main(int argc,
char **argv)
{
char *filein, *fileout;
l_int32 w, h, d, w2, h2, i, ncols;
l_float32 angle, conf;
BOX *box;
BOXA *boxa, *boxas, *boxad, *boxa2;
NUMA *numa;
PIX *pixs, *pixt, *pixb, *pixb2, *pixd;
PIX *pixtlm, *pixvws;
PIX *pixt1, *pixt2, *pixt3, *pixt4, *pixt5, *pixt6;
PIXA *pixam, *pixac, *pixad, *pixat;
PIXAA *pixaa, *pixaa2;
PTA *pta;
SEL *selsplit;
static char mainName[] = "textlinemask";
if (argc != 3)
exit(ERROR_INT(" Syntax: textlinemask filein fileout", mainName, 1));
filein = argv[1];
fileout = argv[2];
pixDisplayWrite(NULL, -1); /* init debug output */
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
pixGetDimensions(pixs, &w, &h, &d);
/* Binarize input */
if (d == 8)
pixt = pixThresholdToBinary(pixs, 128);
else if (d == 1)
pixt = pixClone(pixs);
else {
fprintf(stderr, "depth is %d\n", d);
exit(1);
}
/* Deskew */
pixb = pixFindSkewAndDeskew(pixt, 1, &angle, &conf);
pixDestroy(&pixt);
fprintf(stderr, "Skew angle: %7.2f degrees; %6.2f conf\n", angle, conf);
pixDisplayWrite(pixb, DEBUG_OUTPUT);
#if 1
/* Use full image morphology to find columns, at 2x reduction.
* This only works for very simple layouts where each column
* of text extends the full height of the input image.
* pixam has a pix component over each column. */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixt1 = pixMorphCompSequence(pixb2, "c5.500", 0);
boxa = pixConnComp(pixt1, &pixam, 8);
ncols = boxaGetCount(boxa);
fprintf(stderr, "Num columns: %d\n", ncols);
pixDisplayWrite(pixt1, DEBUG_OUTPUT);
/* Use selective region-based morphology to get the textline mask. */
pixad = pixaMorphSequenceByRegion(pixb2, pixam, "c100.3", 0, 0);
pixGetDimensions(pixb2, &w2, &h2, NULL);
if (DEBUG_OUTPUT) {
pixt2 = pixaDisplay(pixad, w2, h2);
pixDisplayWrite(pixt2, DEBUG_OUTPUT);
pixDestroy(&pixt2);
}
/* Some of the lines may be touching, so use a HMT to split the
* lines in each column, and use a pixaa to save the results. */
selsplit = selCreateFromString(seltext, 17, 7, "selsplit");
pixaa = pixaaCreate(ncols);
for (i = 0; i < ncols; i++) {
pixt3 = pixaGetPix(pixad, i, L_CLONE);
box = pixaGetBox(pixad, i, L_COPY);
pixt4 = pixHMT(NULL, pixt3, selsplit);
pixXor(pixt4, pixt4, pixt3);
boxa2 = pixConnComp(pixt4, &pixac, 8);
pixaaAddPixa(pixaa, pixac, L_INSERT);
pixaaAddBox(pixaa, box, L_INSERT);
if (DEBUG_OUTPUT) {
pixt5 = pixaDisplayRandomCmap(pixac, 0, 0);
pixDisplayWrite(pixt5, DEBUG_OUTPUT);
fprintf(stderr, "Num textlines in col %d: %d\n", i,
boxaGetCount(boxa2));
pixDestroy(&pixt5);
}
pixDestroy(&pixt3);
pixDestroy(&pixt4);
boxaDestroy(&boxa2);
}
/* Visual output */
if (DEBUG_OUTPUT) {
pixDisplayMultiple("/tmp/junk_write_display*");
pixat = pixaReadFiles("/tmp", "junk_write_display");
pixt5 = selDisplayInPix(selsplit, 31, 2);
pixaAddPix(pixat, pixt5, L_INSERT);
pixt6 = pixaDisplayTiledAndScaled(pixat, 32, 400, 3, 0, 35, 3);
pixWrite(fileout, pixt6, IFF_PNG);
pixaDestroy(&pixat);
pixDestroy(&pixt6);
}
/* Test pixaa I/O */
pixaaWrite("/tmp/junkpixaa", pixaa);
pixaa2 = pixaaRead("/tmp/junkpixaa");
pixaaWrite("/tmp/junkpixaa2", pixaa2);
/* Test pixaa display */
pixd = pixaaDisplay(pixaa, w2, h2);
pixWrite("/tmp/junkdisplay", pixd, IFF_PNG);
pixDestroy(&pixd);
/* Cleanup */
pixDestroy(&pixb2);
pixDestroy(&pixt1);
pixaDestroy(&pixam);
pixaDestroy(&pixad);
pixaaDestroy(&pixaa);
pixaaDestroy(&pixaa2);
boxaDestroy(&boxa);
selDestroy(&selsplit);
#endif
#if 0
/* Use the baseline finder; not really what is needed */
numa = pixFindBaselines(pixb, &pta, 1);
#endif
#if 0
/* Use the textline mask function; parameters are not quite right */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixtlm = pixGenTextlineMask(pixb2, &pixvws, NULL, 1);
pixDisplay(pixtlm, 0, 100);
pixDisplay(pixvws, 500, 100);
pixDestroy(&pixb2);
pixDestroy(&pixtlm);
pixDestroy(&pixvws);
#endif
#if 0
/* Use the Breuel whitespace partition method; slow and we would
* still need to work to extract the fg regions. */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
boxas = pixConnComp(pixb2, NULL, 8);
boxad = boxaGetWhiteblocks(boxas, NULL, L_SORT_BY_HEIGHT,
3, 0.1, 200, 0.2, 0);
pixd = pixDrawBoxa(pixb2, boxad, 7, 0xe0708000);
pixDisplay(pixd, 100, 500);
pixDestroy(&pixb2);
pixDestroy(&pixd);
boxaDestroy(&boxas);
boxaDestroy(&boxad);
#endif
#if 0
/* Use morphology to find columns and then selective
* region-based morphology to get the textline mask.
* This is for display; we really want to get a pixa of the
* specific textline masks. */
startTimer();
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixt1 = pixMorphCompSequence(pixb2, "c5.500", 0); /* column mask */
pixt2 = pixMorphSequenceByRegion(pixb2, pixt1, "c100.3", 8, 0, 0, &boxa);
fprintf(stderr, "time = %7.3f sec\n", stopTimer());
pixDisplay(pixt1, 100, 500);
pixDisplay(pixt2, 800, 500);
pixDestroy(&pixb2);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
boxaDestroy(&boxa);
#endif
pixDestroy(&pixs);
pixDestroy(&pixb);
exit(0);
}
int main(int argc,
char **argv)
{
l_int32 w, h, d, w2, h2, i, ncols, ret;
l_float32 angle, conf;
BOX *box;
BOXA *boxa, *boxa2;
PIX *pix, *pixs, *pixb, *pixb2, *pixd;
PIX *pix1, *pix2, *pix3, *pix4, *pix5, *pix6;
PIXA *pixam; /* mask with a single component over each column */
PIXA *pixac, *pixad, *pixat;
PIXAA *pixaa, *pixaa2;
SEL *selsplit;
static char mainName[] = "arabic_lines";
if (argc != 1)
return ERROR_INT(" Syntax: arabic_lines", mainName, 1);
pixDisplayWrite(NULL, -1); /* init debug output */
/* Binarize input */
pixs = pixRead("arabic.png");
pixGetDimensions(pixs, &w, &h, &d);
pix = pixConvertTo1(pixs, 128);
/* Deskew */
pixb = pixFindSkewAndDeskew(pix, 1, &angle, &conf);
pixDestroy(&pix);
fprintf(stderr, "Skew angle: %7.2f degrees; %6.2f conf\n", angle, conf);
pixDisplayWrite(pixb, 1);
/* Use full image morphology to find columns, at 2x reduction.
This only works for very simple layouts where each column
of text extends the full height of the input image. */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pix1 = pixMorphCompSequence(pixb2, "c5.500", 0);
boxa = pixConnComp(pix1, &pixam, 8);
ncols = boxaGetCount(boxa);
fprintf(stderr, "Num columns: %d\n", ncols);
pixDisplayWrite(pix1, 1);
/* Use selective region-based morphology to get the textline mask. */
pixad = pixaMorphSequenceByRegion(pixb2, pixam, "c100.3", 0, 0);
pixGetDimensions(pixb2, &w2, &h2, NULL);
pix2 = pixaDisplay(pixad, w2, h2);
pixDisplayWrite(pix2, 1);
pixDestroy(&pix2);
/* Some of the lines may be touching, so use a HMT to split the
lines in each column, and use a pixaa to save the results. */
selsplit = selCreateFromString(seltext, 17, 7, "selsplit");
pixaa = pixaaCreate(ncols);
for (i = 0; i < ncols; i++) {
pix3 = pixaGetPix(pixad, i, L_CLONE);
box = pixaGetBox(pixad, i, L_COPY);
pix4 = pixHMT(NULL, pix3, selsplit);
pixXor(pix4, pix4, pix3);
boxa2 = pixConnComp(pix4, &pixac, 8);
pixaaAddPixa(pixaa, pixac, L_INSERT);
pixaaAddBox(pixaa, box, L_INSERT);
pix5 = pixaDisplayRandomCmap(pixac, 0, 0);
pixDisplayWrite(pix5, 1);
fprintf(stderr, "Num textlines in col %d: %d\n", i,
boxaGetCount(boxa2));
pixDestroy(&pix5);
pixDestroy(&pix3);
pixDestroy(&pix4);
boxaDestroy(&boxa2);
}
/* Visual output */
ret = system("gthumb /tmp/display/file* &");
pixat = pixaReadFiles("/tmp/display", "file");
pix5 = selDisplayInPix(selsplit, 31, 2);
pixaAddPix(pixat, pix5, L_INSERT);
pix6 = pixaDisplayTiledAndScaled(pixat, 32, 400, 3, 0, 35, 3);
pixWrite("/tmp/result.png", pix6, IFF_PNG);
pixaDestroy(&pixat);
pixDestroy(&pix6);
/* Test pixaa I/O */
pixaaWrite("/tmp/pixaa", pixaa);
pixaa2 = pixaaRead("/tmp/pixaa");
pixaaWrite("/tmp/pixaa2", pixaa2);
/* Test pixaa display */
pixd = pixaaDisplay(pixaa, w2, h2);
pixWrite("/tmp/textlines.png", pixd, IFF_PNG);
pixDestroy(&pixd);
/* Cleanup */
pixDestroy(&pixb2);
pixDestroy(&pix1);
pixaDestroy(&pixam);
pixaDestroy(&pixad);
pixaaDestroy(&pixaa);
pixaaDestroy(&pixaa2);
boxaDestroy(&boxa);
selDestroy(&selsplit);
pixDestroy(&pixs);
pixDestroy(&pixb);
return 0;
}
// Performs line segmentation
bool CubeLineSegmenter::LineSegment() {
// Use full image morphology to find columns
// This only works for simple layouts where each column
// of text extends the full height of the input image.
Pix *pix_temp1 = pixMorphCompSequence(img_, "c5.500", 0);
if (pix_temp1 == NULL) {
return false;
}
// Mask with a single component over each column
Pixa *pixam;
Boxa *boxa = pixConnComp(pix_temp1, &pixam, 8);
if (boxa == NULL) {
return false;
}
int init_morph_min_hgt = kLineSepMorphMinHgt;
char sequence_str[16];
sprintf(sequence_str, "c100.%d", init_morph_min_hgt);
// Use selective region-based morphology to get the textline mask.
Pixa *pixad = pixaMorphSequenceByRegion(img_, pixam, sequence_str, 0, 0);
if (pixad == NULL) {
return false;
}
// for all columns
int col_cnt = boxaGetCount(boxa);
// create columns
columns_ = pixaaCreate(col_cnt);
if (columns_ == NULL) {
return false;
}
// index columns based on readind order (RTL)
int *col_order = IndexRTL(pixad);
if (col_order == NULL) {
return false;
}
line_cnt_ = 0;
for (int col_idx = 0; col_idx < col_cnt; col_idx++) {
int col = col_order[col_idx];
// get the pix and box corresponding to the column
Pix *pixt3 = pixaGetPix(pixad, col, L_CLONE);
if (pixt3 == NULL) {
return false;
}
Box *col_box = pixad->boxa->box[col];
Pixa *pixac;
Boxa *boxa2 = pixConnComp(pixt3, &pixac, 8);
if (boxa2 == NULL) {
return false;
}
// offset the boxes by the column box
for (int line = 0; line < pixac->n; line++) {
pixac->boxa->box[line]->x += col_box->x;
pixac->boxa->box[line]->y += col_box->y;
}
// add the lines
if (AddLines(pixac) == true) {
if (pixaaAddBox(columns_, col_box, L_CLONE) != 0) {
return false;
}
}
pixDestroy(&pixt3);
boxaDestroy(&boxa2);
line_cnt_ += columns_->pixa[col_idx]->n;
}
pixaDestroy(&pixam);
pixaDestroy(&pixad);
boxaDestroy(&boxa);
delete []col_order;
pixDestroy(&pix_temp1);
return true;
}