Exemplo n.º 1
0
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);
}
Exemplo n.º 2
0
jlong Java_com_googlecode_leptonica_android_Edge_nativePixSobelEdgeFilter(JNIEnv *env,
                                                                          jclass clazz,
                                                                          jlong nativePix,
                                                                          jint orientFlag) {
  PIX *pixs = (PIX *) nativePix;
  PIX *pixd = pixSobelEdgeFilter(pixs, (l_int32) orientFlag);

  return (jlong) pixd;
}
Exemplo n.º 3
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;
}
Exemplo n.º 4
0
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);
}
Exemplo n.º 5
0
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;
}
Exemplo n.º 6
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;
}