// Helper to remove an enclosing circle from an image.
// If there isn't one, then the image will most likely get badly mangled.
// The returned pix must be pixDestroyed after use. NULL may be returned
// if the image doesn't meet the trivial conditions that it uses to determine
// success.
static Pix* RemoveEnclosingCircle(Pix* pixs) {
Pix* pixsi = pixInvert(NULL, pixs);
Pix* pixc = pixCreateTemplate(pixs);
pixSetOrClearBorder(pixc, 1, 1, 1, 1, PIX_SET);
pixSeedfillBinary(pixc, pixc, pixsi, 4);
pixInvert(pixc, pixc);
pixDestroy(&pixsi);
Pix* pixt = pixAnd(NULL, pixs, pixc);
l_int32 max_count;
pixCountConnComp(pixt, 8, &max_count);
// The count has to go up before we start looking for the minimum.
l_int32 min_count = MAX_INT32;
Pix* pixout = NULL;
for (int i = 1; i < kMaxCircleErosions; i++) {
pixDestroy(&pixt);
pixErodeBrick(pixc, pixc, 3, 3);
pixt = pixAnd(NULL, pixs, pixc);
l_int32 count;
pixCountConnComp(pixt, 8, &count);
if (i == 1 || count > max_count) {
max_count = count;
min_count = count;
} else if (i > 1 && count < min_count) {
min_count = count;
pixDestroy(&pixout);
pixout = pixCopy(NULL, pixt); // Save the best.
} else if (count >= min_count) {
break; // We have passed by the best.
}
}
pixDestroy(&pixt);
pixDestroy(&pixc);
return pixout;
}
int main(int argc,
char **argv)
{
l_uint8 *array1, *array2;
l_int32 n1, n2, n3;
size_t size1, size2;
FILE *fp;
BOXA *boxa1, *boxa2;
PIX *pixs, *pix1;
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, 100);
boxaDestroy(&boxa1);
pixDestroy(&pix1);
pixaDestroy(&pixa1);
pixDestroy(&pixs);
return regTestCleanup(rp);
}
int main(int argc,
char **argv)
{
char *filein;
l_int32 i, n, count;
BOX *box;
BOXA *boxa;
PIX *pixs, *pixd;
PIXA *pixa;
PIXCMAP *cmap;
static char mainName[] = "cctest1";
if (argc != 2)
return ERROR_INT(" Syntax: cctest1 filein", mainName, 1);
filein = argv[1];
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
if (pixGetDepth(pixs) != 1)
exit(ERROR_INT("pixs not 1 bpp", mainName, 1));
/* Test speed of pixCountConnComp() */
startTimer();
for (i = 0; i < NTIMES; i++)
pixCountConnComp(pixs, 4, &count);
fprintf(stderr, "Time to compute 4-cc: %6.3f sec\n", stopTimer()/NTIMES);
fprintf(stderr, "Number of 4-cc: %d\n", count);
startTimer();
for (i = 0; i < NTIMES; i++)
pixCountConnComp(pixs, 8, &count);
fprintf(stderr, "Time to compute 8-cc: %6.3f sec\n", stopTimer()/NTIMES);
fprintf(stderr, "Number of 8-cc: %d\n", count);
/* Test speed of pixConnComp(), with only boxa output */
startTimer();
for (i = 0; i < NTIMES; i++) {
boxa = pixConnComp(pixs, NULL, 4);
boxaDestroy(&boxa);
}
fprintf(stderr, "Time to compute 4-cc: %6.3f sec\n", stopTimer()/NTIMES);
startTimer();
for (i = 0; i < NTIMES; i++) {
boxa = pixConnComp(pixs, NULL, 8);
boxaDestroy(&boxa);
}
fprintf(stderr, "Time to compute 8-cc: %6.3f sec\n", stopTimer()/NTIMES);
/* Draw outline of each c.c. box */
boxa = pixConnComp(pixs, NULL, 4);
n = boxaGetCount(boxa);
fprintf(stderr, "Num 4-cc boxes: %d\n", n);
for (i = 0; i < n; i++) {
box = boxaGetBox(boxa, i, L_CLONE);
pixRenderBox(pixs, box, 3, L_FLIP_PIXELS);
boxDestroy(&box); /* remember, clones need to be destroyed */
}
pixDisplayWrite(pixs, 1);
boxaDestroy(&boxa);
/* Display each component as a random color in cmapped 8 bpp.
* Background is color 0; it is set to white. */
boxa = pixConnComp(pixs, &pixa, 4);
pixd = pixaDisplayRandomCmap(pixa, pixGetWidth(pixs), pixGetHeight(pixs));
cmap = pixGetColormap(pixd);
pixcmapResetColor(cmap, 0, 255, 255, 255); /* reset background to white */
pixDisplay(pixd, 100, 100);
pixDisplayWrite(pixd, 1);
boxaDestroy(&boxa);
pixDestroy(&pixd);
pixaDestroy(&pixa);
pixDestroy(&pixs);
return 0;
}
/*!
* \brief pixThresholdByConnComp()
*
* \param[in] pixs depth > 1, colormap OK
* \param[in] pixm [optional] 1 bpp mask giving region to ignore by setting
* pixels to white; use NULL if no mask
* \param[in] start, end, incr binarization threshold levels to test
* \param[in] thresh48 threshold on normalized difference between the
* numbers of 4 and 8 connected components
* \param[in] threshdiff threshold on normalized difference between the
* number of 4 cc at successive iterations
* \param[out] pglobthresh [optional] best global threshold; 0
* if no threshold is found
* \param[out] ppixd [optional] image thresholded to binary, or
* null if no threshold is found
* \param[in] debugflag 1 for plotted results
* \return 0 if OK, 1 on error or if no threshold is found
*
* <pre>
* Notes:
* (1) This finds a global threshold based on connected components.
* Although slow, it is reasonable to use it in a situation where
* (a) the background in the image is relatively uniform, and
* (b) the result will be fed to an OCR program that accepts 1 bpp
* images and works best with easily segmented characters.
* The reason for (b) is that this selects a threshold with a
* minimum number of both broken characters and merged characters.
* (2) If the pix has color, it is converted to gray using the
* max component.
* (3) Input 0 to use default values for any of these inputs:
* %start, %end, %incr, %thresh48, %threshdiff.
* (4) This approach can be understood as follows. When the
* binarization threshold is varied, the numbers of c.c. identify
* four regimes:
* (a) For low thresholds, text is broken into small pieces, and
* the number of c.c. is large, with the 4 c.c. significantly
* exceeding the 8 c.c.
* (b) As the threshold rises toward the optimum value, the text
* characters coalesce and there is very little difference
* between the numbers of 4 and 8 c.c, which both go
* through a minimum.
* (c) Above this, the image background gets noisy because some
* pixels are(thresholded to foreground, and the numbers
* of c.c. quickly increase, with the 4 c.c. significantly
* larger than the 8 c.c.
* (d) At even higher thresholds, the image background noise
* coalesces as it becomes mostly foreground, and the
* number of c.c. drops quickly.
* (5) If there is no global threshold that distinguishes foreground
* text from background (e.g., weak text over a background that
* has significant variation and/or bleedthrough), this returns 1,
* which the caller should check.
* </pre>
*/
l_int32
pixThresholdByConnComp(PIX *pixs,
PIX *pixm,
l_int32 start,
l_int32 end,
l_int32 incr,
l_float32 thresh48,
l_float32 threshdiff,
l_int32 *pglobthresh,
PIX **ppixd,
l_int32 debugflag)
{
l_int32 i, thresh, n, n4, n8, mincounts, found, globthresh;
l_float32 count4, count8, firstcount4, prevcount4, diff48, diff4;
GPLOT *gplot;
NUMA *na4, *na8;
PIX *pix1, *pix2, *pix3;
PROCNAME("pixThresholdByConnComp");
if (pglobthresh) *pglobthresh = 0;
if (ppixd) *ppixd = NULL;
if (!pixs || pixGetDepth(pixs) == 1)
return ERROR_INT("pixs undefined or 1 bpp", procName, 1);
if (pixm && pixGetDepth(pixm) != 1)
return ERROR_INT("pixm must be 1 bpp", procName, 1);
/* Assign default values if requested */
if (start <= 0) start = 80;
if (end <= 0) end = 200;
if (incr <= 0) incr = 10;
if (thresh48 <= 0.0) thresh48 = 0.01;
if (threshdiff <= 0.0) threshdiff = 0.01;
if (start > end)
return ERROR_INT("invalid start,end", procName, 1);
/* Make 8 bpp, using the max component if color. */
if (pixGetColormap(pixs))
pix1 = pixRemoveColormap(pixs, REMOVE_CMAP_BASED_ON_SRC);
else
pix1 = pixClone(pixs);
if (pixGetDepth(pix1) == 32)
pix2 = pixConvertRGBToGrayMinMax(pix1, L_CHOOSE_MAX);
else
pix2 = pixConvertTo8(pix1, 0);
pixDestroy(&pix1);
/* Mask out any non-text regions. Do this in-place, because pix2
* can never be the same pix as pixs. */
if (pixm)
pixSetMasked(pix2, pixm, 255);
/* Make sure there are enough components to get a valid signal */
pix3 = pixConvertTo1(pix2, start);
pixCountConnComp(pix3, 4, &n4);
pixDestroy(&pix3);
mincounts = 500;
if (n4 < mincounts) {
L_INFO("Insufficient component count: %d\n", procName, n4);
pixDestroy(&pix2);
return 1;
}
/* Compute the c.c. data */
na4 = numaCreate(0);
na8 = numaCreate(0);
numaSetParameters(na4, start, incr);
numaSetParameters(na8, start, incr);
for (thresh = start, i = 0; thresh <= end; thresh += incr, i++) {
pix3 = pixConvertTo1(pix2, thresh);
pixCountConnComp(pix3, 4, &n4);
pixCountConnComp(pix3, 8, &n8);
numaAddNumber(na4, n4);
numaAddNumber(na8, n8);
pixDestroy(&pix3);
}
if (debugflag) {
gplot = gplotCreate("/tmp/threshroot", GPLOT_PNG,
"number of cc vs. threshold",
"threshold", "number of cc");
gplotAddPlot(gplot, NULL, na4, GPLOT_LINES, "plot 4cc");
gplotAddPlot(gplot, NULL, na8, GPLOT_LINES, "plot 8cc");
gplotMakeOutput(gplot);
gplotDestroy(&gplot);
}
n = numaGetCount(na4);
found = FALSE;
for (i = 0; i < n; i++) {
if (i == 0) {
numaGetFValue(na4, i, &firstcount4);
prevcount4 = firstcount4;
} else {
numaGetFValue(na4, i, &count4);
numaGetFValue(na8, i, &count8);
diff48 = (count4 - count8) / firstcount4;
diff4 = L_ABS(prevcount4 - count4) / firstcount4;
if (debugflag) {
fprintf(stderr, "diff48 = %7.3f, diff4 = %7.3f\n",
diff48, diff4);
}
if (diff48 < thresh48 && diff4 < threshdiff) {
found = TRUE;
break;
}
prevcount4 = count4;
}
}
numaDestroy(&na4);
numaDestroy(&na8);
if (found) {
globthresh = start + i * incr;
if (pglobthresh) *pglobthresh = globthresh;
if (ppixd) {
*ppixd = pixConvertTo1(pix2, globthresh);
pixCopyResolution(*ppixd, pixs);
}
if (debugflag) fprintf(stderr, "global threshold = %d\n", globthresh);
pixDestroy(&pix2);
return 0;
}
if (debugflag) fprintf(stderr, "no global threshold found\n");
pixDestroy(&pix2);
return 1;
}