);
pixt = pixRead("/tmp/lept/regout/h_10bin.png");
pixDisplayWithTitle(pixt, 650, 0, NULL, rp->display);
pixDestroy(&pixt);
pixt = pixRead("/tmp/lept/regout/w_30bin.png");
pixDisplayWithTitle(pixt, 0, 550, NULL, rp->display);
pixDestroy(&pixt);
pixt = pixRead("/tmp/lept/regout/h_30bin.png");
pixDisplayWithTitle(pixt, 650, 550, NULL, rp->display);
pixDestroy(&pixt);
pixDestroy(&pixs);
numaDestroy(&naw);
numaDestroy(&nah);
return regTestCleanup(rp);
}
