dvb output as transcript

src/cc_encoders_common.c

@@ -5,6 +5,7 @@
 #include "608_spupng.h"
 #include "utility.h"
 #include "xds.h"
+#include "ocr.h"
 
 static const char *sami_header= // TODO: Revise the <!-- comments
 "<SAMI>\n\
@@ -205,7 +206,176 @@ int write_cc_buffer_as_transcript2(struct eia608_screen *data, struct encoder_ct
 	dbg_print(CCX_DMT_608, "- - - - - - - - - - - -\r\n");
 	return wrote_something;
 }
+int write_cc_bitmap_as_transcript(struct cc_subtitle *sub, struct encoder_ctx *context)
+{
+	struct spupng_t *sp = (struct spupng_t *)context->out->spupng_data;
+	int x_pos, y_pos, width, height, i;
+	int x, y, y_off, x_off, ret;
+	uint8_t *pbuf;
+	char *filename;
+	struct cc_bitmap* rect;
+	png_color *palette = NULL;
+	png_byte *alpha = NULL;
+#ifdef ENABLE_OCR
+	char*str = NULL;
+#endif
+	int used;
+	unsigned h1,m1,s1,ms1;
+	unsigned h2,m2,s2,ms2;
+	LLONG start_time, end_time;
+	char timeline[128];
+	int len = 0;
 
+        x_pos = -1;
+        y_pos = -1;
+        width = 0;
+        height = 0;
+
+	if (context->prev_start != -1 && (sub->flags & SUB_EOD_MARKER))
+	{
+		start_time = context->prev_start + subs_delay;
+		end_time = sub->start_time - 1;
+	}
+	else if ( !(sub->flags & SUB_EOD_MARKER))
+	{
+		start_time = sub->start_time + subs_delay;
+		end_time = sub->end_time - 1;
+	}
+
+	if(sub->nb_data == 0 )
+		return 0;
+	rect = sub->data;
+	for(i = 0;i < sub->nb_data;i++)
+	{
+		if(x_pos == -1)
+		{
+			x_pos = rect[i].x;
+			y_pos = rect[i].y;
+			width = rect[i].w;
+			height = rect[i].h;
+		}
+		else
+		{
+			if(x_pos > rect[i].x)
+			{
+				width += (x_pos - rect[i].x);
+				x_pos = rect[i].x;
+			}
+
+                        if (rect[i].y < y_pos)
+                        {
+                                height += (y_pos - rect[i].y);
+                                y_pos = rect[i].y;
+                        }
+
+                        if (rect[i].x + rect[i].w > x_pos + width)
+                        {
+                                width = rect[i].x + rect[i].w - x_pos;
+                        }
+
+                        if (rect[i].y + rect[i].h > y_pos + height)
+                        {
+                                height = rect[i].y + rect[i].h - y_pos;
+                        }
+
+		}
+	}
+	if ( sub->flags & SUB_EOD_MARKER )
+		context->prev_start =  sub->start_time;
+	pbuf = (uint8_t*) malloc(width * height);
+	memset(pbuf, 0x0, width * height);
+
+	for(i = 0;i < sub->nb_data;i++)
+	{
+		x_off = rect[i].x - x_pos;
+		y_off = rect[i].y - y_pos;
+		for (y = 0; y < rect[i].h; y++)
+		{
+			for (x = 0; x < rect[i].w; x++)
+				pbuf[((y + y_off) * width) + x_off + x] = rect[i].data[0][y * rect[i].w + x];
+
+		}
+	}
+	palette = (png_color*) malloc(rect[0].nb_colors * sizeof(png_color));
+	if(!palette)
+	{
+		ret = -1;
+		goto end;
+	}
+        alpha = (png_byte*) malloc(rect[0].nb_colors * sizeof(png_byte));
+        if(!alpha)
+        {
+                ret = -1;
+                goto end;
+        }
+	/* TODO do rectangle, wise one color table should not be used for all rectangle */
+        mapclut_paletee(palette, alpha, (uint32_t *)rect[0].data[1],rect[0].nb_colors);
+	quantize_map(alpha, palette, pbuf, width*height, 3, rect[0].nb_colors);
+#ifdef ENABLE_OCR
+	str = ocr_bitmap(palette,alpha,pbuf,width,height);
+	if(str && str[0])
+	{
+		if (context->prev_start != -1 || !(sub->flags & SUB_EOD_MARKER))
+		{
+			char *token = NULL;
+			token = strtok(str,"\r\n");
+			while (token)
+			{
+
+				if (ccx_options.transcript_settings.showStartTime)
+				{
+					char buf1[80];
+					if (ccx_options.transcript_settings.relativeTimestamp)
+					{
+						millis_to_date(start_time + subs_delay, buf1);
+						fdprintf(context->out->fh, "%s|", buf1);
+					}
+					else
+					{
+						mstotime(start_time + subs_delay, &h1, &m1, &s1, &ms1);
+						time_t start_time_int = (start_time + subs_delay) / 1000;
+						int start_time_dec = (start_time + subs_delay) % 1000;
+						struct tm *start_time_struct = gmtime(&start_time_int);
+						strftime(buf1, sizeof(buf1), "%Y%m%d%H%M%S", start_time_struct);
+						fdprintf(context->out->fh, "%s%c%03d|", buf1,ccx_options.millis_separator,start_time_dec);
+					}
+				}
+
+				if (ccx_options.transcript_settings.showEndTime)
+				{
+					char buf2[80];
+					if (ccx_options.transcript_settings.relativeTimestamp)
+					{
+						millis_to_date(end_time, buf2);
+						fdprintf(context->out->fh, "%s|", buf2);
+					}
+					else
+					{
+						mstotime(get_fts() + subs_delay, &h2, &m2, &s2, &ms2);
+						time_t end_time_int = end_time / 1000;
+						int end_time_dec = end_time % 1000;
+						struct tm *end_time_struct = gmtime(&end_time_int);
+						strftime(buf2, sizeof(buf2), "%Y%m%d%H%M%S", end_time_struct);
+						fdprintf(context->out->fh, "%s%c%03d|", buf2,ccx_options.millis_separator,end_time_dec);
+					}
+				}
+				fdprintf(context->out->fh,"DVB|%s\n",token);
+				token = strtok(NULL,"\r\n");
+
+			}
+
+		}
+	}
+#endif
+
+end:
+	sub->nb_data = 0;
+	freep(&sub->data);
+	freep(&palette);
+	freep(&alpha);
+	return ret;
+
+}
 void try_to_add_end_credits(struct encoder_ctx *context, struct ccx_s_write *out)
 {
 	LLONG window, length, st, end;
@@ -400,7 +570,7 @@ int encode_sub(struct encoder_ctx *context, struct cc_subtitle *sub)
 				try_to_add_start_credits(context, sub->start_time);
 			wrote_something = write_cc_bitmap_as_smptett(sub, context);
 		case CCX_OF_TRANSCRIPT:
-//			wrote_something = write_cc_bitmap_as_transcript(sub, context);
+			wrote_something = write_cc_bitmap_as_transcript(sub, context);
 			break;
 		case CCX_OF_SPUPNG:
 			wrote_something = write_cc_bitmap_as_spupng(sub, context);