static void
GetImageMask(PIX *pixs,
l_int32 res,
BOXA **pboxa,
const char *debugfile)
{
PIX *pix1, *pix2, *pix3, *pix4;
PIXA *pixa;
pixSetResolution(pixs, 200, 200);
pix1 = pixConvertTo1(pixs, 100);
pix2 = pixGenHalftoneMask(pix1, NULL, NULL, 0);
pix3 = pixMorphSequence(pix2, "c20.1 + c1.20", 0);
*pboxa = pixConnComp(pix3, NULL, 8);
if (debugfile) {
pixa = pixaCreate(0);
pixaAddPix(pixa, pixs, L_COPY);
pixaAddPix(pixa, pix1, L_INSERT);
pixaAddPix(pixa, pix2, L_INSERT);
pixaAddPix(pixa, pix3, L_INSERT);
pix4 = pixaDisplayTiledInRows(pixa, 32, 1800, 0.25, 0, 25, 2);
pixWrite(debugfile, pix4, IFF_JFIF_JPEG);
pixDisplay(pix4, 100, 100);
pixDestroy(&pix4);
pixaDestroy(&pixa);
} else {
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
}
return;
}
/*!
* pixGetRegionsBinary()
*
* Input: pixs (1 bpp, assumed to be 300 to 400 ppi)
* &pixhm (<optional return> halftone mask)
* &pixtm (<optional return> textline mask)
* &pixtb (<optional return> textblock mask)
* debug (flag: set to 1 for debug output)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) It is best to deskew the image before segmenting.
* (2) The debug flag enables a number of outputs. These
* are included to show how to generate and save/display
* these results.
*/
l_int32
pixGetRegionsBinary(PIX *pixs,
PIX **ppixhm,
PIX **ppixtm,
PIX **ppixtb,
l_int32 debug)
{
char *tempname;
l_int32 htfound, tlfound;
PIX *pixr, *pixt1, *pixt2;
PIX *pixtext; /* text pixels only */
PIX *pixhm2; /* halftone mask; 2x reduction */
PIX *pixhm; /* halftone mask; */
PIX *pixtm2; /* textline mask; 2x reduction */
PIX *pixtm; /* textline mask */
PIX *pixvws; /* vertical white space mask */
PIX *pixtb2; /* textblock mask; 2x reduction */
PIX *pixtbf2; /* textblock mask; 2x reduction; small comps filtered */
PIX *pixtb; /* textblock mask */
PROCNAME("pixGetRegionsBinary");
if (ppixhm) *ppixhm = NULL;
if (ppixtm) *ppixtm = NULL;
if (ppixtb) *ppixtb = NULL;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (pixGetDepth(pixs) != 1)
return ERROR_INT("pixs not 1 bpp", procName, 1);
/* 2x reduce, to 150 -200 ppi */
pixr = pixReduceRankBinaryCascade(pixs, 1, 0, 0, 0);
pixDisplayWrite(pixr, debug);
/* Get the halftone mask */
pixhm2 = pixGenHalftoneMask(pixr, &pixtext, &htfound, debug);
/* Get the textline mask from the text pixels */
pixtm2 = pixGenTextlineMask(pixtext, &pixvws, &tlfound, debug);
/* Get the textblock mask from the textline mask */
pixtb2 = pixGenTextblockMask(pixtm2, pixvws, debug);
pixDestroy(&pixr);
pixDestroy(&pixtext);
pixDestroy(&pixvws);
/* Remove small components from the mask, where a small
* component is defined as one with both width and height < 60 */
pixtbf2 = pixSelectBySize(pixtb2, 60, 60, 4, L_SELECT_IF_EITHER,
L_SELECT_IF_GTE, NULL);
pixDestroy(&pixtb2);
pixDisplayWriteFormat(pixtbf2, debug, IFF_PNG);
/* Expand all masks to full resolution, and do filling or
* small dilations for better coverage. */
pixhm = pixExpandReplicate(pixhm2, 2);
pixt1 = pixSeedfillBinary(NULL, pixhm, pixs, 8);
pixOr(pixhm, pixhm, pixt1);
pixDestroy(&pixt1);
pixDisplayWriteFormat(pixhm, debug, IFF_PNG);
pixt1 = pixExpandReplicate(pixtm2, 2);
pixtm = pixDilateBrick(NULL, pixt1, 3, 3);
pixDestroy(&pixt1);
pixDisplayWriteFormat(pixtm, debug, IFF_PNG);
pixt1 = pixExpandReplicate(pixtbf2, 2);
pixtb = pixDilateBrick(NULL, pixt1, 3, 3);
pixDestroy(&pixt1);
pixDisplayWriteFormat(pixtb, debug, IFF_PNG);
pixDestroy(&pixhm2);
pixDestroy(&pixtm2);
pixDestroy(&pixtbf2);
/* Debug: identify objects that are neither text nor halftone image */
if (debug) {
pixt1 = pixSubtract(NULL, pixs, pixtm); /* remove text pixels */
pixt2 = pixSubtract(NULL, pixt1, pixhm); /* remove halftone pixels */
pixDisplayWriteFormat(pixt2, 1, IFF_PNG);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
}
/* Debug: display textline components with random colors */
if (debug) {
l_int32 w, h;
BOXA *boxa;
PIXA *pixa;
boxa = pixConnComp(pixtm, &pixa, 8);
pixGetDimensions(pixtm, &w, &h, NULL);
pixt1 = pixaDisplayRandomCmap(pixa, w, h);
pixcmapResetColor(pixGetColormap(pixt1), 0, 255, 255, 255);
pixDisplay(pixt1, 100, 100);
pixDisplayWriteFormat(pixt1, 1, IFF_PNG);
pixaDestroy(&pixa);
boxaDestroy(&boxa);
pixDestroy(&pixt1);
}
/* Debug: identify the outlines of each textblock */
if (debug) {
PIXCMAP *cmap;
PTAA *ptaa;
ptaa = pixGetOuterBordersPtaa(pixtb);
tempname = genTempFilename("/tmp", "tb_outlines.ptaa", 0, 0);
ptaaWrite(tempname, ptaa, 1);
FREE(tempname);
pixt1 = pixRenderRandomCmapPtaa(pixtb, ptaa, 1, 16, 1);
cmap = pixGetColormap(pixt1);
pixcmapResetColor(cmap, 0, 130, 130, 130);
pixDisplay(pixt1, 500, 100);
pixDisplayWriteFormat(pixt1, 1, IFF_PNG);
pixDestroy(&pixt1);
ptaaDestroy(&ptaa);
}
/* Debug: get b.b. for all mask components */
if (debug) {
BOXA *bahm, *batm, *batb;
bahm = pixConnComp(pixhm, NULL, 4);
batm = pixConnComp(pixtm, NULL, 4);
batb = pixConnComp(pixtb, NULL, 4);
tempname = genTempFilename("/tmp", "htmask.boxa", 0, 0);
boxaWrite(tempname, bahm);
FREE(tempname);
tempname = genTempFilename("/tmp", "textmask.boxa", 0, 0);
boxaWrite(tempname, batm);
FREE(tempname);
tempname = genTempFilename("/tmp", "textblock.boxa", 0, 0);
boxaWrite(tempname, batb);
FREE(tempname);
boxaDestroy(&bahm);
boxaDestroy(&batm);
boxaDestroy(&batb);
}
if (ppixhm)
*ppixhm = pixhm;
else
pixDestroy(&pixhm);
if (ppixtm)
*ppixtm = pixtm;
else
pixDestroy(&pixtm);
if (ppixtb)
*ppixtb = pixtb;
else
pixDestroy(&pixtb);
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);
}
main(int argc,
char **argv)
{
char buffer[512];
char *tempfile1, *tempfile2;
l_uint8 *data;
l_int32 i, j, w, h, seq, ret, same;
size_t nbytes;
const char *title;
BOX *box;
BOXA *boxa1, *boxa2;
L_BYTEA *ba;
L_PDF_DATA *lpd;
PIX *pix1, *pix2, *pix3, *pix4, *pix5, *pix6;
PIX *pixs, *pixt, *pixg, *pixgc, *pixc;
static char mainName[] = "pdfiotest";
if (argc != 1)
exit(ERROR_INT("syntax: pdfiotest", mainName, 1));
l_pdfSetDateAndVersion(0);
#if 1
/* --------------- Single image tests ------------------- */
fprintf(stderr, "\n*** Writing single images as pdf files\n");
convertToPdf("weasel2.4c.png", L_FLATE_ENCODE, 0, "/tmp/pdffile01.pdf",
0, 0, 72, NULL, 0, "weasel2.4c.png");
convertToPdf("test24.jpg", L_JPEG_ENCODE, 0, "/tmp/pdffile02.pdf",
0, 0, 72, NULL, 0, "test24.jpg");
convertToPdf("feyn.tif", L_G4_ENCODE, 0, "/tmp/pdffile03.pdf",
0, 0, 300, NULL, 0, "feyn.tif");
pixs = pixRead("feyn.tif");
pixConvertToPdf(pixs, L_G4_ENCODE, 0, "/tmp/pdffile04.pdf", 0, 0, 300,
NULL, 0, "feyn.tif");
pixDestroy(&pixs);
pixs = pixRead("test24.jpg");
pixConvertToPdf(pixs, L_JPEG_ENCODE, 5, "/tmp/pdffile05.pdf", 0, 0, 72,
NULL, 0, "test24.jpg");
pixDestroy(&pixs);
pixs = pixRead("feyn.tif");
pixt = pixScaleToGray2(pixs);
pixWrite("junkfeyn8.png", pixt, IFF_PNG);
convertToPdf("junkfeyn8.png", L_JPEG_ENCODE, 0, "/tmp/pdffile06.pdf",
0, 0, 150, NULL, 0, "junkfeyn8.png");
pixDestroy(&pixs);
pixDestroy(&pixt);
convertToPdf("weasel4.16g.png", L_FLATE_ENCODE, 0, "/tmp/pdffile07.pdf",
0, 0, 30, NULL, 0, "weasel4.16g.png");
pixs = pixRead("test24.jpg");
pixg = pixConvertTo8(pixs, 0);
box = boxCreate(100, 100, 100, 100);
pixc = pixClipRectangle(pixs, box, NULL);
pixgc = pixClipRectangle(pixg, box, NULL);
pixWrite("junkpix32.jpg", pixc, IFF_JFIF_JPEG);
pixWrite("junkpix8.jpg", pixgc, IFF_JFIF_JPEG);
convertToPdf("junkpix32.jpg", L_FLATE_ENCODE, 0, "/tmp/pdffile08.pdf",
0, 0, 72, NULL, 0, "junkpix32.jpg");
convertToPdf("junkpix8.jpg", L_FLATE_ENCODE, 0, "/tmp/pdffile09.pdf",
0, 0, 72, NULL, 0, "junkpix8.jpg");
pixDestroy(&pixs);
pixDestroy(&pixg);
pixDestroy(&pixc);
pixDestroy(&pixgc);
boxDestroy(&box);
#endif
#if 1
/* --------------- Multiple image tests ------------------- */
fprintf(stderr, "\n*** Writing multiple images as single page pdf files\n");
pix1 = pixRead("feyn-fract.tif");
pix2 = pixRead("weasel8.240c.png");
/* l_pdfSetDateAndVersion(0); */
/* First, write the 1 bpp image through the mask onto the weasels */
for (i = 0; i < 5; i++) {
for (j = 0; j < 10; j++) {
seq = (i == 0 && j == 0) ? L_FIRST_IMAGE : L_NEXT_IMAGE;
title = (i == 0 && j == 0) ? "feyn-fract.tif" : NULL;
pixConvertToPdf(pix2, L_FLATE_ENCODE, 0, NULL, 100 * j,
100 * i, 70, &lpd, seq, title);
}
}
pixConvertToPdf(pix1, L_G4_ENCODE, 0, "/tmp/pdffile10.pdf", 0, 0, 80, &lpd,
L_LAST_IMAGE, NULL);
/* Now, write the 1 bpp image over the weasels */
l_pdfSetG4ImageMask(0);
for (i = 0; i < 5; i++) {
for (j = 0; j < 10; j++) {
seq = (i == 0 && j == 0) ? L_FIRST_IMAGE : L_NEXT_IMAGE;
title = (i == 0 && j == 0) ? "feyn-fract.tif" : NULL;
pixConvertToPdf(pix2, L_FLATE_ENCODE, 0, NULL, 100 * j,
100 * i, 70, &lpd, seq, title);
}
}
pixConvertToPdf(pix1, L_G4_ENCODE, 0, "/tmp/pdffile11.pdf", 0, 0, 80, &lpd,
L_LAST_IMAGE, NULL);
l_pdfSetG4ImageMask(1);
pixDestroy(&pix1);
pixDestroy(&pix2);
#endif
#if 1
/* -------- pdf convert segmented with no image regions -------- */
fprintf(stderr, "\n*** Writing segmented images without image regions\n");
pix1 = pixRead("rabi.png");
pix2 = pixScaleToGray2(pix1);
pixWrite("/tmp/rabi8.jpg", pix2, IFF_JFIF_JPEG);
pix3 = pixThresholdTo4bpp(pix2, 16, 1);
pixWrite("/tmp/rabi4.png", pix3, IFF_PNG);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* 1 bpp input */
convertToPdfSegmented("rabi.png", 300, L_G4_ENCODE, 128, NULL, 0, 0,
"/tmp/pdffile12.pdf");
convertToPdfSegmented("rabi.png", 300, L_JPEG_ENCODE, 128, NULL, 0, 0,
"/tmp/pdffile13.pdf");
convertToPdfSegmented("rabi.png", 300, L_FLATE_ENCODE, 128, NULL, 0, 0,
"/tmp/pdffile14.pdf");
/* 8 bpp input, no cmap */
convertToPdfSegmented("/tmp/rabi8.jpg", 150, L_G4_ENCODE, 128,
NULL, 0, 0, "/tmp/pdffile15.pdf");
convertToPdfSegmented("/tmp/rabi8.jpg", 150, L_JPEG_ENCODE, 128,
NULL, 0, 0, "/tmp/pdffile16.pdf");
convertToPdfSegmented("/tmp/rabi8.jpg", 150, L_FLATE_ENCODE, 128,
NULL, 0, 0, "/tmp/pdffile17.pdf");
/* 4 bpp input, cmap */
convertToPdfSegmented("/tmp/rabi4.png", 150, L_G4_ENCODE, 128,
NULL, 0, 0, "/tmp/pdffile18.pdf");
convertToPdfSegmented("/tmp/rabi4.png", 150, L_JPEG_ENCODE, 128,
NULL, 0, 0, "/tmp/pdffile19.pdf");
convertToPdfSegmented("/tmp/rabi4.png", 150, L_FLATE_ENCODE, 128,
NULL, 0, 0, "/tmp/pdffile20.pdf");
#endif
#if 1
/* ---------- pdf convert segmented with image regions ---------- */
fprintf(stderr, "\n*** Writing segmented images with image regions\n");
/* Get the image region(s) for rabi.png. There are two
* small bogus regions at the top, but we'll keep them for
* the demonstration. */
pix1 = pixRead("rabi.png");
pixSetResolution(pix1, 300, 300);
pixGetDimensions(pix1, &w, &h, NULL);
pix2 = pixGenHalftoneMask(pix1, NULL, NULL, 0);
pix3 = pixMorphSequence(pix2, "c20.1 + c1.20", 0);
boxa1 = pixConnComp(pix3, NULL, 8);
boxa2 = boxaTransform(boxa1, 0, 0, 0.5, 0.5);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* 1 bpp input */
convertToPdfSegmented("rabi.png", 300, L_G4_ENCODE, 128, boxa1,
0, 0.25, "/tmp/pdffile21.pdf");
convertToPdfSegmented("rabi.png", 300, L_JPEG_ENCODE, 128, boxa1,
0, 0.25, "/tmp/pdffile22.pdf");
convertToPdfSegmented("rabi.png", 300, L_FLATE_ENCODE, 128, boxa1,
0, 0.25, "/tmp/pdffile23.pdf");
/* 8 bpp input, no cmap */
convertToPdfSegmented("/tmp/rabi8.jpg", 150, L_G4_ENCODE, 128, boxa2,
0, 0.5, "/tmp/pdffile24.pdf");
convertToPdfSegmented("/tmp/rabi8.jpg", 150, L_JPEG_ENCODE, 128, boxa2,
0, 0.5, "/tmp/pdffile25.pdf");
convertToPdfSegmented("/tmp/rabi8.jpg", 150, L_FLATE_ENCODE, 128, boxa2,
0, 0.5, "/tmp/pdffile26.pdf");
/* 4 bpp input, cmap */
convertToPdfSegmented("/tmp/rabi4.png", 150, L_G4_ENCODE, 128, boxa2,
0, 0.5, "/tmp/pdffile27.pdf");
convertToPdfSegmented("/tmp/rabi4.png", 150, L_JPEG_ENCODE, 128, boxa2,
0, 0.5, "/tmp/pdffile28.pdf");
convertToPdfSegmented("/tmp/rabi4.png", 150, L_FLATE_ENCODE, 128, boxa2,
0, 0.5, "/tmp/pdffile29.pdf");
/* 4 bpp input, cmap, data output */
data = NULL;
convertToPdfDataSegmented("/tmp/rabi4.png", 150, L_G4_ENCODE, 128, boxa2,
0, 0.5, &data, &nbytes);
l_binaryWrite("/tmp/pdffile30.pdf", "w", data, nbytes);
lept_free(data);
convertToPdfDataSegmented("/tmp/rabi4.png", 150, L_JPEG_ENCODE, 128, boxa2,
0, 0.5, &data, &nbytes);
l_binaryWrite("/tmp/pdffile31.pdf", "w", data, nbytes);
lept_free(data);
convertToPdfDataSegmented("/tmp/rabi4.png", 150, L_FLATE_ENCODE, 128, boxa2,
0, 0.5, &data, &nbytes);
l_binaryWrite("/tmp/pdffile32.pdf", "w", data, nbytes);
lept_free(data);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
#endif
#if 1
/* -------- pdf convert segmented from color image -------- */
fprintf(stderr, "\n*** Writing color segmented images\n");
pix1 = pixRead("candelabrum-11.jpg");
pix2 = pixScale(pix1, 3.0, 3.0);
pixWrite("/tmp/candelabrum3.jpg", pix2, IFF_JFIF_JPEG);
GetImageMask(pix2, 200, &boxa1, "/tmp/seg1.jpg");
convertToPdfSegmented("/tmp/candelabrum3.jpg", 200, L_G4_ENCODE,
100, boxa1, 0, 0.25, "/tmp/pdffile33.pdf");
convertToPdfSegmented("/tmp/candelabrum3.jpg", 200, L_JPEG_ENCODE,
100, boxa1, 0, 0.25, "/tmp/pdffile34.pdf");
convertToPdfSegmented("/tmp/candelabrum3.jpg", 200, L_FLATE_ENCODE,
100, boxa1, 0, 0.25, "/tmp/pdffile35.pdf");
pixDestroy(&pix1);
pixDestroy(&pix2);
boxaDestroy(&boxa1);
pix1 = pixRead("lion-page.00016.jpg");
pix2 = pixScale(pix1, 3.0, 3.0);
pixWrite("/tmp/lion16.jpg", pix2, IFF_JFIF_JPEG);
pix3 = pixRead("lion-mask.00016.tif");
boxa1 = pixConnComp(pix3, NULL, 8);
boxa2 = boxaTransform(boxa1, 0, 0, 3.0, 3.0);
convertToPdfSegmented("/tmp/lion16.jpg", 200, L_G4_ENCODE,
190, boxa2, 0, 0.5, "/tmp/pdffile36.pdf");
convertToPdfSegmented("/tmp/lion16.jpg", 200, L_JPEG_ENCODE,
190, boxa2, 0, 0.5, "/tmp/pdffile37.pdf");
convertToPdfSegmented("/tmp/lion16.jpg", 200, L_FLATE_ENCODE,
190, boxa2, 0, 0.5, "/tmp/pdffile38.pdf");
/* Quantize the non-image part and flate encode.
* This is useful because it results in a smaller file than
* when you flate-encode the un-quantized non-image regions. */
pix4 = pixScale(pix3, 3.0, 3.0); /* higher res mask, for combining */
pix5 = QuantizeNonImageRegion(pix2, pix4, 12);
pixWrite("/tmp/lion16-quant.png", pix5, IFF_PNG);
convertToPdfSegmented("/tmp/lion16-quant.png", 200, L_FLATE_ENCODE,
190, boxa2, 0, 0.5, "/tmp/pdffile39.pdf");
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
pixDestroy(&pix5);
boxaDestroy(&boxa1);
boxaDestroy(&boxa2);
#endif
#if 1
/* ------------------ Test multipage pdf generation ----------------- */
fprintf(stderr, "\n*** Writing multipage pdfs from single page pdfs\n");
/* Generate a multi-page pdf from all these files */
startTimer();
concatenatePdf("/tmp", "pdffile", "/tmp/cat_lept.pdf");
fprintf(stderr, "Time: %7.3f\n", stopTimer());
/* Put two good pdf files in a directory */
lept_mkdir("good");
lept_cp("testfile1.pdf", "/tmp/good");
lept_cp("testfile2.pdf", "/tmp/good");
concatenatePdf("/tmp/good", "file", "/tmp/good.pdf");
/* Make a version with the pdf id removed, so that it is not
* recognized as a pdf */
ba = l_byteaInitFromFile("testfile2.pdf");
data = l_byteaGetData(ba, &nbytes);
l_binaryWrite("testfile0.notpdf.pdf", "w", data + 10, nbytes - 10);
/* Make a version with a corrupted trailer */
data[2297] = '2'; /* munge trailer object 6: change 458 --> 428 */
l_binaryWrite("testfile2.bad.pdf", "w", data, nbytes);
/* Put these two bad files, along with a good file, in a directory */
lept_mkdir("bad");
lept_mv("testfile0.notpdf.pdf", "/tmp/bad");
lept_cp("testfile1.pdf", "/tmp/bad");
lept_mv("testfile2.bad.pdf", "/tmp/bad");
l_byteaDestroy(&ba);
/* Run concat on the bad files. In the /tmp/bad/ directory,
* the "not pdf" file should be ignored, and the corrupted pdf
* file should be properly parsed, so the resulting
* concatenated files should be identical. */
fprintf(stderr, "\nWe attempt to build from the bad directory\n");
concatenatePdf("/tmp/bad", "file", "/tmp/bad.pdf");
filesAreIdentical("/tmp/good.pdf", "/tmp/bad.pdf", &same);
if (same)
fprintf(stderr, "Fixed: files are the same\n"
"Attempt succeeded\n\n");
else
fprintf(stderr, "Busted: files are different\n");
/* pdftk fails because the first file is not a pdf */
fprintf(stderr, "pdftk attempts to build from the bad directory\n");
tempfile1 = genPathname("/tmp/bad", "*.pdf");
tempfile2 = genPathname("/tmp", "pdftk.bad.pdf");
snprintf(buffer, sizeof(buffer), "pdftk %s output %s",
tempfile1, tempfile2);
ret = system(buffer);
lept_free(tempfile1);
lept_free(tempfile2);
fprintf(stderr, "Attempt failed\n\n");
#endif
#if 1
fprintf(stderr, "\n*** pdftk writes multipage pdfs from images\n");
tempfile1 = genPathname("/tmp", "pdffile*.pdf");
tempfile2 = genPathname("/tmp", "cat_pdftk.pdf");
snprintf(buffer, sizeof(buffer), "pdftk %s output %s",
tempfile1, tempfile2);
ret = system(buffer);
lept_free(tempfile1);
lept_free(tempfile2);
#endif
#if 1
/* -- Test simple interface for generating multi-page pdf from images -- */
fprintf(stderr, "\n*** Writing multipage pdfs from images\n");
/* Put four image files in a directory. They will be encoded thus:
* file1.png: flate (8 bpp, only 10 colors)
* file2.jpg: dct (8 bpp, 256 colors because of the jpeg encoding)
* file3.tif: g4 (1 bpp)
* file4.jpg: dct (32 bpp) */
lept_mkdir("image");
pix1 = pixRead("feyn.tif");
pix2 = pixRead("rabi.png");
pix3 = pixScaleToGray3(pix1);
pix4 = pixScaleToGray3(pix2);
pix5 = pixScale(pix1, 0.33, 0.33);
pix6 = pixRead("test24.jpg");
pixWrite("/tmp/image/file1.png", pix3, IFF_PNG); /* 10 colors */
pixWrite("/tmp/image/file2.jpg", pix4, IFF_JFIF_JPEG); /* 256 colors */
pixWrite("/tmp/image/file3.tif", pix5, IFF_TIFF_G4);
pixWrite("/tmp/image/file4.jpg", pix6, IFF_JFIF_JPEG);
startTimer();
convertFilesToPdf("/tmp/image", "file", 100, 0.8, 75, "4 file test",
"/tmp/fourimages.pdf");
fprintf(stderr, "Time: %7.3f\n", stopTimer());
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
pixDestroy(&pix5);
pixDestroy(&pix6);
#endif
return 0;
}
// Finds image regions within the source pix (page image) and returns
// the image regions as a Boxa, Pixa pair, analgous to pixConnComp.
// The returned boxa, pixa may be NULL, meaning no images found.
// If not NULL, they must be destroyed by the caller.
void ImageFinder::FindImages(Pix* pix, Boxa** boxa, Pixa** pixa) {
*boxa = NULL;
*pixa = NULL;
#ifdef HAVE_LIBLEPT
if (pixGetWidth(pix) < kMinImageFindSize ||
pixGetHeight(pix) < kMinImageFindSize)
return; // Not worth looking at small images.
// Reduce by factor 2.
Pix *pixr = pixReduceRankBinaryCascade(pix, 1, 0, 0, 0);
pixDisplayWrite(pixr, textord_tabfind_show_images);
// Get the halftone mask directly from Leptonica.
Pix *pixht2 = pixGenHalftoneMask(pixr, NULL, NULL,
textord_tabfind_show_images);
pixDestroy(&pixr);
if (pixht2 == NULL)
return;
// Expand back up again.
Pix *pixht = pixExpandReplicate(pixht2, 2);
pixDisplayWrite(pixht, textord_tabfind_show_images);
pixDestroy(&pixht2);
// Fill to capture pixels near the mask edges that were missed
Pix *pixt = pixSeedfillBinary(NULL, pixht, pix, 8);
pixOr(pixht, pixht, pixt);
pixDestroy(&pixt);
// Eliminate lines and bars that may be joined to images.
Pix* pixfinemask = pixReduceRankBinaryCascade(pixht, 1, 1, 3, 3);
pixDilateBrick(pixfinemask, pixfinemask, 5, 5);
pixDisplayWrite(pixfinemask, textord_tabfind_show_images);
Pix* pixreduced = pixReduceRankBinaryCascade(pixht, 1, 1, 1, 1);
Pix* pixreduced2 = pixReduceRankBinaryCascade(pixreduced, 3, 3, 3, 0);
pixDestroy(&pixreduced);
pixDilateBrick(pixreduced2, pixreduced2, 5, 5);
Pix* pixcoarsemask = pixExpandReplicate(pixreduced2, 8);
pixDestroy(&pixreduced2);
pixDisplayWrite(pixcoarsemask, textord_tabfind_show_images);
// Combine the coarse and fine image masks.
pixAnd(pixcoarsemask, pixcoarsemask, pixfinemask);
pixDestroy(&pixfinemask);
// Dilate a bit to make sure we get everything.
pixDilateBrick(pixcoarsemask, pixcoarsemask, 3, 3);
Pix* pixmask = pixExpandReplicate(pixcoarsemask, 16);
pixDestroy(&pixcoarsemask);
pixDisplayWrite(pixmask, textord_tabfind_show_images);
// And the image mask with the line and bar remover.
pixAnd(pixht, pixht, pixmask);
pixDestroy(&pixmask);
pixDisplayWrite(pixht, textord_tabfind_show_images);
// Find the individual image regions in the mask image.
*boxa = pixConnComp(pixht, pixa, 8);
pixDestroy(&pixht);
// Rectangularize the individual images. If a sharp edge in vertical and/or
// horizontal occupancy can be found, it indicates a probably rectangular
// image with unwanted bits merged on, so clip to the approximate rectangle.
int npixes = pixaGetCount(*pixa);
for (int i = 0; i < npixes; ++i) {
int x_start, x_end, y_start, y_end;
Pix* img_pix = pixaGetPix(*pixa, i, L_CLONE);
pixDisplayWrite(img_pix, textord_tabfind_show_images);
if (pixNearlyRectangular(img_pix, kMinRectangularFraction,
kMaxRectangularFraction,
kMaxRectangularGradient,
&x_start, &y_start, &x_end, &y_end)) {
// Add 1 to the size as a kludgy flag to indicate to the later stages
// of processing that it is a clipped rectangular image .
Pix* simple_pix = pixCreate(pixGetWidth(img_pix) + 1,
pixGetHeight(img_pix), 1);
pixDestroy(&img_pix);
pixRasterop(simple_pix, x_start, y_start, x_end - x_start,
y_end - y_start, PIX_SET, NULL, 0, 0);
// pixaReplacePix takes ownership of the simple_pix.
pixaReplacePix(*pixa, i, simple_pix, NULL);
img_pix = pixaGetPix(*pixa, i, L_CLONE);
}
// Subtract the pix from the correct location in the master image.
l_int32 x, y, width, height;
pixDisplayWrite(img_pix, textord_tabfind_show_images);
boxaGetBoxGeometry(*boxa, i, &x, &y, &width, &height);
pixRasterop(pix, x, y, width, height, PIX_NOT(PIX_SRC) & PIX_DST,
img_pix, 0, 0);
pixDestroy(&img_pix);
}
#endif
}
