3 * Copyright (c) 2009 Stephen Backway
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * PGS subtitle decoder
29 #include "bytestream.h"
30 #include "libavutil/colorspace.h"
31 #include "libavutil/imgutils.h"
32 #include "libavutil/opt.h"
34 #define RGBA(r,g,b,a) (((a) << 24) | ((r) << 16) | ((g) << 8) | (b))
37 PALETTE_SEGMENT = 0x14,
38 PICTURE_SEGMENT = 0x15,
39 PRESENTATION_SEGMENT = 0x16,
40 WINDOW_SEGMENT = 0x17,
41 DISPLAY_SEGMENT = 0x80,
44 typedef struct PGSSubPictureReference {
49 } PGSSubPictureReference;
51 typedef struct PGSSubPresentation {
54 PGSSubPictureReference *objects;
57 typedef struct PGSSubPicture {
61 unsigned int rle_buffer_size, rle_data_len;
62 unsigned int rle_remaining_len;
65 typedef struct PGSSubContext {
67 PGSSubPresentation presentation;
69 PGSSubPicture pictures[UINT16_MAX];
74 static av_cold int init_decoder(AVCodecContext *avctx)
76 avctx->pix_fmt = PIX_FMT_PAL8;
81 static av_cold int close_decoder(AVCodecContext *avctx)
85 PGSSubContext *ctx = avctx->priv_data;
87 av_freep(&ctx->presentation.objects);
88 ctx->presentation.object_count = 0;
90 for (picture = 0; picture < UINT16_MAX; ++picture) {
91 av_freep(&ctx->pictures[picture].rle);
92 ctx->pictures[picture].rle_buffer_size = 0;
99 * Decode the RLE data.
101 * The subtitle is stored as an Run Length Encoded image.
103 * @param avctx contains the current codec context
104 * @param sub pointer to the processed subtitle data
105 * @param buf pointer to the RLE data to process
106 * @param buf_size size of the RLE data to process
108 static int decode_rle(AVCodecContext *avctx, AVSubtitle *sub, int rect,
109 const uint8_t *buf, unsigned int buf_size)
111 const uint8_t *rle_bitmap_end;
112 int pixel_count, line_count;
114 rle_bitmap_end = buf + buf_size;
116 sub->rects[rect]->pict.data[0] = av_malloc(sub->rects[rect]->w * sub->rects[rect]->h);
118 if (!sub->rects[rect]->pict.data[0])
124 while (buf < rle_bitmap_end && line_count < sub->rects[rect]->h) {
125 uint8_t flags, color;
128 color = bytestream_get_byte(&buf);
132 flags = bytestream_get_byte(&buf);
135 run = (run << 8) + bytestream_get_byte(&buf);
136 color = flags & 0x80 ? bytestream_get_byte(&buf) : 0;
139 if (run > 0 && pixel_count + run <= sub->rects[rect]->w * sub->rects[rect]->h) {
140 memset(sub->rects[rect]->pict.data[0] + pixel_count, color, run);
144 * New Line. Check if correct pixels decoded, if not display warning
145 * and adjust bitmap pointer to correct new line position.
147 if (pixel_count % sub->rects[rect]->w > 0)
148 av_log(avctx, AV_LOG_ERROR, "Decoded %d pixels, when line should be %d pixels\n",
149 pixel_count % sub->rects[rect]->w, sub->rects[rect]->w);
154 if (pixel_count < sub->rects[rect]->w * sub->rects[rect]->h) {
155 av_log(avctx, AV_LOG_ERROR, "Insufficient RLE data for subtitle\n");
159 av_dlog(avctx, "Pixel Count = %d, Area = %d\n", pixel_count, sub->rects[rect]->w * sub->rects[rect]->h);
165 * Parse the picture segment packet.
167 * The picture segment contains details on the sequence id,
168 * width, height and Run Length Encoded (RLE) bitmap data.
170 * @param avctx contains the current codec context
171 * @param buf pointer to the packet to process
172 * @param buf_size size of packet to process
173 * @todo TODO: Enable support for RLE data over multiple packets
175 static int parse_picture_segment(AVCodecContext *avctx,
176 const uint8_t *buf, int buf_size)
178 PGSSubContext *ctx = avctx->priv_data;
180 uint8_t sequence_desc;
181 unsigned int rle_bitmap_len, width, height;
188 picture_id = bytestream_get_be16(&buf);
190 /* skip 1 unknown byte: Version Number */
193 /* Read the Sequence Description to determine if start of RLE data or appended to previous RLE */
194 sequence_desc = bytestream_get_byte(&buf);
196 if (!(sequence_desc & 0x80)) {
197 /* Additional RLE data */
198 if (buf_size > ctx->pictures[picture_id].rle_remaining_len)
201 memcpy(ctx->pictures[picture_id].rle + ctx->pictures[picture_id].rle_data_len, buf, buf_size);
202 ctx->pictures[picture_id].rle_data_len += buf_size;
203 ctx->pictures[picture_id].rle_remaining_len -= buf_size;
212 /* Decode rle bitmap length, stored size includes width/height data */
213 rle_bitmap_len = bytestream_get_be24(&buf) - 2*2;
215 /* Get bitmap dimensions from data */
216 width = bytestream_get_be16(&buf);
217 height = bytestream_get_be16(&buf);
219 /* Make sure the bitmap is not too large */
220 if (avctx->width < width || avctx->height < height) {
221 av_log(avctx, AV_LOG_ERROR, "Bitmap dimensions larger than video.\n");
225 ctx->pictures[picture_id].w = width;
226 ctx->pictures[picture_id].h = height;
228 av_fast_malloc(&ctx->pictures[picture_id].rle, &ctx->pictures[picture_id].rle_buffer_size, rle_bitmap_len);
230 if (!ctx->pictures[picture_id].rle)
233 memcpy(ctx->pictures[picture_id].rle, buf, buf_size);
234 ctx->pictures[picture_id].rle_data_len = buf_size;
235 ctx->pictures[picture_id].rle_remaining_len = rle_bitmap_len - buf_size;
241 * Parse the palette segment packet.
243 * The palette segment contains details of the palette,
244 * a maximum of 256 colors can be defined.
246 * @param avctx contains the current codec context
247 * @param buf pointer to the packet to process
248 * @param buf_size size of packet to process
250 static void parse_palette_segment(AVCodecContext *avctx,
251 const uint8_t *buf, int buf_size)
253 PGSSubContext *ctx = avctx->priv_data;
255 const uint8_t *buf_end = buf + buf_size;
256 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
258 int y, cb, cr, alpha;
259 int r, g, b, r_add, g_add, b_add;
261 /* Skip two null bytes */
264 while (buf < buf_end) {
265 color_id = bytestream_get_byte(&buf);
266 y = bytestream_get_byte(&buf);
267 cr = bytestream_get_byte(&buf);
268 cb = bytestream_get_byte(&buf);
269 alpha = bytestream_get_byte(&buf);
272 YUV_TO_RGB2(r, g, b, y);
274 av_dlog(avctx, "Color %d := (%d,%d,%d,%d)\n", color_id, r, g, b, alpha);
276 /* Store color in palette */
277 ctx->clut[color_id] = RGBA(r,g,b,alpha);
282 * Parse the presentation segment packet.
284 * The presentation segment contains details on the video
285 * width, video height, x & y subtitle position.
287 * @param avctx contains the current codec context
288 * @param buf pointer to the packet to process
289 * @param buf_size size of packet to process
290 * @todo TODO: Implement cropping
292 static void parse_presentation_segment(AVCodecContext *avctx,
293 const uint8_t *buf, int buf_size)
295 PGSSubContext *ctx = avctx->priv_data;
297 int w = bytestream_get_be16(&buf);
298 int h = bytestream_get_be16(&buf);
300 uint16_t object_index;
302 av_dlog(avctx, "Video Dimensions %dx%d\n",
304 if (av_image_check_size(w, h, 0, avctx) >= 0)
305 avcodec_set_dimensions(avctx, w, h);
307 /* Skip 1 bytes of unknown, frame rate? */
310 ctx->presentation.id_number = bytestream_get_be16(&buf);
313 * Skip 3 bytes of unknown:
315 * palette_update_flag (0x80),
320 ctx->presentation.object_count = bytestream_get_byte(&buf);
321 if (!ctx->presentation.object_count)
324 /* Verify that enough bytes are remaining for all of the objects. */
326 if (buf_size < ctx->presentation.object_count * 8) {
327 ctx->presentation.object_count = 0;
331 av_freep(&ctx->presentation.objects);
332 ctx->presentation.objects = av_malloc(sizeof(PGSSubPictureReference) * ctx->presentation.object_count);
333 if (!ctx->presentation.objects) {
334 ctx->presentation.object_count = 0;
338 for (object_index = 0; object_index < ctx->presentation.object_count; ++object_index) {
339 PGSSubPictureReference *reference = &ctx->presentation.objects[object_index];
340 reference->picture_id = bytestream_get_be16(&buf);
342 /* Skip window_id_ref */
344 /* composition_flag (0x80 - object cropped, 0x40 - object forced) */
345 reference->composition = bytestream_get_byte(&buf);
347 reference->x = bytestream_get_be16(&buf);
348 reference->y = bytestream_get_be16(&buf);
350 /* TODO If cropping, cropping_x, cropping_y, cropping_width, cropping_height (all 2 bytes).*/
351 av_dlog(avctx, "Subtitle Placement ID=%d, x=%d, y=%d\n", reference->picture_id, reference->x, reference->y);
353 if (reference->x > avctx->width || reference->y > avctx->height) {
354 av_log(avctx, AV_LOG_ERROR, "Subtitle out of video bounds. x = %d, y = %d, video width = %d, video height = %d.\n",
355 reference->x, reference->y, avctx->width, avctx->height);
363 * Parse the display segment packet.
365 * The display segment controls the updating of the display.
367 * @param avctx contains the current codec context
368 * @param data pointer to the data pertaining the subtitle to display
369 * @param buf pointer to the packet to process
370 * @param buf_size size of packet to process
371 * @todo TODO: Fix start time, relies on correct PTS, currently too late
373 * @todo TODO: Fix end time, normally cleared by a second display
374 * @todo segment, which is currently ignored as it clears
375 * @todo the subtitle too early.
377 static int display_end_segment(AVCodecContext *avctx, void *data,
378 const uint8_t *buf, int buf_size)
380 AVSubtitle *sub = data;
381 PGSSubContext *ctx = avctx->priv_data;
387 * The end display time is a timeout value and is only reached
388 * if the next subtitle is later than timeout or subtitle has
389 * not been cleared by a subsequent empty display command.
392 pts = ctx->pts != AV_NOPTS_VALUE ? ctx->pts : sub->pts;
393 memset(sub, 0, sizeof(*sub));
395 ctx->pts = AV_NOPTS_VALUE;
397 // Blank if last object_count was 0.
398 if (!ctx->presentation.object_count)
401 sub->start_display_time = 0;
402 sub->end_display_time = 20000;
405 sub->num_rects = ctx->presentation.object_count;
406 sub->rects = av_mallocz(sizeof(*sub->rects) * sub->num_rects);
408 for (rect = 0; rect < sub->num_rects; ++rect) {
409 uint16_t picture_id = ctx->presentation.objects[rect].picture_id;
410 sub->rects[rect] = av_mallocz(sizeof(*sub->rects[rect]));
411 sub->rects[rect]->x = ctx->presentation.objects[rect].x;
412 sub->rects[rect]->y = ctx->presentation.objects[rect].y;
413 sub->rects[rect]->w = ctx->pictures[picture_id].w;
414 sub->rects[rect]->h = ctx->pictures[picture_id].h;
415 sub->rects[rect]->type = SUBTITLE_BITMAP;
418 sub->rects[rect]->pict.linesize[0] = ctx->pictures[picture_id].w;
419 if (ctx->pictures[picture_id].rle) {
420 if (ctx->pictures[picture_id].rle_remaining_len)
421 av_log(avctx, AV_LOG_ERROR, "RLE data length %u is %u bytes shorter than expected\n",
422 ctx->pictures[picture_id].rle_data_len, ctx->pictures[picture_id].rle_remaining_len);
423 if (decode_rle(avctx, sub, rect, ctx->pictures[picture_id].rle, ctx->pictures[picture_id].rle_data_len) < 0)
427 /* Allocate memory for colors */
428 sub->rects[rect]->nb_colors = 256;
429 sub->rects[rect]->pict.data[1] = av_mallocz(AVPALETTE_SIZE);
431 /* Copy the forced flag */
432 sub->rects[rect]->forced = (ctx->presentation.objects[rect].composition & 0x40) != 0;
434 if (!ctx->forced_subs_only || ctx->presentation.objects[rect].composition & 0x40)
435 memcpy(sub->rects[rect]->pict.data[1], ctx->clut, sub->rects[rect]->nb_colors * sizeof(uint32_t));
441 static int decode(AVCodecContext *avctx, void *data, int *data_size,
444 PGSSubContext *ctx = avctx->priv_data;
445 const uint8_t *buf = avpkt->data;
446 int buf_size = avpkt->size;
447 AVSubtitle *sub = data;
449 const uint8_t *buf_end;
450 uint8_t segment_type;
454 av_dlog(avctx, "PGS sub packet:\n");
456 for (i = 0; i < buf_size; i++) {
457 av_dlog(avctx, "%02x ", buf[i]);
459 av_dlog(avctx, "\n");
463 av_dlog(avctx, "\n");
467 /* Ensure that we have received at a least a segment code and segment length */
471 buf_end = buf + buf_size;
473 /* Step through buffer to identify segments */
474 while (buf < buf_end) {
475 segment_type = bytestream_get_byte(&buf);
476 segment_length = bytestream_get_be16(&buf);
478 av_dlog(avctx, "Segment Length %d, Segment Type %x\n", segment_length, segment_type);
480 if (segment_type != DISPLAY_SEGMENT && segment_length > buf_end - buf)
483 switch (segment_type) {
484 case PALETTE_SEGMENT:
485 parse_palette_segment(avctx, buf, segment_length);
487 case PICTURE_SEGMENT:
488 parse_picture_segment(avctx, buf, segment_length);
490 case PRESENTATION_SEGMENT:
491 parse_presentation_segment(avctx, buf, segment_length);
496 * Window Segment Structure (No new information provided):
498 * 2 bytes: X position of subtitle,
499 * 2 bytes: Y position of subtitle,
500 * 2 bytes: Width of subtitle,
501 * 2 bytes: Height of subtitle.
504 case DISPLAY_SEGMENT:
505 *data_size = display_end_segment(avctx, data, buf, segment_length);
508 av_log(avctx, AV_LOG_ERROR, "Unknown subtitle segment type 0x%x, length %d\n",
509 segment_type, segment_length);
513 buf += segment_length;
519 #define OFFSET(x) offsetof(PGSSubContext, x)
520 #define SD AV_OPT_FLAG_SUBTITLE_PARAM | AV_OPT_FLAG_DECODING_PARAM
521 static const AVOption options[] = {
522 {"forced_subs_only", "Only show forced subtitles", OFFSET(forced_subs_only), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, SD},
526 static const AVClass pgsdec_class = {
527 .class_name = "PGS subtitle decoder",
528 .item_name = av_default_item_name,
530 .version = LIBAVUTIL_VERSION_INT,
533 AVCodec ff_pgssub_decoder = {
535 .type = AVMEDIA_TYPE_SUBTITLE,
536 .id = AV_CODEC_ID_HDMV_PGS_SUBTITLE,
537 .priv_data_size = sizeof(PGSSubContext),
538 .init = init_decoder,
539 .close = close_decoder,
541 .long_name = NULL_IF_CONFIG_SMALL("HDMV Presentation Graphic Stream subtitles"),
542 .priv_class = &pgsdec_class,