void
AddTestSet(PIXA *pixa,
PIX *pixs,
l_int32 filtertype,
l_int32 edgethresh,
l_int32 smoothx,
l_int32 smoothy,
l_float32 gamma,
l_int32 minval,
l_int32 maxval,
l_int32 targetthresh)
{
PIX *pixb, *pixd, *pixth;
pixThresholdSpreadNorm(pixs, filtertype, edgethresh,
smoothx, smoothy, gamma, minval,
maxval, targetthresh, &pixth, NULL, &pixd);
pixSaveTiled(pixth, pixa, 1.0, 1, 20, 0);
pixSaveTiled(pixd, pixa, 1.0, 0, 20, 0);
pixb = pixThresholdToBinary(pixd, targetthresh - 20);
pixSaveTiled(pixb, pixa, 1.0, 0, 20, 0);
pixDestroy(&pixb);
pixb = pixThresholdToBinary(pixd, targetthresh);
pixSaveTiled(pixb, pixa, 1.0, 0, 20, 0);
pixDestroy(&pixb);
pixb = pixThresholdToBinary(pixd, targetthresh + 20);
pixSaveTiled(pixb, pixa, 1.0, 0, 20, 0);
pixDestroy(&pixb);
pixb = pixThresholdToBinary(pixd, targetthresh + 40);
pixSaveTiled(pixb, pixa, 1.0, 0, 20, 0);
pixDestroy(&pixb);
pixDestroy(&pixth);
pixDestroy(&pixd);
return;
}
main(int argc,
char **argv)
{
l_int32 i, w, h, d;
l_float32 time;
PIX *pixs, *pixf, *pixd;
PIXA *pixa;
char *filein, *fileout;
static char mainName[] = "edgetest";
if (argc != 3)
exit(ERROR_INT(" Syntax: edgetest filein fileout", mainName, 1));
filein = argv[1];
fileout = argv[2];
if ((pixs = pixRead(filein)) == NULL)
exit(ERROR_INT("pix not made", mainName, 1));
pixGetDimensions(pixs, &w, &h, &d);
if (d != 8)
exit(ERROR_INT("pix not 8 bpp", mainName, 1));
/* Speed: about 12 Mpix/GHz/sec */
startTimer();
pixf = pixSobelEdgeFilter(pixs, L_HORIZONTAL_EDGES);
pixd = pixThresholdToBinary(pixf, 60);
pixInvert(pixd, pixd);
time = stopTimer();
fprintf(stderr, "Time = %7.3f sec\n", time);
fprintf(stderr, "MPix/sec: %7.3f\n", 0.000001 * w * h / time);
pixDisplay(pixs, 0, 0);
pixInvert(pixf, pixf);
pixDisplay(pixf, 480, 0);
pixDisplay(pixd, 960, 0);
pixWrite(fileout, pixf, IFF_PNG);
pixDestroy(&pixd);
/* Threshold at different values */
pixInvert(pixf, pixf);
for (i = 10; i <= 120; i += 10) {
pixd = pixThresholdToBinary(pixf, i);
pixInvert(pixd, pixd);
pixDisplayWrite(pixd, 1);
pixDestroy(&pixd);
}
pixDestroy(&pixf);
/* Display tiled */
pixa = pixaReadFiles("/tmp", "junk_write_display");
pixd = pixaDisplayTiledAndScaled(pixa, 8, 400, 3, 0, 25, 2);
pixWrite("/tmp/junktiles.jpg", pixd, IFF_JFIF_JPEG);
pixDestroy(&pixd);
pixaDestroy(&pixa);
pixDestroy(&pixs);
exit(0);
}
l_int32 main(int argc,
char **argv)
{
l_int32 pageno;
L_DEWARP *dew1;
L_DEWARPA *dewa;
PIX *pixs, *pixn, *pixg, *pixb;
static char mainName[] = "dewarptest2";
if (argc != 1 && argc != 3)
return ERROR_INT("Syntax: dewarptest2 [image pageno]", mainName, 1);
if (argc == 1) {
pixs = pixRead("cat-35.jpg");
pageno = 35;
}
else {
pixs = pixRead(argv[1]);
pageno = atoi(argv[2]);
}
if (!pixs)
return ERROR_INT("image not read", mainName, 1);
dewa = dewarpaCreate(40, 30, 1, 6, 50);
#if NORMALIZE
/* Normalize for varying background and binarize */
pixn = pixBackgroundNormSimple(pixs, NULL, NULL);
pixg = pixConvertRGBToGray(pixn, 0.5, 0.3, 0.2);
pixb = pixThresholdToBinary(pixg, 130);
pixDestroy(&pixn);
#else
/* Don't normalize; just threshold and clean edges */
pixg = pixConvertTo8(pixs, 0);
pixb = pixThresholdToBinary(pixg, 100);
pixSetOrClearBorder(pixb, 30, 30, 40, 40, PIX_CLR);
#endif
/* Run the basic functions */
dew1 = dewarpCreate(pixb, pageno);
dewarpaInsertDewarp(dewa, dew1);
dewarpBuildModel(dew1, "/tmp/dewarp_model1.pdf");
dewarpaApplyDisparity(dewa, pageno, pixg, "/tmp/dewarp_apply1.pdf");
dewarpaDestroy(&dewa);
pixDestroy(&pixs);
pixDestroy(&pixg);
pixDestroy(&pixb);
return 0;
}
jlong Java_com_googlecode_leptonica_android_GrayQuant_nativePixThresholdToBinary(JNIEnv *env, jclass clazz,
jlong nativePix, jint thresh) {
PIX *pixs = (PIX *) nativePix;
PIX *pixd = pixThresholdToBinary(pixs, (l_int32) thresh);
return (jlong) pixd;
}
/* load PNG, TIFF, JPG, GIF or BMP to PIX datastructure. The actual supported
* formats depends on how the leptonica was compiled */
PIX *loadimage(char *filename){
PIX *pix, *pixt;
int format, bpp;
format=fileformat(filename);
// In later versions of leptonica you will have to do this
// pixReadHeader(filename, format,NULL,NULL,NULL,bpp,NULL);
if(format!=IFF_PNG && format!=IFF_JFIF_JPEG && format!=IFF_TIFF && format!=
IFF_GIF && format!=7 && format!=8){
dfprintf(stderr,"Not recognised file format %i", format);
return NULL;
}
if ((pix = pixRead(filename)) == NULL) return NULL;
/* TODO: convert image to 1-bpp 300dpi regardless of scan */
bpp=pixGetDepth(pix);
if(bpp>1){
/*
printf("Bits per pixel=%i",bpp);
exit(1); */
//pixThresholdForFgBg(pix,5,100,NULL,NULL);
//pixContrastTRC(pix, pix, 1000);
pixt = pixContrastNorm(NULL, pix, 10, 10, 40, 2, 2);
pixDestroy(&pix);
pix = pixGammaTRC(NULL, pixt, 1.5, 50, 235);
pixt=pixThresholdToBinary(pix, 200);
//pixt=pixThreshold8(pix,1,1,0);
pixDestroy(&pix);
pix=pixt;
}
return pix;
}
/*!
* pixRotateBinaryNice()
*
* Input: pixs (1 bpp)
* angle (radians; clockwise is positive; about the center)
* incolor (L_BRING_IN_WHITE, L_BRING_IN_BLACK)
* Return: pixd, or null on error
*
* Notes:
* (1) For very small rotations, just return a clone.
* (2) This does a computationally expensive rotation of 1 bpp images.
* The fastest rotators (using shears or subsampling) leave
* visible horizontal and vertical shear lines across which
* the image shear changes by one pixel. To ameliorate the
* visual effect one can introduce random dithering. One
* way to do this in a not-too-random fashion is given here.
* We convert to 8 bpp, do a very small blur, rotate using
* linear interpolation (same as area mapping), do a
* small amount of sharpening to compensate for the initial
* blur, and threshold back to binary. The shear lines
* are magically removed.
* (3) This operation is about 5x slower than rotation by sampling.
*/
PIX *
pixRotateBinaryNice(PIX *pixs,
l_float32 angle,
l_int32 incolor)
{
PIX *pixt1, *pixt2, *pixt3, *pixt4, *pixd;
PROCNAME("pixRotateBinaryNice");
if (!pixs || pixGetDepth(pixs) != 1)
return (PIX *)ERROR_PTR("pixs undefined or not 1 bpp", procName, NULL);
if (incolor != L_BRING_IN_WHITE && incolor != L_BRING_IN_BLACK)
return (PIX *)ERROR_PTR("invalid incolor", procName, NULL);
pixt1 = pixConvertTo8(pixs, 0);
pixt2 = pixBlockconv(pixt1, 1, 1); /* smallest blur allowed */
pixt3 = pixRotateAM(pixt2, angle, incolor);
pixt4 = pixUnsharpMasking(pixt3, 1, 1.0); /* sharpen a bit */
pixd = pixThresholdToBinary(pixt4, 128);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
pixDestroy(&pixt4);
return pixd;
}
main(int argc,
char **argv)
{
char *infile, *outfile;
l_int32 d;
PIX *pixs, *pixgr, *pixb;
static char mainName[] = "showedges";
if (argc != 3)
exit(ERROR_INT(" Syntax: showedges infile outfile", mainName, 1));
infile = argv[1];
outfile = argv[2];
pixs = pixRead(infile);
d = pixGetDepth(pixs);
if (d != 8 && d != 32)
exit(ERROR_INT("d not 8 or 32 bpp", mainName, 1));
pixgr = pixHalfEdgeByBandpass(pixs, SMOOTH_WIDTH_1, SMOOTH_WIDTH_1,
SMOOTH_WIDTH_2, SMOOTH_WIDTH_2);
pixb = pixThresholdToBinary(pixgr, THRESHOLD);
pixInvert(pixb, pixb);
/* pixWrite("junkpixgr", pixgr, IFF_JFIF_JPEG); */
pixWrite(outfile, pixb, IFF_PNG);
return 0;
}
/*!
* dewarpSinglePageInit()
*
* Input: pixs (with text, any depth)
* thresh (for global thresholding to 1 bpp; ignored otherwise)
* adaptive (1 for adaptive thresholding; 0 for global threshold)
* use_both (1 for horizontal and vertical; 0 for vertical only)
* &pixb (<return> 1 bpp image)
* &dewa (<return> initialized dewa)
* Return: 0 if OK, 1 on error (list of page numbers), or null on error
*
* Notes:
* (1) This binarizes the input pixs if necessary, returning the
* binarized image. It also initializes the dewa to default values
* for the model parameters.
* (2) If pixs is 1 bpp, the parameters @adaptive and @thresh are ignored.
* (3) To change the model parameters, call dewarpaSetCurvatures()
* before running dewarpSinglePageRun(). For DC:
* dewarpSinglePageInit(pixs, 0, 1, 1, &pixb, &dewa);
* dewarpaSetCurvatures(dewa, 250, -1, -1, 80, 70, 150);
* dewarpSinglePageRun(pixs, pixb, dewa, &pixd, 0);
* dewarpaDestroy(&dewa);
* pixDestroy(&pixb);
*/
l_int32
dewarpSinglePageInit(PIX *pixs,
l_int32 thresh,
l_int32 adaptive,
l_int32 use_both,
PIX **ppixb,
L_DEWARPA **pdewa)
{
PIX *pix1;
PROCNAME("dewarpSinglePageInit");
if (ppixb) *ppixb = NULL;
if (pdewa) *pdewa = NULL;
if (!ppixb || !pdewa)
return ERROR_INT("&pixb and &dewa not both defined", procName, 1);
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
*pdewa = dewarpaCreate(1, 0, 1, 0, -1);
dewarpaUseBothArrays(*pdewa, use_both);
/* Generate a binary image, if necessary */
if (pixGetDepth(pixs) > 1) {
pix1 = pixConvertTo8(pixs, 0);
if (adaptive)
*ppixb = pixAdaptThresholdToBinary(pix1, NULL, 1.0);
else
*ppixb = pixThresholdToBinary(pix1, thresh);
pixDestroy(&pix1);
} else {
*ppixb = pixClone(pixs);
}
return 0;
}
// threshold the image -- warning, operates ONLY on grayscale pixmaps
// original image is unchanged - a new modified image at raster's end
/////////////////////////////////////////////////////////////////////////////////////////////////////////////
Pix Pix::ThresholdToBinary(int threshold)
{
if(IsEmpty())
return Pix();
PIX *dPix = pixThresholdToBinary(pix, threshold);
if(!dPix)
return Pix();
return Pix(&dPix);
} // END Pix::ThresholdToBinary()
/*
* Binarize and fix uneven illumination
*/
void MainWindow::on_actionBinarizeUnIl_triggered() {
QApplication::setOverrideCursor(Qt::WaitCursor);
PIX *pixc, *pixg, *pixsg, *pixd;
/* Convert the RGB image to grayscale. */
this->statusBar()->showMessage(tr("Convert the RGB image to grayscale."));
pixsg = pixConvertRGBToLuminance(pixs);
setPixToScene(pixsg);
/* Remove the text in the fg. */
this->statusBar()->showMessage(tr("Remove the text in the fg."));
pixc = pixCloseGray(pixsg, 25, 25);
setPixToScene(pixc);
/* Smooth the bg with a convolution. */
// pixsm = pixBlockconv(pixc, 15, 15);
// pixDestroy(&pixsm);
/* Normalize for uneven illumination on gray image. */
this->statusBar()->showMessage(tr("Normalize for uneven illumination on gray image."));
pixBackgroundNormGrayArrayMorph(pixsg, NULL, 4, 5, 200, &pixg);
pixc = pixApplyInvBackgroundGrayMap(pixsg, pixg, 4, 4);
pixDestroy(&pixsg);
pixDestroy(&pixg);
setPixToScene(pixc);
/* Increase the dynamic range. */
// make dark gray *black* and light gray *white*
this->statusBar()->showMessage(tr("Increase the dynamic range."));
pixd = pixGammaTRC(NULL, pixc, 1.0, 50, 220);
setPixToScene(pixd);
/* Threshold to 1 bpp. */
this->statusBar()->showMessage(tr("Threshold to 1 bpp."));
pixs = pixThresholdToBinary(pixd, 120);
pixDestroy(&pixd);
setPixToScene();
this->statusBar()->showMessage(tr("Finished..."), 2000);
modified = true;
updateTitle();
QApplication::restoreOverrideCursor();
}
jint Java_com_googlecode_leptonica_android_Dewarp_nativeDewarp(JNIEnv *env, jclass clazz, jint nativePix) {
LOGV("%s",__FUNCTION__);
PIX* pixd=NULL;
PIX *pixs = (PIX *) nativePix;
PIX* pixb = pixThresholdToBinary(pixs, 130);
// Basic functioning:
L_Dewarp *dew = dewarpCreate(pixb,0,60,5,1);
int buildresult=dewarpBuildModel(dew, 0);
if(buildresult==0){
int applyresult=dewarpApplyDisparity(dew, pixs, 0);
if(applyresult==0){
pixd = pixClone(dew->pixd);
}
}
pixDestroy(&pixb);
dewarpDestroy(&dew);
if(pixd!=NULL){
return (jint) pixd;
} else {
return (jint) 0;
}
}
char* _process_frame_tickertext(struct lib_hardsubx_ctx *ctx, AVFrame *frame, int width, int height, int index)
{
PIX *im;
PIX *edge_im;
PIX *lum_im;
PIX *feat_im;
char *subtitle_text=NULL;
im = pixCreate(width,height,32);
lum_im = pixCreate(width,height,32);
feat_im = pixCreate(width,height,32);
int i,j;
for(i=(92*height)/100;i<height;i++)
{
for(j=0;j<width;j++)
{
int p=j*3+i*frame->linesize[0];
int r=frame->data[0][p];
int g=frame->data[0][p+1];
int b=frame->data[0][p+2];
pixSetRGBPixel(im,j,i,r,g,b);
float L,A,B;
rgb_to_lab((float)r,(float)g,(float)b,&L,&A,&B);
if(L > ctx->lum_thresh)
pixSetRGBPixel(lum_im,j,i,255,255,255);
else
pixSetRGBPixel(lum_im,j,i,0,0,0);
}
}
//Handle the edge image
edge_im = pixCreate(width,height,8);
edge_im = pixConvertRGBToGray(im,0.0,0.0,0.0);
edge_im = pixSobelEdgeFilter(edge_im, L_VERTICAL_EDGES);
edge_im = pixDilateGray(edge_im, 21, 11);
edge_im = pixThresholdToBinary(edge_im,50);
for(i=92*(height/100);i<height;i++)
{
for(j=0;j<width;j++)
{
unsigned int p1,p2,p3;
pixGetPixel(edge_im,j,i,&p1);
// pixGetPixel(pixd,j,i,&p2);
pixGetPixel(lum_im,j,i,&p3);
if(p1==0&&p3>0)
pixSetRGBPixel(feat_im,j,i,255,255,255);
else
pixSetRGBPixel(feat_im,j,i,0,0,0);
}
}
// Tesseract OCR for the ticker text here
subtitle_text = get_ocr_text_simple(ctx, lum_im);
char write_path[100];
sprintf(write_path,"./lum_im%04d.jpg",index);
pixWrite(write_path,lum_im,IFF_JFIF_JPEG);
sprintf(write_path,"./im%04d.jpg",index);
pixWrite(write_path,im,IFF_JFIF_JPEG);
pixDestroy(&lum_im);
pixDestroy(&im);
pixDestroy(&edge_im);
pixDestroy(&feat_im);
return subtitle_text;
}
void _display_frame(struct lib_hardsubx_ctx *ctx, AVFrame *frame, int width, int height, int timestamp)
{
// Debug: Display the frame after processing
PIX *im;
im = pixCreate(width,height,32);
PIX *hue_im = pixCreate(width,height,32);
int i,j;
for(i=0;i<height;i++)
{
for(j=0;j<width;j++)
{
int p=j*3+i*frame->linesize[0];
int r=frame->data[0][p];
int g=frame->data[0][p+1];
int b=frame->data[0][p+2];
pixSetRGBPixel(im,j,i,r,g,b);
float H,S,V;
rgb_to_hsv((float)r,(float)g,(float)b,&H,&S,&V);
if(abs(H-ctx->hue)<20)
{
pixSetRGBPixel(hue_im,j,i,r,g,b);
}
}
}
PIX *edge_im = pixCreate(width,height,8),*edge_im_2 = pixCreate(width,height,8);
edge_im = pixConvertRGBToGray(im,0.0,0.0,0.0);
edge_im = pixSobelEdgeFilter(edge_im, L_VERTICAL_EDGES);
edge_im = pixDilateGray(edge_im, 21, 1);
edge_im = pixThresholdToBinary(edge_im,50);
PIX *pixd = pixCreate(width,height,1);
pixSauvolaBinarize(pixConvertRGBToGray(hue_im,0.0,0.0,0.0), 15, 0.3, 1, NULL, NULL, NULL, &pixd);
edge_im_2 = pixConvertRGBToGray(hue_im,0.0,0.0,0.0);
edge_im_2 = pixDilateGray(edge_im_2, 5, 5);
PIX *feat_im = pixCreate(width,height,32);
for(i=3*(height/4);i<height;i++)
{
for(j=0;j<width;j++)
{
unsigned int p1,p2,p3,p4;
pixGetPixel(edge_im,j,i,&p1);
pixGetPixel(pixd,j,i,&p2);
// pixGetPixel(hue_im,j,i,&p3);
pixGetPixel(edge_im_2,j,i,&p4);
if(p1==0&&p2==0&&p4>0)//if(p4>0&&p1==0)//if(p2==0&&p1==0&&p3>0)
{
pixSetRGBPixel(feat_im,j,i,255,255,255);
}
}
}
char *txt=NULL;
// txt = get_ocr_text_simple(ctx, feat_im);
// txt=get_ocr_text_wordwise_threshold(ctx, feat_im, ctx->conf_thresh);
// if(txt != NULL)printf("%s\n", txt);
pixDestroy(&im);
pixDestroy(&edge_im);
pixDestroy(&feat_im);
pixDestroy(&edge_im_2);
pixDestroy(&pixd);
}
char* _process_frame_white_basic(struct lib_hardsubx_ctx *ctx, AVFrame *frame, int width, int height, int index)
{
//printf("frame : %04d\n", index);
PIX *im;
PIX *edge_im;
PIX *lum_im;
PIX *feat_im;
char *subtitle_text=NULL;
im = pixCreate(width,height,32);
lum_im = pixCreate(width,height,32);
feat_im = pixCreate(width,height,32);
int i,j;
for(i=(3*height)/4;i<height;i++)
{
for(j=0;j<width;j++)
{
int p=j*3+i*frame->linesize[0];
int r=frame->data[0][p];
int g=frame->data[0][p+1];
int b=frame->data[0][p+2];
pixSetRGBPixel(im,j,i,r,g,b);
float L,A,B;
rgb_to_lab((float)r,(float)g,(float)b,&L,&A,&B);
if(L > ctx->lum_thresh)
pixSetRGBPixel(lum_im,j,i,255,255,255);
else
pixSetRGBPixel(lum_im,j,i,0,0,0);
}
}
//Handle the edge image
edge_im = pixCreate(width,height,8);
edge_im = pixConvertRGBToGray(im,0.0,0.0,0.0);
edge_im = pixSobelEdgeFilter(edge_im, L_VERTICAL_EDGES);
edge_im = pixDilateGray(edge_im, 21, 11);
edge_im = pixThresholdToBinary(edge_im,50);
for(i=3*(height/4);i<height;i++)
{
for(j=0;j<width;j++)
{
unsigned int p1,p2,p3;
pixGetPixel(edge_im,j,i,&p1);
// pixGetPixel(pixd,j,i,&p2);
pixGetPixel(lum_im,j,i,&p3);
if(p1==0&&p3>0)
pixSetRGBPixel(feat_im,j,i,255,255,255);
else
pixSetRGBPixel(feat_im,j,i,0,0,0);
}
}
if(ctx->detect_italics)
{
ctx->ocr_mode = HARDSUBX_OCRMODE_WORD;
}
// TESSERACT OCR FOR THE FRAME HERE
switch(ctx->ocr_mode)
{
case HARDSUBX_OCRMODE_WORD:
if(ctx->conf_thresh > 0)
subtitle_text = get_ocr_text_wordwise_threshold(ctx, lum_im, ctx->conf_thresh);
else
subtitle_text = get_ocr_text_wordwise(ctx, lum_im);
break;
case HARDSUBX_OCRMODE_LETTER:
if(ctx->conf_thresh > 0)
subtitle_text = get_ocr_text_letterwise_threshold(ctx, lum_im, ctx->conf_thresh);
else
subtitle_text = get_ocr_text_letterwise(ctx, lum_im);
break;
case HARDSUBX_OCRMODE_FRAME:
if(ctx->conf_thresh > 0)
subtitle_text = get_ocr_text_simple_threshold(ctx, lum_im, ctx->conf_thresh);
else
subtitle_text = get_ocr_text_simple(ctx, lum_im);
break;
default:
fatal(EXIT_MALFORMED_PARAMETER,"Invalid OCR Mode");
}
pixDestroy(&lum_im);
pixDestroy(&im);
pixDestroy(&edge_im);
pixDestroy(&feat_im);
return subtitle_text;
}
l_int32 main(int argc,
char **argv) {
l_int32 i, n;
l_float32 a, b, c, d, e;
NUMA *nax, *nafit;
PIX *pixs, *pixn, *pixg, *pixb, *pixt1, *pixt2;
PIXA *pixa;
PTA *pta, *ptad;
PTAA *ptaa1, *ptaa2;
pixs = pixRead("cat-35.jpg");
/* pixs = pixRead("zanotti-78.jpg"); */
/* Normalize for varying background and binarize */
pixn = pixBackgroundNormSimple(pixs, NULL, NULL);
pixg = pixConvertRGBToGray(pixn, 0.5, 0.3, 0.2);
pixb = pixThresholdToBinary(pixg, 130);
pixDestroy(&pixn);
pixDestroy(&pixg);
/* Get the textline centers */
pixa = pixaCreate(6);
ptaa1 = dewarpGetTextlineCenters(pixb, 0);
pixt1 = pixCreateTemplate(pixs);
pixSetAll(pixt1);
pixt2 = pixDisplayPtaa(pixt1, ptaa1);
pixWrite("/tmp/textline1.png", pixt2, IFF_PNG);
pixDisplayWithTitle(pixt2, 0, 100, "textline centers 1", 1);
pixaAddPix(pixa, pixt2, L_INSERT);
pixDestroy(&pixt1);
/* Remove short lines */
fprintf(stderr, "Num all lines = %d\n", ptaaGetCount(ptaa1));
ptaa2 = dewarpRemoveShortLines(pixb, ptaa1, 0.8, 0);
pixt1 = pixCreateTemplate(pixs);
pixSetAll(pixt1);
pixt2 = pixDisplayPtaa(pixt1, ptaa2);
pixWrite("/tmp/textline2.png", pixt2, IFF_PNG);
pixDisplayWithTitle(pixt2, 300, 100, "textline centers 2", 1);
pixaAddPix(pixa, pixt2, L_INSERT);
pixDestroy(&pixt1);
n = ptaaGetCount(ptaa2);
fprintf(stderr, "Num long lines = %d\n", n);
ptaaDestroy(&ptaa1);
pixDestroy(&pixb);
/* Long lines over input image */
pixt1 = pixCopy(NULL, pixs);
pixt2 = pixDisplayPtaa(pixt1, ptaa2);
pixWrite("/tmp/textline3.png", pixt2, IFF_PNG);
pixDisplayWithTitle(pixt2, 600, 100, "textline centers 3", 1);
pixaAddPix(pixa, pixt2, L_INSERT);
pixDestroy(&pixt1);
/* Quadratic fit to curve */
pixt1 = pixCopy(NULL, pixs);
for (i = 0; i < n; i++) {
pta = ptaaGetPta(ptaa2, i, L_CLONE);
ptaGetArrays(pta, &nax, NULL);
ptaGetQuadraticLSF(pta, &a, &b, &c, &nafit);
fprintf(stderr, "Quadratic: a = %10.6f, b = %7.3f, c = %7.3f\n",
a, b, c);
ptad = ptaCreateFromNuma(nax, nafit);
pixDisplayPta(pixt1, pixt1, ptad);
ptaDestroy(&pta);
ptaDestroy(&ptad);
numaDestroy(&nax);
numaDestroy(&nafit);
}
pixWrite("/tmp/textline4.png", pixt1, IFF_PNG);
pixDisplayWithTitle(pixt1, 900, 100, "textline centers 4", 1);
pixaAddPix(pixa, pixt1, L_INSERT);
/* Cubic fit to curve */
pixt1 = pixCopy(NULL, pixs);
for (i = 0; i < n; i++) {
pta = ptaaGetPta(ptaa2, i, L_CLONE);
ptaGetArrays(pta, &nax, NULL);
ptaGetCubicLSF(pta, &a, &b, &c, &d, &nafit);
fprintf(stderr, "Cubic: a = %10.6f, b = %10.6f, c = %7.3f, d = %7.3f\n",
a, b, c, d);
ptad = ptaCreateFromNuma(nax, nafit);
pixDisplayPta(pixt1, pixt1, ptad);
ptaDestroy(&pta);
ptaDestroy(&ptad);
numaDestroy(&nax);
numaDestroy(&nafit);
}
pixWrite("/tmp/textline5.png", pixt1, IFF_PNG);
pixDisplayWithTitle(pixt1, 1200, 100, "textline centers 5", 1);
pixaAddPix(pixa, pixt1, L_INSERT);
/* Quartic fit to curve */
pixt1 = pixCopy(NULL, pixs);
for (i = 0; i < n; i++) {
pta = ptaaGetPta(ptaa2, i, L_CLONE);
ptaGetArrays(pta, &nax, NULL);
ptaGetQuarticLSF(pta, &a, &b, &c, &d, &e, &nafit);
fprintf(stderr,
"Quartic: a = %7.3f, b = %7.3f, c = %9.5f, d = %7.3f, e = %7.3f\n",
a, b, c, d, e);
ptad = ptaCreateFromNuma(nax, nafit);
pixDisplayPta(pixt1, pixt1, ptad);
ptaDestroy(&pta);
ptaDestroy(&ptad);
numaDestroy(&nax);
numaDestroy(&nafit);
}
pixWrite("/tmp/textline6.png", pixt1, IFF_PNG);
pixDisplayWithTitle(pixt1, 1500, 100, "textline centers 6", 1);
pixaAddPix(pixa, pixt1, L_INSERT);
pixaConvertToPdf(pixa, 300, 0.5, L_JPEG_ENCODE, 75,
"LS fittings to textlines", "/tmp/dewarp_fittings.pdf");
pixaDestroy(&pixa);
pixDestroy(&pixs);
ptaaDestroy(&ptaa2);
return 0;
}
int main(int argc,
char **argv) {
char *infile;
l_int32 w, d, threshval, ival, newval;
l_uint32 val;
PIX *pixs, *pixg, *pixg2;
PIX *pix1, *pix2;
PIXA *pixa;
static char mainName[] = "binarize_set";
if (argc != 2)
return ERROR_INT(" Syntax: binarize_set infile", mainName, 1);
infile = argv[1];
pixa = pixaCreate(5);
pixs = pixRead(infile);
pixGetDimensions(pixs, &w, NULL, &d);
pixSaveTiled(pixs, pixa, 1.0, 1, 50, 32);
pixDisplay(pixs, 100, 0);
#if ALL
/* 1. Standard background normalization with a global threshold. */
pixg = pixConvertTo8(pixs, 0);
pix1 = pixBackgroundNorm(pixg, NULL, NULL, 10, 15, 100, 50, 255, 2, 2);
pix2 = pixThresholdToBinary(pix1, 160);
pixWrite("/tmp/binar1.png", pix2, IFF_PNG);
pixDisplay(pix2, 100, 0);
pixSaveTiled(pix2, pixa, 1.0, 1, 50, 32);
pixDestroy(&pixg);
pixDestroy(&pix1);
pixDestroy(&pix2);
#endif
#if ALL
/* 2. Background normalization followed by Otsu thresholding. Otsu
* binarization attempts to split the image into two roughly equal
* sets of pixels, and it does a very poor job when there are large
* amounts of dark background. By doing a background normalization
* first (to get the background near 255), we remove this problem.
* Then we use a modified Otsu to estimate the best global
* threshold on the normalized image. */
pixg = pixConvertTo8(pixs, 0);
pix1 = pixOtsuThreshOnBackgroundNorm(pixg, NULL, 10, 15, 100,
50, 255, 2, 2, 0.10, &threshval);
fprintf(stderr, "thresh val = %d\n", threshval);
pixSaveTiled(pix1, pixa, 1.0, 1, 50, 32);
pixWrite("/tmp/binar2.png", pix1, IFF_PNG);
pixDisplay(pix1, 100, 200);
pixDestroy(&pixg);
pixDestroy(&pix1);
#endif
#if ALL
/* 3. Background normalization with Otsu threshold estimation and
* masking for threshold selection. */
pixg = pixConvertTo8(pixs, 0);
pix1 = pixMaskedThreshOnBackgroundNorm(pixg, NULL, 10, 15, 100,
50, 2, 2, 0.10, &threshval);
fprintf(stderr, "thresh val = %d\n", threshval);
pixSaveTiled(pix1, pixa, 1.0, 1, 50, 32);
pixWrite("/tmp/binar3.png", pix1, IFF_PNG);
pixDisplay(pix1, 100, 400);
pixDestroy(&pixg);
pixDestroy(&pix1);
#endif
#if ALL
/* 4. Background normalization followed by Sauvola binarization */
if (d == 32)
pixg = pixConvertRGBToGray(pixs, 0.2, 0.7, 0.1);
else
pixg = pixConvertTo8(pixs, 0);
pixg2 = pixContrastNorm(NULL, pixg, 20, 20, 130, 2, 2);
pixSauvolaBinarizeTiled(pixg2, 25, 0.40, 1, 1, NULL, &pix1);
pixSaveTiled(pix1, pixa, 1.0, 1, 50, 32);
pixWrite("/tmp/binar4.png", pix1, IFF_PNG);
pixDisplay(pix1, 100, 600);
pixDestroy(&pixg);
pixDestroy(&pixg2);
pixDestroy(&pix1);
#endif
#if ALL
/* 5. Contrast normalization followed by background normalization, and
* thresholding. */
if (d == 32)
pixg = pixConvertRGBToGray(pixs, 0.2, 0.7, 0.1);
else
pixg = pixConvertTo8(pixs, 0);
pixOtsuAdaptiveThreshold(pixg, 5000, 5000, 0, 0, 0.1, &pix1, NULL);
pixGetPixel(pix1, 0, 0, &val);
ival = (l_int32) val;
newval = ival + (l_int32)(0.6 * (110 - ival));
fprintf(stderr, "th1 = %d, th2 = %d\n", ival, newval);
pixDestroy(&pix1);
pixContrastNorm(pixg, pixg, 50, 50, 130, 2, 2);
pixg2 = pixBackgroundNorm(pixg, NULL, NULL, 20, 20, 70, 40, 200, 2, 2);
ival = L_MIN(ival, 110);
pix1 = pixThresholdToBinary(pixg2, ival);
pixSaveTiled(pix1, pixa, 1.0, 1, 50, 32);
pixWrite("/tmp/binar5.png", pix1, IFF_PNG);
pixDisplay(pix1, 100, 800);
pixDestroy(&pixg);
pixDestroy(&pixg2);
pixDestroy(&pix1);
#endif
pix1 = pixaDisplayTiledInRows(pixa, 32, w + 100, 1.0, 0, 30, 2);
pixWrite("/tmp/binar6.png", pix1, IFF_PNG);
pixDisplay(pix1, 1000, 0);
pixDestroy(&pix1);
pixaDestroy(&pixa);
pixDestroy(&pixs);
return 0;
}
l_int32 main(int argc,
char **argv)
{
l_int32 method, pageno;
L_DEWARP *dew1;
L_DEWARPA *dewa;
PIX *pixs, *pixn, *pixg, *pixb, *pixd;
static char mainName[] = "dewarptest2";
if (argc != 2 && argc != 4)
return ERROR_INT("Syntax: dewarptest2 method [image pageno]",
mainName, 1);
if (argc == 2) {
pixs = pixRead("cat-35.jpg");
pageno = 35;
}
else {
pixs = pixRead(argv[2]);
pageno = atoi(argv[3]);
}
if (!pixs)
return ERROR_INT("image not read", mainName, 1);
method = atoi(argv[1]);
lept_mkdir("lept");
if (method == 1) { /* Use single page dewarp function */
dewarpSinglePage(pixs, 1, 100, 1, &pixd, NULL, 1);
pixDisplay(pixd, 100, 100);
} else { /* Break down into multiple steps; require min of only 6 lines */
dewa = dewarpaCreate(40, 30, 1, 6, 50);
dewarpaUseBothArrays(dewa, 1);
#if NORMALIZE
/* Normalize for varying background and binarize */
pixn = pixBackgroundNormSimple(pixs, NULL, NULL);
pixg = pixConvertRGBToGray(pixn, 0.5, 0.3, 0.2);
pixb = pixThresholdToBinary(pixg, 130);
pixDestroy(&pixn);
#else
/* Don't normalize; just threshold and clean edges */
pixg = pixConvertTo8(pixs, 0);
pixb = pixThresholdToBinary(pixg, 100);
pixSetOrClearBorder(pixb, 30, 30, 40, 40, PIX_CLR);
#endif
/* Run the basic functions */
dew1 = dewarpCreate(pixb, pageno);
dewarpaInsertDewarp(dewa, dew1);
dewarpBuildPageModel(dew1, "/tmp/lept/test2_model.pdf");
dewarpaApplyDisparity(dewa, pageno, pixg, -1, 0, 0, &pixd,
"/tmp/lept/test2_apply.pdf");
dewarpaInfo(stderr, dewa);
dewarpaDestroy(&dewa);
pixDestroy(&pixg);
pixDestroy(&pixb);
}
pixDestroy(&pixs);
pixDestroy(&pixd);
return 0;
}
int main(int argc,
char **argv)
{
l_int32 index;
l_uint32 val32;
BOX *box, *box1, *box2, *box3, *box4, *box5;
BOXA *boxa;
L_KERNEL *kel;
PIX *pixs, *pixg, *pixb, *pixd, *pixt, *pix1, *pix2, *pix3, *pix4;
PIXA *pixa;
PIXCMAP *cmap;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
pixa = pixaCreate(0);
/* Color non-white pixels on RGB */
pixs = pixRead("lucasta-frag.jpg");
pixt = pixConvert8To32(pixs);
box = boxCreate(120, 30, 200, 200);
pixColorGray(pixt, box, L_PAINT_DARK, 220, 0, 0, 255);
regTestWritePixAndCheck(rp, pixt, IFF_JFIF_JPEG); /* 0 */
pixaAddPix(pixa, pixt, L_COPY);
pixColorGray(pixt, NULL, L_PAINT_DARK, 220, 255, 100, 100);
regTestWritePixAndCheck(rp, pixt, IFF_JFIF_JPEG); /* 1 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
/* Color non-white pixels on colormap */
pixt = pixThresholdTo4bpp(pixs, 6, 1);
box = boxCreate(120, 30, 200, 200);
pixColorGray(pixt, box, L_PAINT_DARK, 220, 0, 0, 255);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 2 */
pixaAddPix(pixa, pixt, L_COPY);
pixColorGray(pixt, NULL, L_PAINT_DARK, 220, 255, 100, 100);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 3 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
/* Color non-black pixels on RGB */
pixt = pixConvert8To32(pixs);
box = boxCreate(120, 30, 200, 200);
pixColorGray(pixt, box, L_PAINT_LIGHT, 20, 0, 0, 255);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 4 */
pixaAddPix(pixa, pixt, L_COPY);
pixColorGray(pixt, NULL, L_PAINT_LIGHT, 80, 255, 100, 100);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 5 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
/* Color non-black pixels on colormap */
pixt = pixThresholdTo4bpp(pixs, 6, 1);
box = boxCreate(120, 30, 200, 200);
pixColorGray(pixt, box, L_PAINT_LIGHT, 20, 0, 0, 255);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 6 */
pixaAddPix(pixa, pixt, L_COPY);
pixColorGray(pixt, NULL, L_PAINT_LIGHT, 20, 255, 100, 100);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 7 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
/* Add highlight color to RGB */
pixt = pixConvert8To32(pixs);
box = boxCreate(507, 5, 385, 45);
pixg = pixClipRectangle(pixs, box, NULL);
pixb = pixThresholdToBinary(pixg, 180);
pixInvert(pixb, pixb);
pixDisplayWrite(pixb, 1);
composeRGBPixel(50, 0, 250, &val32);
pixPaintThroughMask(pixt, pixb, box->x, box->y, val32);
boxDestroy(&box);
pixDestroy(&pixg);
pixDestroy(&pixb);
box = boxCreate(236, 107, 262, 40);
pixg = pixClipRectangle(pixs, box, NULL);
pixb = pixThresholdToBinary(pixg, 180);
pixInvert(pixb, pixb);
composeRGBPixel(250, 0, 50, &val32);
pixPaintThroughMask(pixt, pixb, box->x, box->y, val32);
boxDestroy(&box);
pixDestroy(&pixg);
pixDestroy(&pixb);
box = boxCreate(222, 208, 247, 43);
pixg = pixClipRectangle(pixs, box, NULL);
pixb = pixThresholdToBinary(pixg, 180);
pixInvert(pixb, pixb);
composeRGBPixel(60, 250, 60, &val32);
pixPaintThroughMask(pixt, pixb, box->x, box->y, val32);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 8 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
pixDestroy(&pixg);
pixDestroy(&pixb);
/* Add highlight color to colormap */
pixt = pixThresholdTo4bpp(pixs, 5, 1);
cmap = pixGetColormap(pixt);
pixcmapGetIndex(cmap, 255, 255, 255, &index);
box = boxCreate(507, 5, 385, 45);
pixSetSelectCmap(pixt, box, index, 50, 0, 250);
boxDestroy(&box);
box = boxCreate(236, 107, 262, 40);
pixSetSelectCmap(pixt, box, index, 250, 0, 50);
boxDestroy(&box);
box = boxCreate(222, 208, 247, 43);
pixSetSelectCmap(pixt, box, index, 60, 250, 60);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 9 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
/* Paint lines on RGB */
pixt = pixConvert8To32(pixs);
pixRenderLineArb(pixt, 450, 20, 850, 320, 5, 200, 50, 125);
pixRenderLineArb(pixt, 30, 40, 440, 40, 5, 100, 200, 25);
box = boxCreate(70, 80, 300, 245);
pixRenderBoxArb(pixt, box, 3, 200, 200, 25);
regTestWritePixAndCheck(rp, pixt, IFF_JFIF_JPEG); /* 10 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
/* Paint lines on colormap */
pixt = pixThresholdTo4bpp(pixs, 5, 1);
pixRenderLineArb(pixt, 450, 20, 850, 320, 5, 200, 50, 125);
pixRenderLineArb(pixt, 30, 40, 440, 40, 5, 100, 200, 25);
box = boxCreate(70, 80, 300, 245);
pixRenderBoxArb(pixt, box, 3, 200, 200, 25);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 11 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
/* Blend lines on RGB */
pixt = pixConvert8To32(pixs);
pixRenderLineBlend(pixt, 450, 20, 850, 320, 5, 200, 50, 125, 0.35);
pixRenderLineBlend(pixt, 30, 40, 440, 40, 5, 100, 200, 25, 0.35);
box = boxCreate(70, 80, 300, 245);
pixRenderBoxBlend(pixt, box, 3, 200, 200, 25, 0.6);
regTestWritePixAndCheck(rp, pixt, IFF_JFIF_JPEG); /* 12 */
pixaAddPix(pixa, pixt, L_INSERT);
boxDestroy(&box);
/* Colorize gray on cmapped image. */
pix1 = pixRead("lucasta.150.jpg");
pix2 = pixThresholdTo4bpp(pix1, 7, 1);
box1 = boxCreate(73, 206, 140, 27);
pixColorGrayCmap(pix2, box1, L_PAINT_LIGHT, 130, 207, 43);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 13 */
pixaAddPix(pixa, pix2, L_COPY);
if (rp->display)
pixPrintStreamInfo(stderr, pix2, "One box added");
box2 = boxCreate(255, 404, 197, 25);
pixColorGrayCmap(pix2, box2, L_PAINT_LIGHT, 230, 67, 119);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 14 */
pixaAddPix(pixa, pix2, L_COPY);
if (rp->display)
pixPrintStreamInfo(stderr, pix2, "Two boxes added");
box3 = boxCreate(122, 756, 224, 22);
pixColorGrayCmap(pix2, box3, L_PAINT_DARK, 230, 67, 119);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 15 */
pixaAddPix(pixa, pix2, L_COPY);
if (rp->display)
pixPrintStreamInfo(stderr, pix2, "Three boxes added");
box4 = boxCreate(11, 780, 147, 22);
pixColorGrayCmap(pix2, box4, L_PAINT_LIGHT, 70, 137, 229);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 16 */
pixaAddPix(pixa, pix2, L_COPY);
if (rp->display)
pixPrintStreamInfo(stderr, pix2, "Four boxes added");
box5 = boxCreate(163, 605, 78, 22);
pixColorGrayCmap(pix2, box5, L_PAINT_LIGHT, 70, 137, 229);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 17 */
pixaAddPix(pixa, pix2, L_INSERT);
if (rp->display)
pixPrintStreamInfo(stderr, pix2, "Five boxes added");
pixDestroy(&pix1);
boxDestroy(&box1);
boxDestroy(&box2);
boxDestroy(&box3);
boxDestroy(&box4);
boxDestroy(&box5);
pixDestroy(&pixs);
/* Make a gray image and identify the fg pixels (val > 230) */
pixs = pixRead("feyn-fract.tif");
pix1 = pixConvertTo8(pixs, 0);
kel = makeGaussianKernel(2, 2, 1.5, 1.0);
pix2 = pixConvolve(pix1, kel, 8, 1);
pix3 = pixThresholdToBinary(pix2, 230);
boxa = pixConnComp(pix3, NULL, 8);
pixDestroy(&pixs);
pixDestroy(&pix1);
pixDestroy(&pix3);
kernelDestroy(&kel);
/* Color the individual components in the gray image */
pix4 = pixColorGrayRegions(pix2, boxa, L_PAINT_DARK, 230, 255, 0, 0);
regTestWritePixAndCheck(rp, pix4, IFF_PNG); /* 18 */
pixaAddPix(pixa, pix4, L_INSERT);
pixDisplayWithTitle(pix4, 0, 0, NULL, rp->display);
/* Threshold to 10 levels of gray */
pix3 = pixThresholdOn8bpp(pix2, 10, 1);
regTestWritePixAndCheck(rp, pix3, IFF_PNG); /* 19 */
pixaAddPix(pixa, pix3, L_COPY);
/* Color the individual components in the cmapped image */
pix4 = pixColorGrayRegions(pix3, boxa, L_PAINT_DARK, 230, 255, 0, 0);
regTestWritePixAndCheck(rp, pix4, IFF_PNG); /* 20 */
pixaAddPix(pixa, pix4, L_INSERT);
pixDisplayWithTitle(pix4, 0, 100, NULL, rp->display);
boxaDestroy(&boxa);
/* Color the entire gray image (not component-wise) */
pixColorGray(pix2, NULL, L_PAINT_DARK, 230, 255, 0, 0);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 21 */
pixaAddPix(pixa, pix2, L_INSERT);
/* Color the entire cmapped image (not component-wise) */
pixColorGray(pix3, NULL, L_PAINT_DARK, 230, 255, 0, 0);
regTestWritePixAndCheck(rp, pix3, IFF_PNG); /* 22 */
pixaAddPix(pixa, pix3, L_INSERT);
/* Reconstruct cmapped images */
pixd = ReconstructByValue(rp, "weasel2.4c.png");
regTestWritePixAndCheck(rp, pixd, IFF_PNG); /* 23 */
pixaAddPix(pixa, pixd, L_INSERT);
pixd = ReconstructByValue(rp, "weasel4.11c.png");
regTestWritePixAndCheck(rp, pixd, IFF_PNG); /* 24 */
pixaAddPix(pixa, pixd, L_INSERT);
pixd = ReconstructByValue(rp, "weasel8.240c.png");
regTestWritePixAndCheck(rp, pixd, IFF_PNG); /* 25 */
pixaAddPix(pixa, pixd, L_INSERT);
/* Fake reconstruct cmapped images, with one color into a band */
pixd = FakeReconstructByBand(rp, "weasel2.4c.png");
regTestWritePixAndCheck(rp, pixd, IFF_PNG); /* 26 */
pixaAddPix(pixa, pixd, L_INSERT);
pixd = FakeReconstructByBand(rp, "weasel4.11c.png");
regTestWritePixAndCheck(rp, pixd, IFF_PNG); /* 27 */
pixaAddPix(pixa, pixd, L_INSERT);
pixd = FakeReconstructByBand(rp, "weasel8.240c.png");
regTestWritePixAndCheck(rp, pixd, IFF_PNG); /* 28 */
pixaAddPix(pixa, pixd, L_INSERT);
/* If in testing mode, make a pdf */
if (rp->display) {
pixaConvertToPdf(pixa, 100, 1.0, L_FLATE_ENCODE, 0,
"Colorize and paint", "/tmp/lept/regout/paint.pdf");
L_INFO("Output pdf: /tmp/lept/regout/paint.pdf\n", rp->testname);
}
pixaDestroy(&pixa);
return regTestCleanup(rp);
}
main(int argc,
char **argv)
{
l_float32 sum, sumx, sumy, diff;
L_DEWARP *dew;
L_DEWARPA *dewa;
FPIX *fpixs, *fpixs2, *fpixs3, *fpixs4, *fpixg, *fpixd;
FPIX *fpix1, *fpix2, *fpixt1, *fpixt2;
DPIX *dpix, *dpix2;
L_KERNEL *kel, *kelx, *kely;
PIX *pixs, *pixs2, *pixs3, *pixt, *pixd, *pixg, *pixb, *pixn;
PIX *pixt1, *pixt2, *pixt3, *pixt4, *pixt5, *pixt6;
PIXA *pixa;
PTA *ptas, *ptad;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
pixa = pixaCreate(0);
/* Gaussian kernel */
kel = makeGaussianKernel(5, 5, 3.0, 4.0);
kernelGetSum(kel, &sum);
if (rp->display) fprintf(stderr, "Sum for 2d gaussian kernel = %f\n", sum);
pixt = kernelDisplayInPix(kel, 41, 2);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 0 */
pixSaveTiled(pixt, pixa, 1, 1, 20, 8);
pixDestroy(&pixt);
/* Separable gaussian kernel */
makeGaussianKernelSep(5, 5, 3.0, 4.0, &kelx, &kely);
kernelGetSum(kelx, &sumx);
if (rp->display) fprintf(stderr, "Sum for x gaussian kernel = %f\n", sumx);
kernelGetSum(kely, &sumy);
if (rp->display) fprintf(stderr, "Sum for y gaussian kernel = %f\n", sumy);
if (rp->display) fprintf(stderr, "Sum for x * y gaussian kernel = %f\n",
sumx * sumy);
pixt = kernelDisplayInPix(kelx, 41, 2);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 1 */
pixSaveTiled(pixt, pixa, 1, 0, 20, 8);
pixDestroy(&pixt);
pixt = kernelDisplayInPix(kely, 41, 2);
regTestWritePixAndCheck(rp, pixt, IFF_PNG); /* 2 */
pixSaveTiled(pixt, pixa, 1, 0, 20, 8);
pixDestroy(&pixt);
/* Use pixRasterop() to generate source image */
pixs = pixRead("test8.jpg");
pixs2 = pixRead("karen8.jpg");
pixRasterop(pixs, 150, 125, 150, 100, PIX_SRC, pixs2, 75, 100);
regTestWritePixAndCheck(rp, pixs, IFF_JFIF_JPEG); /* 3 */
/* Convolution directly with pix */
pixt1 = pixConvolve(pixs, kel, 8, 1);
regTestWritePixAndCheck(rp, pixt1, IFF_JFIF_JPEG); /* 4 */
pixSaveTiled(pixt1, pixa, 1, 1, 20, 8);
pixt2 = pixConvolveSep(pixs, kelx, kely, 8, 1);
regTestWritePixAndCheck(rp, pixt2, IFF_JFIF_JPEG); /* 5 */
pixSaveTiled(pixt2, pixa, 1, 0, 20, 8);
/* Convolution indirectly with fpix, using fpixRasterop()
* to generate the source image. */
fpixs = pixConvertToFPix(pixs, 3);
fpixs2 = pixConvertToFPix(pixs2, 3);
fpixRasterop(fpixs, 150, 125, 150, 100, fpixs2, 75, 100);
fpixt1 = fpixConvolve(fpixs, kel, 1);
pixt3 = fpixConvertToPix(fpixt1, 8, L_CLIP_TO_ZERO, 1);
regTestWritePixAndCheck(rp, pixt3, IFF_JFIF_JPEG); /* 6 */
pixSaveTiled(pixt3, pixa, 1, 1, 20, 8);
fpixt2 = fpixConvolveSep(fpixs, kelx, kely, 1);
pixt4 = fpixConvertToPix(fpixt2, 8, L_CLIP_TO_ZERO, 1);
regTestWritePixAndCheck(rp, pixt4, IFF_JFIF_JPEG); /* 7 */
pixSaveTiled(pixt4, pixa, 1, 0, 20, 8);
pixDestroy(&pixs2);
fpixDestroy(&fpixs2);
fpixDestroy(&fpixt1);
fpixDestroy(&fpixt2);
/* Comparison of results */
pixCompareGray(pixt1, pixt2, L_COMPARE_ABS_DIFF, 0, NULL,
&diff, NULL, NULL);
if (rp->display)
fprintf(stderr, "Ave diff of pixConvolve and pixConvolveSep: %f\n",
diff);
pixCompareGray(pixt3, pixt4, L_COMPARE_ABS_DIFF, 0, NULL,
&diff, NULL, NULL);
if (rp->display)
fprintf(stderr, "Ave diff of fpixConvolve and fpixConvolveSep: %f\n",
diff);
pixCompareGray(pixt1, pixt3, L_COMPARE_ABS_DIFF, 0, NULL,
&diff, NULL, NULL);
if (rp->display)
fprintf(stderr, "Ave diff of pixConvolve and fpixConvolve: %f\n", diff);
pixCompareGray(pixt2, pixt4, L_COMPARE_ABS_DIFF, GPLOT_PNG, NULL,
&diff, NULL, NULL);
if (rp->display)
fprintf(stderr, "Ave diff of pixConvolveSep and fpixConvolveSep: %f\n",
diff);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
pixDestroy(&pixt4);
/* Test arithmetic operations; add in a fraction rotated by 180 */
pixs3 = pixRotate180(NULL, pixs);
regTestWritePixAndCheck(rp, pixs3, IFF_JFIF_JPEG); /* 8 */
pixSaveTiled(pixs3, pixa, 1, 1, 20, 8);
fpixs3 = pixConvertToFPix(pixs3, 3);
fpixd = fpixLinearCombination(NULL, fpixs, fpixs3, 20.0, 5.0);
fpixAddMultConstant(fpixd, 0.0, 23.174); /* multiply up in magnitude */
pixd = fpixDisplayMaxDynamicRange(fpixd); /* bring back to 8 bpp */
regTestWritePixAndCheck(rp, pixd, IFF_JFIF_JPEG); /* 9 */
pixSaveTiled(pixd, pixa, 1, 0, 20, 8);
pixDestroy(&pixs3);
fpixDestroy(&fpixs3);
fpixDestroy(&fpixd);
pixDestroy(&pixd);
pixDestroy(&pixs);
fpixDestroy(&fpixs);
/* Save the comparison graph; gnuplot should have made it by now! */
#ifndef _WIN32
sleep(2);
#else
Sleep(2000);
#endif /* _WIN32 */
pixt5 = pixRead("/tmp/grayroot.png");
regTestWritePixAndCheck(rp, pixt5, IFF_PNG); /* 10 */
pixSaveTiled(pixt5, pixa, 1, 1, 20, 8);
pixDestroy(&pixt5);
/* Display results */
pixd = pixaDisplay(pixa, 0, 0);
regTestWritePixAndCheck(rp, pixd, IFF_JFIF_JPEG); /* 11 */
pixDisplayWithTitle(pixd, 100, 100, NULL, rp->display);
pixDestroy(&pixd);
pixaDestroy(&pixa);
/* Test some more convolutions, with sampled output. First on pix */
pixa = pixaCreate(0);
pixs = pixRead("1555-7.jpg");
pixg = pixConvertTo8(pixs, 0);
l_setConvolveSampling(5, 5);
pixt1 = pixConvolve(pixg, kel, 8, 1);
regTestWritePixAndCheck(rp, pixt1, IFF_JFIF_JPEG); /* 12 */
pixSaveTiled(pixt1, pixa, 1, 1, 20, 32);
pixt2 = pixConvolveSep(pixg, kelx, kely, 8, 1);
regTestWritePixAndCheck(rp, pixt2, IFF_JFIF_JPEG); /* 13 */
pixSaveTiled(pixt2, pixa, 1, 0, 20, 32);
pixt3 = pixConvolveRGB(pixs, kel);
regTestWritePixAndCheck(rp, pixt3, IFF_JFIF_JPEG); /* 14 */
pixSaveTiled(pixt3, pixa, 1, 0, 20, 32);
pixt4 = pixConvolveRGBSep(pixs, kelx, kely);
regTestWritePixAndCheck(rp, pixt4, IFF_JFIF_JPEG); /* 15 */
pixSaveTiled(pixt4, pixa, 1, 0, 20, 32);
/* Then on fpix */
fpixg = pixConvertToFPix(pixg, 1);
fpixt1 = fpixConvolve(fpixg, kel, 1);
pixt5 = fpixConvertToPix(fpixt1, 8, L_CLIP_TO_ZERO, 0);
regTestWritePixAndCheck(rp, pixt5, IFF_JFIF_JPEG); /* 16 */
pixSaveTiled(pixt5, pixa, 1, 1, 20, 32);
fpixt2 = fpixConvolveSep(fpixg, kelx, kely, 1);
pixt6 = fpixConvertToPix(fpixt2, 8, L_CLIP_TO_ZERO, 0);
regTestWritePixAndCheck(rp, pixt6, IFF_JFIF_JPEG); /* 17 */
pixSaveTiled(pixt2, pixa, 1, 0, 20, 32);
regTestCompareSimilarPix(rp, pixt1, pixt5, 2, 0.00, 0); /* 18 */
regTestCompareSimilarPix(rp, pixt2, pixt6, 2, 0.00, 0); /* 19 */
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
pixDestroy(&pixt4);
pixDestroy(&pixt5);
pixDestroy(&pixt6);
fpixDestroy(&fpixg);
fpixDestroy(&fpixt1);
fpixDestroy(&fpixt2);
pixd = pixaDisplay(pixa, 0, 0);
regTestWritePixAndCheck(rp, pixd, IFF_JFIF_JPEG); /* 20 */
pixDisplayWithTitle(pixd, 600, 100, NULL, rp->display);
pixDestroy(&pixs);
pixDestroy(&pixg);
pixDestroy(&pixd);
pixaDestroy(&pixa);
/* Test extension (continued and slope).
* First, build a smooth vertical disparity array;
* then extend and show the contours. */
pixs = pixRead("cat-35.jpg");
pixn = pixBackgroundNormSimple(pixs, NULL, NULL);
pixg = pixConvertRGBToGray(pixn, 0.5, 0.3, 0.2);
pixb = pixThresholdToBinary(pixg, 130);
dewa = dewarpaCreate(1, 30, 1, 15, 0);
dew = dewarpCreate(pixb, 35);
dewarpaInsertDewarp(dewa, dew);
dewarpBuildModel(dew, NULL);
dewarpPopulateFullRes(dew, NULL);
fpixs = dew->fullvdispar;
fpixs2 = fpixAddContinuedBorder(fpixs, 200, 200, 100, 300);
fpixs3 = fpixAddSlopeBorder(fpixs, 200, 200, 100, 300);
dpix = fpixConvertToDPix(fpixs3);
fpixs4 = dpixConvertToFPix(dpix);
pixt1 = fpixRenderContours(fpixs, 2.0, 0.2);
pixt2 = fpixRenderContours(fpixs2, 2.0, 0.2);
pixt3 = fpixRenderContours(fpixs3, 2.0, 0.2);
pixt4 = fpixRenderContours(fpixs4, 2.0, 0.2);
pixt5 = pixRead("karen8.jpg");
dpix2 = pixConvertToDPix(pixt5, 1);
pixt6 = dpixConvertToPix(dpix2, 8, L_CLIP_TO_ZERO, 0);
regTestWritePixAndCheck(rp, pixt1, IFF_PNG); /* 21 */
pixDisplayWithTitle(pixt1, 0, 100, NULL, rp->display);
regTestWritePixAndCheck(rp, pixt2, IFF_PNG); /* 22 */
pixDisplayWithTitle(pixt2, 470, 100, NULL, rp->display);
regTestWritePixAndCheck(rp, pixt3, IFF_PNG); /* 23 */
pixDisplayWithTitle(pixt3, 1035, 100, NULL, rp->display);
regTestComparePix(rp, pixt3, pixt4); /* 24 */
regTestComparePix(rp, pixt5, pixt6); /* 25 */
pixDestroy(&pixs);
pixDestroy(&pixn);
pixDestroy(&pixg);
pixDestroy(&pixb);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
pixDestroy(&pixt4);
pixDestroy(&pixt5);
pixDestroy(&pixt6);
fpixDestroy(&fpixs2);
fpixDestroy(&fpixs3);
fpixDestroy(&fpixs4);
dpixDestroy(&dpix);
dpixDestroy(&dpix2);
/* Test affine and projective transforms on fpix */
fpixWrite("/tmp/fpix1.fp", dew->fullvdispar);
fpix1 = fpixRead("/tmp/fpix1.fp");
pixt1 = fpixAutoRenderContours(fpix1, 40);
regTestWritePixAndCheck(rp, pixt1, IFF_PNG); /* 26 */
pixDisplayWithTitle(pixt1, 0, 500, NULL, rp->display);
pixDestroy(&pixt1);
MakePtasAffine(1, &ptas, &ptad);
fpix2 = fpixAffinePta(fpix1, ptad, ptas, 200, 0.0);
pixt2 = fpixAutoRenderContours(fpix2, 40);
regTestWritePixAndCheck(rp, pixt2, IFF_PNG); /* 27 */
pixDisplayWithTitle(pixt2, 400, 500, NULL, rp->display);
fpixDestroy(&fpix2);
pixDestroy(&pixt2);
ptaDestroy(&ptas);
ptaDestroy(&ptad);
MakePtas(1, &ptas, &ptad);
fpix2 = fpixProjectivePta(fpix1, ptad, ptas, 200, 0.0);
pixt3 = fpixAutoRenderContours(fpix2, 40);
regTestWritePixAndCheck(rp, pixt3, IFF_PNG); /* 28 */
pixDisplayWithTitle(pixt3, 400, 500, NULL, rp->display);
fpixDestroy(&fpix2);
pixDestroy(&pixt3);
ptaDestroy(&ptas);
ptaDestroy(&ptad);
fpixDestroy(&fpix1);
dewarpaDestroy(&dewa);
kernelDestroy(&kel);
kernelDestroy(&kelx);
kernelDestroy(&kely);
return regTestCleanup(rp);
}
/*!
* deskew()
*
* Input: pixs
* redsearch (for binary search: reduction factor = 1, 2 or 4)
* Return: deskewed pix, or NULL on error
*/
PIX *
deskew(PIX *pixs,
l_int32 redsearch)
{
l_float32 angle, conf, deg2rad;
PIX *pixg; /* gray version */
PIX *pixb; /* binary version */
PIX *pixd; /* destination image */
PROCNAME("deskew");
if (!pixs)
return (PIX *)ERROR_PTR("pixs not defined", procName, NULL);
/* Calculate a skew angle. We may need to make a binary version of the
* image for this calculation.
*/
if (pixGetDepth(pixs) != 1) {
/* FIX ME: We should probably pick a threshold value with more care. */
/* Create a grayscale image if we need one. */
if (pixGetDepth(pixs) >= 24) {
pixg = pixConvertRGBToGray(pixs, 0.0, 0.0, 0.0);
} else {
pixg = pixs;
}
pixb = pixThresholdToBinary(pixg, 127);
if (pixg != pixs) {
pixDestroy(&pixg);
}
/* Assert: We are done with any gray image. */
} else {
pixb = pixs;
}
/* Assert: We have a valid binary image. */
if (redsearch != 1 && redsearch != 2 && redsearch != 4)
return (PIX *)ERROR_PTR("redsearch not in {1,2,4}", procName, NULL);
deg2rad = 3.1415926535 / 180.;
if (pixFindSkewSweepAndSearch(pixb, &angle, &conf,
DEFAULT_SWEEP_REDUCTION, redsearch,
DEFAULT_SWEEP_RANGE, DEFAULT_SWEEP_DELTA,
DEFAULT_MINBS_DELTA)) {
pixd = pixClone(pixs);
goto finish;
}
if (L_ABS(angle) < MIN_DESKEW_ANGLE || conf < MIN_ALLOWED_CONFIDENCE) {
pixd = pixClone(pixs);
goto finish;
}
/* If the pixel depth of pixs is 1, we need to use a bit-depth
* independent rotate instead of the more accurate area mapping rotate.
*/
if (pixGetDepth(pixs) == 1) {
if ((pixd = pixRotateShear(pixs, 0, 0, deg2rad * angle, 0xffffff00)) == NULL) {
pixd = pixClone(pixs);
}
} else {
#if defined(COLOR_ROTATE)
if ((pixd = pixRotateAMColorFast(pixs, deg2rad * angle)) == NULL) {
pixd = pixClone(pixs);
}
#else
if ((pixd = pixRotateAM(pixs, deg2rad * angle, 0xffffff00)) == NULL) {
pixd = pixClone(pixs);
}
#endif
}
finish:
if (pixb != pixs) {
pixDestroy(&pixb);
}
return pixd;
}
int main(int argc,
char **argv)
{
char label[512];
l_int32 rval, gval, bval, w, h, i, j, rwhite, gwhite, bwhite, count;
l_uint32 pixel;
GPLOT *gplot1, *gplot2;
NUMA *naseq, *na;
NUMAA *naa1, *naa2;
PIX *pixs, *pixt, *pixt0, *pixt1, *pixt2;
PIX *pixr, *pixg, *pixb;
PIXA *pixa;
PIXCMAP *cmap;
static char mainName[] = "colorspacetest";
if (argc != 2)
return ERROR_INT(" Syntax: colorspacetest filein", mainName, 1);
if ((pixs = pixRead(argv[1])) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
/* Generate colors by sampling hue with max sat and value.
* This was used to make the color strip 19-colors.png. */
pixa = pixaCreate(19);
for (i = 0; i < 19; i++) {
convertHSVToRGB((240 * i / 18), 255, 255, &rval, &gval, &bval);
composeRGBPixel(rval, gval, bval, &pixel);
pixt1 = pixCreate(50, 100, 32);
pixSetAllArbitrary(pixt1, pixel);
pixaAddPix(pixa, pixt1, L_INSERT);
}
pixt2 = pixaDisplayTiledInRows(pixa, 32, 1100, 1.0, 0, 0, 0);
pixDisplayWrite(pixt2, 1);
pixDestroy(&pixt2);
pixaDestroy(&pixa);
/* Colorspace conversion in rgb */
pixDisplayWrite(pixs, 1);
pixt = pixConvertRGBToHSV(NULL, pixs);
pixDisplayWrite(pixt, 1);
pixConvertHSVToRGB(pixt, pixt);
pixDisplayWrite(pixt, 1);
pixDestroy(&pixt);
/* Colorspace conversion on a colormap */
pixt = pixOctreeQuantNumColors(pixs, 25, 0);
pixDisplayWrite(pixt, 1);
cmap = pixGetColormap(pixt);
pixcmapWriteStream(stderr, cmap);
pixcmapConvertRGBToHSV(cmap);
pixcmapWriteStream(stderr, cmap);
pixDisplayWrite(pixt, 1);
pixcmapConvertHSVToRGB(cmap);
pixcmapWriteStream(stderr, cmap);
pixDisplayWrite(pixt, 1);
pixDestroy(&pixt);
/* Color content extraction */
pixColorContent(pixs, 0, 0, 0, 0, &pixr, &pixg, &pixb);
pixDisplayWrite(pixr, 1);
pixDisplayWrite(pixg, 1);
pixDisplayWrite(pixb, 1);
pixDestroy(&pixr);
pixDestroy(&pixg);
pixDestroy(&pixb);
/* Color content measurement */
pixa = pixaCreate(20);
naseq = numaMakeSequence(100, 5, 20);
naa1 = numaaCreate(6);
naa2 = numaaCreate(6);
for (i = 0; i < 6; i++) {
na = numaCreate(20);
numaaAddNuma(naa1, na, L_COPY);
numaaAddNuma(naa2, na, L_INSERT);
}
pixGetDimensions(pixs, &w, &h, NULL);
for (i = 0; i < 20; i++) {
rwhite = 100 + 5 * i;
gwhite = 200 - 5 * i;
bwhite = 150;
pixt0 = pixGlobalNormRGB(NULL, pixs, rwhite, gwhite, bwhite, 255);
pixaAddPix(pixa, pixt0, L_INSERT);
pixt1 = pixColorMagnitude(pixs, rwhite, gwhite, bwhite,
L_MAX_DIFF_FROM_AVERAGE_2);
for (j = 0; j < 6; j++) {
pixt2 = pixThresholdToBinary(pixt1, 30 + 10 * j);
pixInvert(pixt2, pixt2);
pixCountPixels(pixt2, &count, NULL);
na = numaaGetNuma(naa1, j, L_CLONE);
numaAddNumber(na, (l_float32)count / (l_float32)(w * h));
numaDestroy(&na);
pixDestroy(&pixt2);
}
pixDestroy(&pixt1);
pixt1 = pixColorMagnitude(pixs, rwhite, gwhite, bwhite,
L_MAX_MIN_DIFF_FROM_2);
for (j = 0; j < 6; j++) {
pixt2 = pixThresholdToBinary(pixt1, 30 + 10 * j);
pixInvert(pixt2, pixt2);
pixCountPixels(pixt2, &count, NULL);
na = numaaGetNuma(naa2, j, L_CLONE);
numaAddNumber(na, (l_float32)count / (l_float32)(w * h));
numaDestroy(&na);
pixDestroy(&pixt2);
}
pixDestroy(&pixt1);
}
gplot1 = gplotCreate("/tmp/junkplot1", GPLOT_X11,
"Fraction with given color (diff from average)",
"white point space for red", "amount of color");
gplot2 = gplotCreate("/tmp/junkplot2", GPLOT_X11,
"Fraction with given color (min diff)",
"white point space for red", "amount of color");
for (j = 0; j < 6; j++) {
na = numaaGetNuma(naa1, j, L_CLONE);
sprintf(label, "thresh %d", 30 + 10 * j);
gplotAddPlot(gplot1, naseq, na, GPLOT_LINES, label);
numaDestroy(&na);
na = numaaGetNuma(naa2, j, L_CLONE);
gplotAddPlot(gplot2, naseq, na, GPLOT_LINES, label);
numaDestroy(&na);
}
gplotMakeOutput(gplot1);
gplotMakeOutput(gplot2);
gplotDestroy(&gplot1);
gplotDestroy(&gplot2);
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, 250, 4, 0, 10, 2);
pixWrite("/tmp/junkcolormag", pixt1, IFF_PNG);
pixDisplayWithTitle(pixt1, 0, 100, "Color magnitude", 1);
pixDestroy(&pixt1);
pixaDestroy(&pixa);
numaDestroy(&naseq);
numaaDestroy(&naa1);
numaaDestroy(&naa2);
pixDisplayMultiple("/tmp/display/file*");
pixDestroy(&pixs);
return 0;
}
/*
* pixWriteSegmentedPageToPS()
*
* Input: pixs (all depths; colormap ok)
* pixm (<optional> 1 bpp segmentation mask over image region)
* textscale (scale of text output relative to pixs)
* imagescale (scale of image output relative to pixs)
* threshold (threshold for binarization; typ. 190)
* pageno (page number in set; use 1 for new output file)
* fileout (output ps file)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) This generates the PS string for a mixed text/image page,
* and adds it to an existing file if @pageno > 1.
* The PS output is determined by fitting the result to
* a letter-size (8.5 x 11 inch) page.
* (2) The two images (pixs and pixm) are at the same resolution
* (typically 300 ppi). They are used to generate two compressed
* images, pixb and pixc, that are put directly into the output
* PS file.
* (3) pixb is the text component. In the PostScript world, we think of
* it as a mask through which we paint black. It is produced by
* scaling pixs by @textscale, and thresholding to 1 bpp.
* (4) pixc is the image component, which is that part of pixs under
* the mask pixm. It is scaled from pixs by @imagescale.
* (5) Typical values are textscale = 2.0 and imagescale = 0.5.
* (6) If pixm == NULL, the page has only text. If it is all black,
* the page is all image and has no text.
* (7) This can be used to write a multi-page PS file, by using
* sequential page numbers with the same output file. It can
* also be used to write separate PS files for each page,
* by using different output files with @pageno = 0 or 1.
*/
l_int32
pixWriteSegmentedPageToPS(PIX *pixs,
PIX *pixm,
l_float32 textscale,
l_float32 imagescale,
l_int32 threshold,
l_int32 pageno,
const char *fileout)
{
l_int32 alltext, notext, d, ret;
l_uint32 val;
l_float32 scaleratio;
PIX *pixmi, *pixmis, *pixt, *pixg, *pixsc, *pixb, *pixc;
PROCNAME("pixWriteSegmentedPageToPS");
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (!fileout)
return ERROR_INT("fileout not defined", procName, 1);
if (imagescale <= 0.0 || textscale <= 0.0)
return ERROR_INT("relative scales must be > 0.0", procName, 1);
/* Analyze the page. Determine the ratio by which the
* binary text mask is scaled relative to the image part.
* If there is no image region (alltext == TRUE), the
* text mask will be rendered directly to fit the page,
* and scaleratio = 1.0. */
alltext = TRUE;
notext = FALSE;
scaleratio = 1.0;
if (pixm) {
pixZero(pixm, &alltext); /* pixm empty: all text */
if (alltext)
pixm = NULL; /* treat it as not existing here */
else {
pixmi = pixInvert(NULL, pixm);
pixZero(pixmi, ¬ext); /* pixm full; no text */
pixDestroy(&pixmi);
scaleratio = textscale / imagescale;
}
}
if (pixGetDepth(pixs) == 1) { /* render tiff g4 */
pixb = pixClone(pixs);
pixc = NULL;
}
else {
pixt = pixConvertTo8Or32(pixs, 0, 0); /* this can be a clone of pixs */
/* Get the binary text mask. Note that pixg cannot be a
* clone of pixs, because it may be altered by pixSetMasked(). */
pixb = NULL;
if (notext == FALSE) {
d = pixGetDepth(pixt);
if (d == 8)
pixg = pixCopy(NULL, pixt);
else /* d == 32 */
pixg = pixConvertRGBToLuminance(pixt);
if (pixm) /* clear out the image parts */
pixSetMasked(pixg, pixm, 255);
if (textscale == 1.0)
pixsc = pixClone(pixg);
else if (textscale >= 0.7)
pixsc = pixScaleGrayLI(pixg, textscale, textscale);
else
pixsc = pixScaleAreaMap(pixg, textscale, textscale);
pixb = pixThresholdToBinary(pixsc, threshold);
pixDestroy(&pixg);
pixDestroy(&pixsc);
}
/* Get the scaled image region */
pixc = NULL;
if (pixm) {
if (imagescale == 1.0)
pixsc = pixClone(pixt); /* can possibly be a clone of pixs */
else
pixsc = pixScale(pixt, imagescale, imagescale);
/* If pixm is not full, clear the pixels in pixsc
* corresponding to bg in pixm, where there can be text
* that is written through the mask pixb. Note that
* we could skip this and use pixsc directly in
* pixWriteMixedToPS(); however, clearing these
* non-image regions to a white background will reduce
* the size of pixc (relative to pixsc), and hence
* reduce the size of the PS file that is generated.
* Use a copy so that we don't accidentally alter pixs. */
if (notext == FALSE) {
pixmis = pixScale(pixm, imagescale, imagescale);
pixmi = pixInvert(NULL, pixmis);
val = (d == 8) ? 0xff : 0xffffff00;
pixc = pixCopy(NULL, pixsc);
pixSetMasked(pixc, pixmi, val); /* clear non-image part */
pixDestroy(&pixmis);
pixDestroy(&pixmi);
}
else
pixc = pixClone(pixsc);
pixDestroy(&pixsc);
}
pixDestroy(&pixt);
}
ret = pixWriteMixedToPS(pixb, pixc, scaleratio, pageno, fileout);
pixDestroy(&pixb);
pixDestroy(&pixc);
return ret;
}
l_int32 main(int argc,
char **argv)
{
l_int32 irval, igval, ibval;
l_float32 rval, gval, bval, fract, fgfract;
L_BMF *bmf;
BOX *box;
BOXA *boxa;
FPIX *fpix;
PIX *pixs, *pix1, *pix2, *pix3, *pix4, *pix5, *pix6, *pix7;
PIX *pix8, *pix9, *pix10, *pix11, *pix12, *pix13, *pix14, *pix15;
PIXA *pixa;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
pixa = pixaCreate(0);
pixs = pixRead("breviar38.150.jpg");
/* pixs = pixRead("breviar32.150.jpg"); */
pixaAddPix(pixa, pixs, L_CLONE);
regTestWritePixAndCheck(rp, pixs, IFF_JFIF_JPEG); /* 0 */
pixDisplayWithTitle(pixs, 0, 0, "Input image", rp->display);
/* Extract the blue component, which is small in all the text
* regions, including in the highlight color region */
pix1 = pixGetRGBComponent(pixs, COLOR_BLUE);
pixaAddPix(pixa, pix1, L_CLONE);
regTestWritePixAndCheck(rp, pix1, IFF_JFIF_JPEG); /* 1 */
pixDisplayWithTitle(pix1, 200, 0, "Blue component", rp->display);
/* Do a background normalization, with the background set to
* approximately 200 */
pix2 = pixBackgroundNormSimple(pix1, NULL, NULL);
pixaAddPix(pixa, pix2, L_COPY);
regTestWritePixAndCheck(rp, pix2, IFF_JFIF_JPEG); /* 2 */
pixDisplayWithTitle(pix2, 400, 0, "BG normalized to 200", rp->display);
/* Do a linear transform on the gray pixels, with 50 going to
* black and 160 going to white. 50 is sufficiently low to
* make both the red and black print quite dark. Quantize
* to a few equally spaced gray levels. This is the image
* to which highlight color will be applied. */
pixGammaTRC(pix2, pix2, 1.0, 50, 160);
pix3 = pixThresholdOn8bpp(pix2, 7, 1);
pixaAddPix(pixa, pix3, L_CLONE);
regTestWritePixAndCheck(rp, pix3, IFF_JFIF_JPEG); /* 3 */
pixDisplayWithTitle(pix3, 600, 0, "Basic quantized with white bg",
rp->display);
/* Identify the regions of red text. First, make a mask
* consisting of all pixels such that (R-B)/B is larger
* than 2.0. This will have all the red, plus a lot of
* the dark pixels. */
fpix = pixComponentFunction(pixs, 1.0, 0.0, -1.0, 0.0, 0.0, 1.0);
pix4 = fpixThresholdToPix(fpix, 2.0);
pixInvert(pix4, pix4); /* red plus some dark text */
pixaAddPix(pixa, pix4, L_CLONE);
regTestWritePixAndCheck(rp, pix4, IFF_PNG); /* 4 */
pixDisplayWithTitle(pix4, 800, 0, "Red plus dark pixels", rp->display);
/* Make a mask consisting of all the red and background pixels */
pix5 = pixGetRGBComponent(pixs, COLOR_RED);
pix6 = pixThresholdToBinary(pix5, 128);
pixInvert(pix6, pix6); /* red plus background (white) */
/* Intersect the two masks to get a mask consisting of pixels
* that are almost certainly red. This is the seed. */
pix7 = pixAnd(NULL, pix4, pix6); /* red only (seed) */
pixaAddPix(pixa, pix7, L_COPY);
regTestWritePixAndCheck(rp, pix7, IFF_PNG); /* 5 */
pixDisplayWithTitle(pix7, 0, 600, "Seed for red color", rp->display);
/* Make the clipping mask by thresholding the image with
* the background cleaned to white. */
pix8 = pixThresholdToBinary(pix2, 230); /* mask */
pixaAddPix(pixa, pix8, L_CLONE);
regTestWritePixAndCheck(rp, pix8, IFF_PNG); /* 6 */
pixDisplayWithTitle(pix8, 200, 600, "Clipping mask for red components",
rp->display);
/* Fill into the mask from the seed */
pixSeedfillBinary(pix7, pix7, pix8, 8); /* filled: red plus touching */
regTestWritePixAndCheck(rp, pix7, IFF_PNG); /* 7 */
pixDisplayWithTitle(pix7, 400, 600, "Red component mask filled",
rp->display);
/* Remove long horizontal and vertical lines from the filled result */
pix9 = pixMorphSequence(pix7, "o40.1", 0);
pixSubtract(pix7, pix7, pix9); /* remove long horizontal lines */
pixDestroy(&pix9);
pix9 = pixMorphSequence(pix7, "o1.40", 0);
pixSubtract(pix7, pix7, pix9); /* remove long vertical lines */
/* Close the regions to be colored */
pix10 = pixMorphSequence(pix7, "c5.1", 0);
pixaAddPix(pixa, pix10, L_CLONE);
regTestWritePixAndCheck(rp, pix10, IFF_PNG); /* 8 */
pixDisplayWithTitle(pix10, 600, 600,
"Components defining regions allowing coloring",
rp->display);
/* Sanity check on amount to be colored. Only accept images
* with less than 10% of all the pixels with highlight color */
pixForegroundFraction(pix10, &fgfract);
if (fgfract >= 0.10) {
L_INFO("too much highlighting: fract = %6.3f; removing it\n",
rp->testname, fgfract);
pixClearAll(pix10);
pixSetPixel(pix10, 0, 0, 1);
}
/* Get the bounding boxes of the regions to be colored */
boxa = pixConnCompBB(pix10, 8);
/* Get a color to paint that is representative of the
* actual highlight color in the image. Scale each
* color component up from the average by an amount necessary
* to saturate the red. Then divide the green and
* blue components by 2.0. */
pixGetAverageMaskedRGB(pixs, pix7, 0, 0, 1, L_MEAN_ABSVAL,
&rval, &gval, &bval);
fract = 255.0 / rval;
irval = lept_roundftoi(fract * rval);
igval = lept_roundftoi(fract * gval / 2.0);
ibval = lept_roundftoi(fract * bval / 2.0);
fprintf(stderr, "(r,g,b) = (%d,%d,%d)\n", irval, igval, ibval);
/* Color the quantized gray version in the selected regions */
pix11 = pixColorGrayRegions(pix3, boxa, L_PAINT_DARK, 220, irval,
igval, ibval);
pixaAddPix(pixa, pix11, L_CLONE);
regTestWritePixAndCheck(rp, pix11, IFF_PNG); /* 9 */
pixDisplayWithTitle(pix11, 800, 600, "Final colored result", rp->display);
pixaAddPix(pixa, pixs, L_CLONE);
/* Test colorization on gray and cmapped gray */
pix12 = pixColorGrayRegions(pix2, boxa, L_PAINT_DARK, 220, 0, 255, 0);
pixaAddPix(pixa, pix12, L_CLONE);
regTestWritePixAndCheck(rp, pix12, IFF_PNG); /* 10 */
pixDisplayWithTitle(pix12, 900, 600, "Colorizing boxa gray", rp->display);
box = boxCreate(200, 200, 250, 350);
pix13 = pixCopy(NULL, pix2);
pixColorGray(pix13, box, L_PAINT_DARK, 220, 0, 0, 255);
pixaAddPix(pixa, pix13, L_CLONE);
regTestWritePixAndCheck(rp, pix13, IFF_PNG); /* 11 */
pixDisplayWithTitle(pix13, 1000, 600, "Colorizing box gray", rp->display);
pix14 = pixThresholdTo4bpp(pix2, 6, 1);
pix15 = pixColorGrayRegions(pix14, boxa, L_PAINT_DARK, 220, 0, 0, 255);
pixaAddPix(pixa, pix15, L_CLONE);
regTestWritePixAndCheck(rp, pix15, IFF_PNG); /* 12 */
pixDisplayWithTitle(pix15, 1100, 600, "Colorizing boxa cmap", rp->display);
pixColorGrayCmap(pix14, box, L_PAINT_DARK, 0, 255, 255);
pixaAddPix(pixa, pix14, L_CLONE);
regTestWritePixAndCheck(rp, pix14, IFF_PNG); /* 13 */
pixDisplayWithTitle(pix14, 1200, 600, "Colorizing box cmap", rp->display);
boxDestroy(&box);
/* Generate a pdf of the intermediate results */
lept_mkdir("lept");
L_INFO("Writing to /tmp/lept/colorize.pdf\n", rp->testname);
pixaConvertToPdf(pixa, 90, 1.0, 0, 0, "Colorizing highlighted text",
"/tmp/lept/colorize.pdf");
pixaDestroy(&pixa);
fpixDestroy(&fpix);
boxDestroy(&box);
boxaDestroy(&boxa);
pixDestroy(&pixs);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
pixDestroy(&pix4);
pixDestroy(&pix5);
pixDestroy(&pix6);
pixDestroy(&pix7);
pixDestroy(&pix8);
pixDestroy(&pix9);
pixDestroy(&pix10);
pixDestroy(&pix11);
pixDestroy(&pix12);
pixDestroy(&pix13);
pixDestroy(&pix14);
pixDestroy(&pix15);
/* Test the color detector */
pixa = pixaCreate(7);
bmf = bmfCreate("./fonts", 4);
pix1 = TestForRedColor(rp, "brev06.75.jpg", 1, bmf); /* 14 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = TestForRedColor(rp, "brev10.75.jpg", 0, bmf); /* 15 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = TestForRedColor(rp, "brev14.75.jpg", 1, bmf); /* 16 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = TestForRedColor(rp, "brev20.75.jpg", 1, bmf); /* 17 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = TestForRedColor(rp, "brev36.75.jpg", 0, bmf); /* 18 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = TestForRedColor(rp, "brev53.75.jpg", 1, bmf); /* 19 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = TestForRedColor(rp, "brev56.75.jpg", 1, bmf); /* 20 */
pixaAddPix(pixa, pix1, L_INSERT);
/* Generate a pdf of the color detector results */
L_INFO("Writing to /tmp/lept/colordetect.pdf\n", rp->testname);
pixaConvertToPdf(pixa, 45, 1.0, 0, 0, "Color detection",
"/tmp/lept/colordetect.pdf");
pixaDestroy(&pixa);
bmfDestroy(&bmf);
return regTestCleanup(rp);
}
/*!
* pixOtsuAdaptiveThreshold()
*
* Input: pixs (8 bpp)
* sx, sy (desired tile dimensions; actual size may vary)
* smoothx, smoothy (half-width of convolution kernel applied to
* threshold array: use 0 for no smoothing)
* scorefract (fraction of the max Otsu score; typ. 0.1;
* use 0.0 for standard Otsu)
* &pixth (<optional return> array of threshold values
* found for each tile)
* &pixd (<optional return> thresholded input pixs, based on
* the threshold array)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) The Otsu method finds a single global threshold for an image.
* This function allows a locally adapted threshold to be
* found for each tile into which the image is broken up.
* (2) The array of threshold values, one for each tile, constitutes
* a highly downscaled image. This array is optionally
* smoothed using a convolution. The full width and height of the
* convolution kernel are (2 * @smoothx + 1) and (2 * @smoothy + 1).
* (3) The minimum tile dimension allowed is 16. If such small
* tiles are used, it is recommended to use smoothing, because
* without smoothing, each small tile determines the splitting
* threshold independently. A tile that is entirely in the
* image bg will then hallucinate fg, resulting in a very noisy
* binarization. The smoothing should be large enough that no
* tile is only influenced by one type (fg or bg) of pixels,
* because it will force a split of its pixels.
* (4) To get a single global threshold for the entire image, use
* input values of @sx and @sy that are larger than the image.
* For this situation, the smoothing parameters are ignored.
* (5) The threshold values partition the image pixels into two classes:
* one whose values are less than the threshold and another
* whose values are greater than or equal to the threshold.
* This is the same use of 'threshold' as in pixThresholdToBinary().
* (6) The scorefract is the fraction of the maximum Otsu score, which
* is used to determine the range over which the histogram minimum
* is searched. See numaSplitDistribution() for details on the
* underlying method of choosing a threshold.
* (7) This uses enables a modified version of the Otsu criterion for
* splitting the distribution of pixels in each tile into a
* fg and bg part. The modification consists of searching for
* a minimum in the histogram over a range of pixel values where
* the Otsu score is within a defined fraction, @scorefract,
* of the max score. To get the original Otsu algorithm, set
* @scorefract == 0.
*/
l_int32
pixOtsuAdaptiveThreshold(PIX *pixs,
l_int32 sx,
l_int32 sy,
l_int32 smoothx,
l_int32 smoothy,
l_float32 scorefract,
PIX **ppixth,
PIX **ppixd)
{
l_int32 w, h, nx, ny, i, j, thresh;
l_uint32 val;
PIX *pixt, *pixb, *pixthresh, *pixth, *pixd;
PIXTILING *pt;
PROCNAME("pixOtsuAdaptiveThreshold");
if (!ppixth && !ppixd){
return ERROR_INT("neither &pixth nor &pixd defined", procName, 1);
LOGE("neither &pixth nor &pixd defined");
}
if (ppixth) *ppixth = NULL;
if (ppixd) *ppixd = NULL;
if (!pixs || pixGetDepth(pixs) != 8){
return ERROR_INT("pixs not defined or not 8 bpp", procName, 1);
LOGE("pixs not defined or not 8 bpp");
}
if (sx < 16 || sy < 16){
return ERROR_INT("sx and sy must be >= 16", procName, 1);
LOGE("sx and sy must be >= 16");
}
/* Compute the threshold array for the tiles */
pixGetDimensions(pixs, &w, &h, NULL);
nx = L_MAX(1, w / sx);
ny = L_MAX(1, h / sy);
smoothx = L_MIN(smoothx, (nx - 1) / 2);
smoothy = L_MIN(smoothy, (ny - 1) / 2);
pt = pixTilingCreate(pixs, nx, ny, 0, 0, 0, 0);
pixthresh = pixCreate(nx, ny, 8);
for (i = 0; i < ny; i++) {
for (j = 0; j < nx; j++) {
pixt = pixTilingGetTile(pt, i, j);
pixSplitDistributionFgBg(pixt, scorefract, 1, &thresh,
NULL, NULL, 0);
pixSetPixel(pixthresh, j, i, thresh); /* see note (4) */
pixDestroy(&pixt);
}
}
/* Optionally smooth the threshold array */
if (smoothx > 0 || smoothy > 0)
pixth = pixBlockconv(pixthresh, smoothx, smoothy);
else
pixth = pixClone(pixthresh);
pixDestroy(&pixthresh);
/* Optionally apply the threshold array to binarize pixs */
if (ppixd) {
pixd = pixCreate(w, h, 1);
for (i = 0; i < ny; i++) {
for (j = 0; j < nx; j++) {
pixt = pixTilingGetTile(pt, i, j);
pixGetPixel(pixth, j, i, &val);
pixb = pixThresholdToBinary(pixt, val);
pixTilingPaintTile(pixd, i, j, pixb, pt);
pixDestroy(&pixt);
pixDestroy(&pixb);
}
}
*ppixd = pixd;
}
if (ppixth)
*ppixth = pixth;
else
pixDestroy(&pixth);
pixTilingDestroy(&pt);
return 0;
}
int main(int argc,
char **argv)
{
char dilateseq[512], erodeseq[512];
char openseq[512], closeseq[512];
char wtophatseq[512], btophatseq[512];
l_int32 w, h;
PIX *pixs, *pix1, *pix2, *pix3, *pix4, *pix5;
PIXA *pixa;
PIXACC *pacc;
PIXCMAP *cmap;
L_REGPARAMS *rp;
if (regTestSetup(argc, argv, &rp))
return 1;
pixs = pixRead("aneurisms8.jpg");
pixa = pixaCreate(0);
/* =========================================================== */
/* -------- Test gray morph, including interpreter ------------ */
pix1 = pixDilateGray(pixs, WSIZE, HSIZE);
sprintf(dilateseq, "D%d.%d", WSIZE, HSIZE);
pix2 = pixGrayMorphSequence(pixs, dilateseq, 0, 0);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 0 */
regTestComparePix(rp, pix1, pix2); /* 1 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pix1 = pixErodeGray(pixs, WSIZE, HSIZE);
sprintf(erodeseq, "E%d.%d", WSIZE, HSIZE);
pix2 = pixGrayMorphSequence(pixs, erodeseq, 0, 100);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 2 */
regTestComparePix(rp, pix1, pix2); /* 3 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pix1 = pixOpenGray(pixs, WSIZE, HSIZE);
sprintf(openseq, "O%d.%d", WSIZE, HSIZE);
pix2 = pixGrayMorphSequence(pixs, openseq, 0, 200);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 4 */
regTestComparePix(rp, pix1, pix2); /* 5 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pix1 = pixCloseGray(pixs, WSIZE, HSIZE);
sprintf(closeseq, "C%d.%d", WSIZE, HSIZE);
pix2 = pixGrayMorphSequence(pixs, closeseq, 0, 300);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 6 */
regTestComparePix(rp, pix1, pix2); /* 7 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pix1 = pixTophat(pixs, WSIZE, HSIZE, L_TOPHAT_WHITE);
sprintf(wtophatseq, "Tw%d.%d", WSIZE, HSIZE);
pix2 = pixGrayMorphSequence(pixs, wtophatseq, 0, 400);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 8 */
regTestComparePix(rp, pix1, pix2); /* 9 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pix1 = pixTophat(pixs, WSIZE, HSIZE, L_TOPHAT_BLACK);
sprintf(btophatseq, "Tb%d.%d", WSIZE, HSIZE);
pix2 = pixGrayMorphSequence(pixs, btophatseq, 0, 500);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 10 */
regTestComparePix(rp, pix1, pix2); /* 11 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
/* ------------- Test erode/dilate duality -------------- */
pix1 = pixDilateGray(pixs, WSIZE, HSIZE);
pix2 = pixInvert(NULL, pixs);
pix3 = pixErodeGray(pix2, WSIZE, HSIZE);
pixInvert(pix3, pix3);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 12 */
regTestComparePix(rp, pix1, pix3); /* 13 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* ------------- Test open/close duality -------------- */
pix1 = pixOpenGray(pixs, WSIZE, HSIZE);
pix2 = pixInvert(NULL, pixs);
pix3 = pixCloseGray(pix2, WSIZE, HSIZE);
pixInvert(pix3, pix3);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 14 */
regTestComparePix(rp, pix1, pix3); /* 15 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* ------------- Test tophat duality -------------- */
pix1 = pixTophat(pixs, WSIZE, HSIZE, L_TOPHAT_WHITE);
pix2 = pixInvert(NULL, pixs);
pix3 = pixTophat(pix2, WSIZE, HSIZE, L_TOPHAT_BLACK);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 16 */
regTestComparePix(rp, pix1, pix3); /* 17 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pixDestroy(&pix3);
pix1 = pixGrayMorphSequence(pixs, "Tw9.5", 0, 100);
pix2 = pixInvert(NULL, pixs);
pix3 = pixGrayMorphSequence(pix2, "Tb9.5", 0, 300);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 18 */
regTestComparePix(rp, pix1, pix3); /* 19 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* ------------- Test opening/closing for large sels -------------- */
pix1 = pixGrayMorphSequence(pixs,
"C9.9 + C19.19 + C29.29 + C39.39 + C49.49", 0, 100);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 20 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = pixGrayMorphSequence(pixs,
"O9.9 + O19.19 + O29.29 + O39.39 + O49.49", 0, 400);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 21 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = pixaDisplayTiledInColumns(pixa, 4, 1.0, 20, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 22 */
pixDisplayWithTitle(pix1, 0, 0, NULL, rp->display);
pixaDestroy(&pixa);
pixDestroy(&pix1);
/* =========================================================== */
pixa = pixaCreate(0);
/* ---------- Closing plus white tophat result ------------ *
* Parameters: wsize, hsize = 9, 29 *
* ---------------------------------------------------------*/
pix1 = pixCloseGray(pixs, 9, 9);
pix2 = pixTophat(pix1, 9, 9, L_TOPHAT_WHITE);
pix3 = pixGrayMorphSequence(pixs, "C9.9 + TW9.9", 0, 0);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 23 */
regTestComparePix(rp, pix2, pix3); /* 24 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = pixMaxDynamicRange(pix2, L_LINEAR_SCALE);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 25 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pixDestroy(&pix3);
pix1 = pixCloseGray(pixs, 29, 29);
pix2 = pixTophat(pix1, 29, 29, L_TOPHAT_WHITE);
pix3 = pixGrayMorphSequence(pixs, "C29.29 + Tw29.29", 0, 0);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 26 */
regTestComparePix(rp, pix2, pix3); /* 27 */
pixaAddPix(pixa, pix1, L_INSERT);
pix1 = pixMaxDynamicRange(pix2, L_LINEAR_SCALE);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 28 */
pixaAddPix(pixa, pix1, L_INSERT);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* --------- hdome with parameter height = 100 ------------*/
pix1 = pixHDome(pixs, 100, 4);
pix2 = pixMaxDynamicRange(pix1, L_LINEAR_SCALE);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 29 */
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 30 */
pixaAddPix(pixa, pix1, L_INSERT);
pixaAddPix(pixa, pix2, L_INSERT);
/* ----- Contrast enhancement with morph parameters 9, 9 -------*/
pixGetDimensions(pixs, &w, &h, NULL);
pix1 = pixInitAccumulate(w, h, 0x8000);
pixAccumulate(pix1, pixs, L_ARITH_ADD);
pixMultConstAccumulate(pix1, 3., 0x8000);
pix2 = pixOpenGray(pixs, 9, 9);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 31 */
pixaAddPix(pixa, pix2, L_INSERT);
pixAccumulate(pix1, pix2, L_ARITH_SUBTRACT);
pix2 = pixCloseGray(pixs, 9, 9);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 32 */
pixaAddPix(pixa, pix2, L_INSERT);
pixAccumulate(pix1, pix2, L_ARITH_SUBTRACT);
pix2 = pixFinalAccumulate(pix1, 0x8000, 8);
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 33 */
pixaAddPix(pixa, pix2, L_INSERT);
pixDestroy(&pix1);
/* Do the same thing with the Pixacc */
pacc = pixaccCreate(w, h, 1);
pixaccAdd(pacc, pixs);
pixaccMultConst(pacc, 3.);
pix1 = pixOpenGray(pixs, 9, 9);
pixaccSubtract(pacc, pix1);
pixDestroy(&pix1);
pix1 = pixCloseGray(pixs, 9, 9);
pixaccSubtract(pacc, pix1);
pixDestroy(&pix1);
pix1 = pixaccFinal(pacc, 8);
pixaccDestroy(&pacc);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 34 */
pixaAddPix(pixa, pix1, L_INSERT);
regTestComparePix(rp, pix1, pix2); /* 35 */
pix1 = pixaDisplayTiledInColumns(pixa, 4, 1.0, 20, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 36 */
pixDisplayWithTitle(pix1, 1100, 0, NULL, rp->display);
pixaDestroy(&pixa);
pixDestroy(&pix1);
pixDestroy(&pixs);
/* =========================================================== */
pixa = pixaCreate(0);
/* ---- Tophat result on feynman stamp, to extract diagrams ----- */
pixs = pixRead("feynman-stamp.jpg");
pixGetDimensions(pixs, &w, &h, NULL);
/* Make output image to hold five intermediate images */
pix1 = pixCreate(5 * w + 18, h + 6, 32); /* composite output image */
pixSetAllArbitrary(pix1, 0x0000ff00); /* set to blue */
/* Paste in the input image */
pix2 = pixRemoveColormap(pixs, REMOVE_CMAP_TO_FULL_COLOR);
pixRasterop(pix1, 3, 3, w, h, PIX_SRC, pix2, 0, 0); /* 1st one */
regTestWritePixAndCheck(rp, pix2, IFF_PNG); /* 37 */
pixaAddPix(pixa, pix2, L_INSERT);
/* Paste in the grayscale version */
cmap = pixGetColormap(pixs);
if (cmap)
pix2 = pixRemoveColormap(pixs, REMOVE_CMAP_TO_GRAYSCALE);
else
pix2 = pixConvertRGBToGray(pixs, 0.33, 0.34, 0.33);
pix3 = pixConvertTo32(pix2); /* 8 --> 32 bpp */
pixRasterop(pix1, w + 6, 3, w, h, PIX_SRC, pix3, 0, 0); /* 2nd one */
regTestWritePixAndCheck(rp, pix3, IFF_PNG); /* 38 */
pixaAddPix(pixa, pix3, L_INSERT);
/* Paste in a log dynamic range scaled version of the white tophat */
pix3 = pixTophat(pix2, 3, 3, L_TOPHAT_WHITE);
pix4 = pixMaxDynamicRange(pix3, L_LOG_SCALE);
pix5 = pixConvertTo32(pix4);
pixRasterop(pix1, 2 * w + 9, 3, w, h, PIX_SRC, pix5, 0, 0); /* 3rd */
regTestWritePixAndCheck(rp, pix5, IFF_PNG); /* 39 */
pixaAddPix(pixa, pix5, L_INSERT);
pixDestroy(&pix2);
pixDestroy(&pix4);
/* Stretch the range and threshold to binary; paste it in */
pix2 = pixGammaTRC(NULL, pix3, 1.0, 0, 80);
pix4 = pixThresholdToBinary(pix2, 70);
pix5 = pixConvertTo32(pix4);
pixRasterop(pix1, 3 * w + 12, 3, w, h, PIX_SRC, pix5, 0, 0); /* 4th */
regTestWritePixAndCheck(rp, pix5, IFF_PNG); /* 40 */
pixaAddPix(pixa, pix5, L_INSERT);
pixDestroy(&pix2);
pixDestroy(&pix3);
/* Invert; this is the final result */
pixInvert(pix4, pix4);
pix5 = pixConvertTo32(pix4);
pixRasterop(pix1, 4 * w + 15, 3, w, h, PIX_SRC, pix5, 0, 0); /* 5th */
regTestWritePixAndCheck(rp, pix5, IFF_PNG); /* 41 */
pixaAddPix(pixa, pix5, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix4);
pix1 = pixaDisplayTiledInRows(pixa, 32, 1700, 1.0, 0, 20, 2);
regTestWritePixAndCheck(rp, pix1, IFF_PNG); /* 42 */
pixDisplayWithTitle(pix1, 0, 800, NULL, rp->display);
pixaDestroy(&pixa);
pixDestroy(&pix1);
pixDestroy(&pixs);
return regTestCleanup(rp);
}
/*!
* pixMaskedThreshOnBackgroundNorm()
*
* Input: pixs (8 bpp grayscale; not colormapped)
* pixim (<optional> 1 bpp 'image' mask; can be null)
* sx, sy (tile size in pixels)
* thresh (threshold for determining foreground)
* mincount (min threshold on counts in a tile)
* smoothx (half-width of block convolution kernel width)
* smoothy (half-width of block convolution kernel height)
* scorefract (fraction of the max Otsu score; typ. ~ 0.1)
* &thresh (<optional return> threshold value that was
* used on the normalized image)
* Return: pixd (1 bpp thresholded image), or null on error
*
* Notes:
* (1) This begins with a standard background normalization.
* Additionally, there is a flexible background norm, that
* will adapt to a rapidly varying background, and this
* puts white pixels in the background near regions with
* significant foreground. The white pixels are turned into
* a 1 bpp selection mask by binarization followed by dilation.
* Otsu thresholding is performed on the input image to get an
* estimate of the threshold in the non-mask regions.
* The background normalized image is thresholded with two
* different values, and the result is combined using
* the selection mask.
* (2) Note that the numbers 255 (for bgval target) and 190 (for
* thresholding on pixn) are tied together, and explicitly
* defined in this function.
* (3) See pixBackgroundNorm() for meaning and typical values
* of input parameters. For a start, you can try:
* sx, sy = 10, 15
* thresh = 100
* mincount = 50
* smoothx, smoothy = 2
*/
PIX *
pixMaskedThreshOnBackgroundNorm(PIX *pixs,
PIX *pixim,
l_int32 sx,
l_int32 sy,
l_int32 thresh,
l_int32 mincount,
l_int32 smoothx,
l_int32 smoothy,
l_float32 scorefract,
l_int32 *pthresh)
{
l_int32 w, h;
l_uint32 val;
PIX *pixn, *pixm, *pixd, *pixt1, *pixt2, *pixt3, *pixt4;
PROCNAME("pixMaskedThreshOnBackgroundNorm");
if (pthresh) *pthresh = 0;
if (!pixs || pixGetDepth(pixs) != 8)
return (PIX *)ERROR_PTR("pixs undefined or not 8 bpp", procName, NULL);
if (pixGetColormap(pixs))
return (PIX *)ERROR_PTR("pixs is colormapped", procName, NULL);
if (sx < 4 || sy < 4)
return (PIX *)ERROR_PTR("sx and sy must be >= 4", procName, NULL);
if (mincount > sx * sy) {
L_WARNING("mincount too large for tile size\n", procName);
mincount = (sx * sy) / 3;
}
/* Standard background normalization */
pixn = pixBackgroundNorm(pixs, pixim, NULL, sx, sy, thresh,
mincount, 255, smoothx, smoothy);
if (!pixn)
return (PIX *)ERROR_PTR("pixn not made", procName, NULL);
/* Special background normalization for adaptation to quickly
* varying background. Threshold on the very light parts,
* which tend to be near significant edges, and dilate to
* form a mask over regions that are typically text. The
* dilation size is chosen to cover the text completely,
* except for very thick fonts. */
pixt1 = pixBackgroundNormFlex(pixs, 7, 7, 1, 1, 20);
pixt2 = pixThresholdToBinary(pixt1, 240);
pixInvert(pixt2, pixt2);
pixm = pixMorphSequence(pixt2, "d21.21", 0);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
/* Use Otsu to get a global threshold estimate for the image,
* which is stored as a single pixel in pixt3. */
pixGetDimensions(pixs, &w, &h, NULL);
pixOtsuAdaptiveThreshold(pixs, w, h, 0, 0, scorefract, &pixt3, NULL);
if (pixt3 && pthresh) {
pixGetPixel(pixt3, 0, 0, &val);
*pthresh = val;
}
pixDestroy(&pixt3);
/* Threshold the background normalized images differentially,
* using a high value correlated with the background normalization
* for the part of the image under the mask (i.e., near the
* darker, thicker foreground), and a value that depends on the Otsu
* threshold for the rest of the image. This gives a solid
* (high) thresholding for the foreground parts of the image,
* while allowing the background and light foreground to be
* reasonably well cleaned using a threshold adapted to the
* input image. */
pixd = pixThresholdToBinary(pixn, val + 30); /* for bg and light fg */
pixt4 = pixThresholdToBinary(pixn, 190); /* for heavier fg */
pixCombineMasked(pixd, pixt4, pixm);
pixDestroy(&pixt4);
pixDestroy(&pixm);
pixDestroy(&pixn);
if (!pixd)
return (PIX *)ERROR_PTR("pixd not made", procName, NULL);
else
return pixd;
}
/*!
* convertFilesTo1bpp()
*
* Input: dirin
* substr (<optional> substring filter on filenames; can be NULL)
* upscaling (1, 2 or 4; only for input color or grayscale)
* thresh (global threshold for binarization; use 0 for default)
* firstpage
* npages (use 0 to do all from @firstpage to the end)
* dirout
* outformat (IFF_PNG, IFF_TIFF_G4)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) Images are sorted lexicographically, and the names in the
* output directory are retained except for the extension.
*/
l_int32
convertFilesTo1bpp(const char *dirin,
const char *substr,
l_int32 upscaling,
l_int32 thresh,
l_int32 firstpage,
l_int32 npages,
const char *dirout,
l_int32 outformat) {
l_int32 i, nfiles;
char buf[512];
char *fname, *tail, *basename;
PIX *pixs, *pixg1, *pixg2, *pixb;
SARRAY *safiles;
PROCNAME("convertFilesTo1bpp");
if (!dirin)
return ERROR_INT("dirin", procName, 1);
if (!dirout)
return ERROR_INT("dirout", procName, 1);
if (upscaling != 1 && upscaling != 2 && upscaling != 4)
return ERROR_INT("invalid upscaling factor", procName, 1);
if (thresh <= 0) thresh = 180;
if (firstpage < 0) firstpage = 0;
if (npages < 0) npages = 0;
if (outformat != IFF_TIFF_G4)
outformat = IFF_PNG;
safiles = getSortedPathnamesInDirectory(dirin, substr, firstpage, npages);
if (!safiles)
return ERROR_INT("safiles not made", procName, 1);
if ((nfiles = sarrayGetCount(safiles)) == 0) {
sarrayDestroy(&safiles);
return ERROR_INT("no matching files in the directory", procName, 1);
}
for (i = 0; i < nfiles; i++) {
fname = sarrayGetString(safiles, i, L_NOCOPY);
if ((pixs = pixRead(fname)) == NULL) {
L_WARNING("Couldn't read file %s\n", procName, fname);
continue;
}
if (pixGetDepth(pixs) == 32)
pixg1 = pixConvertRGBToLuminance(pixs);
else
pixg1 = pixClone(pixs);
pixg2 = pixRemoveColormap(pixg1, REMOVE_CMAP_TO_GRAYSCALE);
if (pixGetDepth(pixg2) == 1) {
pixb = pixClone(pixg2);
} else {
if (upscaling == 1)
pixb = pixThresholdToBinary(pixg2, thresh);
else if (upscaling == 2)
pixb = pixScaleGray2xLIThresh(pixg2, thresh);
else /* upscaling == 4 */
pixb = pixScaleGray4xLIThresh(pixg2, thresh);
}
pixDestroy(&pixs);
pixDestroy(&pixg1);
pixDestroy(&pixg2);
splitPathAtDirectory(fname, NULL, &tail);
splitPathAtExtension(tail, &basename, NULL);
if (outformat == IFF_TIFF_G4) {
snprintf(buf, sizeof(buf), "%s/%s.tif", dirout, basename);
pixWrite(buf, pixb, IFF_TIFF_G4);
} else {
snprintf(buf, sizeof(buf), "%s/%s.png", dirout, basename);
pixWrite(buf, pixb, IFF_PNG);
}
pixDestroy(&pixb);
FREE(tail);
FREE(basename);
}
sarrayDestroy(&safiles);
return 0;
}
char *_process_frame_color_basic(struct lib_hardsubx_ctx *ctx, AVFrame *frame, int width, int height, int index)
{
char *subtitle_text=NULL;
PIX *im;
im = pixCreate(width,height,32);
PIX *hue_im = pixCreate(width,height,32);
int i,j;
for(i=0;i<height;i++)
{
for(j=0;j<width;j++)
{
int p=j*3+i*frame->linesize[0];
int r=frame->data[0][p];
int g=frame->data[0][p+1];
int b=frame->data[0][p+2];
pixSetRGBPixel(im,j,i,r,g,b);
float H,S,V;
rgb_to_hsv((float)r,(float)g,(float)b,&H,&S,&V);
if(abs(H-ctx->hue)<20)
{
pixSetRGBPixel(hue_im,j,i,r,g,b);
}
}
}
PIX *edge_im = pixCreate(width,height,8),*edge_im_2 = pixCreate(width,height,8);
edge_im = pixConvertRGBToGray(im,0.0,0.0,0.0);
edge_im = pixSobelEdgeFilter(edge_im, L_VERTICAL_EDGES);
edge_im = pixDilateGray(edge_im, 21, 1);
edge_im = pixThresholdToBinary(edge_im,50);
PIX *pixd = pixCreate(width,height,1);
pixSauvolaBinarize(pixConvertRGBToGray(hue_im,0.0,0.0,0.0), 15, 0.3, 1, NULL, NULL, NULL, &pixd);
edge_im_2 = pixConvertRGBToGray(hue_im,0.0,0.0,0.0);
edge_im_2 = pixDilateGray(edge_im_2, 5, 5);
PIX *feat_im = pixCreate(width,height,32);
for(i=3*(height/4);i<height;i++)
{
for(j=0;j<width;j++)
{
unsigned int p1,p2,p3,p4;
pixGetPixel(edge_im,j,i,&p1);
pixGetPixel(pixd,j,i,&p2);
// pixGetPixel(hue_im,j,i,&p3);
pixGetPixel(edge_im_2,j,i,&p4);
if(p1==0&&p2==0&&p4>0)//if(p4>0&&p1==0)//if(p2==0&&p1==0&&p3>0)
{
pixSetRGBPixel(feat_im,j,i,255,255,255);
}
}
}
if(ctx->detect_italics)
{
ctx->ocr_mode = HARDSUBX_OCRMODE_WORD;
}
// TESSERACT OCR FOR THE FRAME HERE
switch(ctx->ocr_mode)
{
case HARDSUBX_OCRMODE_WORD:
if(ctx->conf_thresh > 0)
subtitle_text = get_ocr_text_wordwise_threshold(ctx, feat_im, ctx->conf_thresh);
else
subtitle_text = get_ocr_text_wordwise(ctx, feat_im);
break;
case HARDSUBX_OCRMODE_LETTER:
if(ctx->conf_thresh > 0)
subtitle_text = get_ocr_text_letterwise_threshold(ctx, feat_im, ctx->conf_thresh);
else
subtitle_text = get_ocr_text_letterwise(ctx, feat_im);
break;
case HARDSUBX_OCRMODE_FRAME:
if(ctx->conf_thresh > 0)
subtitle_text = get_ocr_text_simple_threshold(ctx, feat_im, ctx->conf_thresh);
else
subtitle_text = get_ocr_text_simple(ctx, feat_im);
break;
default:
fatal(EXIT_MALFORMED_PARAMETER,"Invalid OCR Mode");
}
pixDestroy(&feat_im);
pixDestroy(&im);
pixDestroy(&edge_im);
pixDestroy(&hue_im);
return subtitle_text;
}
int main(int argc,
char **argv) {
char dilateseq[BUF_SIZE], erodeseq[BUF_SIZE];
char openseq[BUF_SIZE], closeseq[BUF_SIZE];
char wtophatseq[BUF_SIZE], btophatseq[BUF_SIZE];
char *filein;
l_int32 w, h, d;
PIX *pixs, *pixt, *pixt2, *pixt3, *pixt3a, *pixt4;
PIX *pixg, *pixd, *pixd1, *pixd2, *pixd3;
PIXACC *pacc;
PIXCMAP *cmap;
static char mainName[] = "graymorph1_reg";
if (argc != 2)
return ERROR_INT(" Syntax: graymorph1_reg filein", mainName, 1);
filein = argv[1];
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
pixGetDimensions(pixs, &w, &h, &d);
if (d != 8)
return ERROR_INT("pixs not 8 bpp", mainName, 1);
/* -------- Test gray morph, including interpreter ------------ */
pixd = pixDilateGray(pixs, WSIZE, HSIZE);
sprintf(dilateseq, "D%d.%d", WSIZE, HSIZE);
pixg = pixGrayMorphSequence(pixs, dilateseq, HORIZ_SEP, 0);
pixCompare(pixd, pixg, "results are the same", "results are different");
pixDestroy(&pixg);
pixDestroy(&pixd);
pixd = pixErodeGray(pixs, WSIZE, HSIZE);
sprintf(erodeseq, "E%d.%d", WSIZE, HSIZE);
pixg = pixGrayMorphSequence(pixs, erodeseq, HORIZ_SEP, 100);
pixCompare(pixd, pixg, "results are the same", "results are different");
pixDestroy(&pixg);
pixDestroy(&pixd);
pixd = pixOpenGray(pixs, WSIZE, HSIZE);
sprintf(openseq, "O%d.%d", WSIZE, HSIZE);
pixg = pixGrayMorphSequence(pixs, openseq, HORIZ_SEP, 200);
pixCompare(pixd, pixg, "results are the same", "results are different");
pixDestroy(&pixg);
pixDestroy(&pixd);
pixd = pixCloseGray(pixs, WSIZE, HSIZE);
sprintf(closeseq, "C%d.%d", WSIZE, HSIZE);
pixg = pixGrayMorphSequence(pixs, closeseq, HORIZ_SEP, 300);
pixCompare(pixd, pixg, "results are the same", "results are different");
pixDestroy(&pixg);
pixDestroy(&pixd);
pixd = pixTophat(pixs, WSIZE, HSIZE, L_TOPHAT_WHITE);
sprintf(wtophatseq, "Tw%d.%d", WSIZE, HSIZE);
pixg = pixGrayMorphSequence(pixs, wtophatseq, HORIZ_SEP, 400);
pixCompare(pixd, pixg, "results are the same", "results are different");
pixDestroy(&pixg);
pixDestroy(&pixd);
pixd = pixTophat(pixs, WSIZE, HSIZE, L_TOPHAT_BLACK);
sprintf(btophatseq, "Tb%d.%d", WSIZE, HSIZE);
pixg = pixGrayMorphSequence(pixs, btophatseq, HORIZ_SEP, 500);
pixCompare(pixd, pixg, "results are the same", "results are different");
pixDestroy(&pixg);
/* ------------- Test erode/dilate duality -------------- */
pixd = pixDilateGray(pixs, WSIZE, HSIZE);
pixInvert(pixs, pixs);
pixd2 = pixErodeGray(pixs, WSIZE, HSIZE);
pixInvert(pixd2, pixd2);
pixCompare(pixd, pixd2, "results are the same", "results are different");
pixDestroy(&pixd);
pixDestroy(&pixd2);
/* ------------- Test open/close duality -------------- */
pixd = pixOpenGray(pixs, WSIZE, HSIZE);
pixInvert(pixs, pixs);
pixd2 = pixCloseGray(pixs, WSIZE, HSIZE);
pixInvert(pixd2, pixd2);
pixCompare(pixd, pixd2, "results are the same", "results are different");
pixDestroy(&pixd);
pixDestroy(&pixd2);
/* ------------- Test tophat duality -------------- */
pixd = pixTophat(pixs, WSIZE, HSIZE, L_TOPHAT_WHITE);
pixInvert(pixs, pixs);
pixd2 = pixTophat(pixs, WSIZE, HSIZE, L_TOPHAT_BLACK);
pixCompare(pixd, pixd2, "Correct: images are duals",
"Error: images are not duals");
pixDestroy(&pixd);
pixDestroy(&pixd2);
pixInvert(pixs, pixs);
pixd = pixGrayMorphSequence(pixs, "Tw9.5", HORIZ_SEP, 100);
pixInvert(pixs, pixs);
pixd2 = pixGrayMorphSequence(pixs, "Tb9.5", HORIZ_SEP, 300);
pixCompare(pixd, pixd2, "Correct: images are duals",
"Error: images are not duals");
pixDestroy(&pixd);
pixDestroy(&pixd2);
/* ------------- Test opening/closing for large sels -------------- */
pixd = pixGrayMorphSequence(pixs,
"C9.9 + C19.19 + C29.29 + C39.39 + C49.49", HORIZ_SEP, 100);
pixDestroy(&pixd);
pixd = pixGrayMorphSequence(pixs,
"O9.9 + O19.19 + O29.29 + O39.39 + O49.49", HORIZ_SEP, 400);
pixDestroy(&pixd);
/* ---------- Closing plus white tophat result ------------ *
* Parameters: wsize, hsize = 9, 29 *
* ---------------------------------------------------------*/
pixd = pixCloseGray(pixs, 9, 9);
pixd1 = pixTophat(pixd, 9, 9, L_TOPHAT_WHITE);
pixd2 = pixGrayMorphSequence(pixs, "C9.9 + TW9.9", HORIZ_SEP, 0);
pixCompare(pixd1, pixd2, "correct: same", "wrong: different");
pixd3 = pixMaxDynamicRange(pixd1, L_LINEAR_SCALE);
pixDisplayWrite(pixd3, 1);
pixDestroy(&pixd);
pixDestroy(&pixd1);
pixDestroy(&pixd2);
pixDestroy(&pixd3);
pixd = pixCloseGray(pixs, 29, 29);
pixd1 = pixTophat(pixd, 29, 29, L_TOPHAT_WHITE);
pixd2 = pixGrayMorphSequence(pixs, "C29.29 + Tw29.29", HORIZ_SEP, 0);
pixCompare(pixd1, pixd2, "correct: same", "wrong: different");
pixd3 = pixMaxDynamicRange(pixd1, L_LINEAR_SCALE);
pixDisplayWrite(pixd3, 1);
pixDestroy(&pixd);
pixDestroy(&pixd1);
pixDestroy(&pixd2);
pixDestroy(&pixd3);
/* --------- hdome with parameter height = 100 ------------*/
pixd = pixHDome(pixs, 100, 4);
pixd2 = pixMaxDynamicRange(pixd, L_LINEAR_SCALE);
pixDisplayWrite(pixd2, 1);
pixDestroy(&pixd2);
/* ----- Contrast enhancement with morph parameters 9, 9 -------*/
pixd1 = pixInitAccumulate(w, h, 0x8000);
pixAccumulate(pixd1, pixs, L_ARITH_ADD);
pixMultConstAccumulate(pixd1, 3., 0x8000);
pixd2 = pixOpenGray(pixs, 9, 9);
pixAccumulate(pixd1, pixd2, L_ARITH_SUBTRACT);
pixDestroy(&pixd2);
pixd2 = pixCloseGray(pixs, 9, 9);
pixAccumulate(pixd1, pixd2, L_ARITH_SUBTRACT);
pixDestroy(&pixd2);
pixd = pixFinalAccumulate(pixd1, 0x8000, 8);
pixDisplayWrite(pixd, 1);
pixDestroy(&pixd1);
/* Do the same thing with the Pixacc */
pacc = pixaccCreate(w, h, 1);
pixaccAdd(pacc, pixs);
pixaccMultConst(pacc, 3.);
pixd1 = pixOpenGray(pixs, 9, 9);
pixaccSubtract(pacc, pixd1);
pixDestroy(&pixd1);
pixd1 = pixCloseGray(pixs, 9, 9);
pixaccSubtract(pacc, pixd1);
pixDestroy(&pixd1);
pixd2 = pixaccFinal(pacc, 8);
pixaccDestroy(&pacc);
pixDisplayWrite(pixd2, 1);
pixCompare(pixd, pixd2, "Correct: same", "Wrong: different");
pixDestroy(&pixd);
pixDestroy(&pixd2);
/* ---- Tophat result on feynman stamp, to extract diagrams ----- */
pixDestroy(&pixs);
pixs = pixRead("feynman-stamp.jpg");
/* Make output image to hold five intermediate images */
w = pixGetWidth(pixs);
h = pixGetHeight(pixs);
pixd = pixCreate(5 * w + 18, h + 6, 32); /* composite output image */
pixSetAllArbitrary(pixd, 0x0000ff00); /* set to blue */
/* Paste in the input image */
pixt = pixRemoveColormap(pixs, REMOVE_CMAP_TO_FULL_COLOR);
pixRasterop(pixd, 3, 3, w, h, PIX_SRC, pixt, 0, 0); /* 1st one */
/* pixWrite("/tmp/junkgray.jpg", pixt, IFF_JFIF_JPEG); */
pixDestroy(&pixt);
/* Paste in the grayscale version */
cmap = pixGetColormap(pixs);
if (cmap)
pixt = pixRemoveColormap(pixs, REMOVE_CMAP_TO_GRAYSCALE);
else
pixt = pixConvertRGBToGray(pixs, 0.33, 0.34, 0.33);
pixt2 = pixConvertTo32(pixt); /* 8 --> 32 bpp */
pixRasterop(pixd, w + 6, 3, w, h, PIX_SRC, pixt2, 0, 0); /* 2nd one */
pixDestroy(&pixt2);
/* Paste in a log dynamic range scaled version of the white tophat */
pixt2 = pixTophat(pixt, 3, 3, L_TOPHAT_WHITE);
pixt3a = pixMaxDynamicRange(pixt2, L_LOG_SCALE);
pixt3 = pixConvertTo32(pixt3a);
pixRasterop(pixd, 2 * w + 9, 3, w, h, PIX_SRC, pixt3, 0, 0); /* 3rd */
/* pixWrite("/tmp/junktophat.jpg", pixt2, IFF_JFIF_JPEG); */
pixDestroy(&pixt3);
pixDestroy(&pixt3a);
pixDestroy(&pixt);
/* Stretch the range and threshold to binary; paste it in */
pixt3a = pixGammaTRC(NULL, pixt2, 1.0, 0, 80);
pixt3 = pixThresholdToBinary(pixt3a, 70);
pixt4 = pixConvertTo32(pixt3);
pixRasterop(pixd, 3 * w + 12, 3, w, h, PIX_SRC, pixt4, 0, 0); /* 4th */
/* pixWrite("/tmp/junkbin.png", pixt3, IFF_PNG); */
pixDestroy(&pixt2);
pixDestroy(&pixt3a);
pixDestroy(&pixt4);
/* Invert; this is the final result */
pixInvert(pixt3, pixt3);
pixt4 = pixConvertTo32(pixt3);
pixRasterop(pixd, 4 * w + 15, 3, w, h, PIX_SRC, pixt4, 0, 0); /* 5th */
pixWrite("/tmp/junkbininvert.png", pixt3, IFF_PNG);
pixDisplayWrite(pixd, 1);
/* pixWrite("/tmp/junkall.jpg", pixd, IFF_JFIF_JPEG); */
pixDestroy(&pixt3);
pixDestroy(&pixt4);
pixDestroy(&pixd);
pixDisplayMultiple("/tmp/display/file*");
pixDestroy(&pixs);
return 0;
}
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);
}