FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
pgssubdec.c
Go to the documentation of this file.
1 /*
2  * PGS subtitle decoder
3  * Copyright (c) 2009 Stephen Backway
4  *
5  * This file is part of FFmpeg.
6  *
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.
11  *
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.
16  *
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
20  */
21 
22 /**
23  * @file
24  * PGS subtitle decoder
25  */
26 
27 #include "avcodec.h"
28 #include "bytestream.h"
29 #include "internal.h"
30 #include "mathops.h"
31 
32 #include "libavutil/colorspace.h"
33 #include "libavutil/imgutils.h"
34 #include "libavutil/opt.h"
35 
36 #define RGBA(r,g,b,a) (((a) << 24) | ((r) << 16) | ((g) << 8) | (b))
37 #define MAX_EPOCH_PALETTES 8 // Max 8 allowed per PGS epoch
38 #define MAX_EPOCH_OBJECTS 64 // Max 64 allowed per PGS epoch
39 #define MAX_OBJECT_REFS 2 // Max objects per display set
40 
47 };
48 
49 typedef struct PGSSubObjectRef {
50  int id;
51  int window_id;
53  int x;
54  int y;
55  int crop_x;
56  int crop_y;
57  int crop_w;
58  int crop_h;
60 
61 typedef struct PGSSubPresentation {
62  int id_number;
66  int64_t pts;
68 
69 typedef struct PGSSubObject {
70  int id;
71  int w;
72  int h;
75  unsigned int rle_remaining_len;
76 } PGSSubObject;
77 
78 typedef struct PGSSubObjects {
79  int count;
82 
83 typedef struct PGSSubPalette {
84  int id;
85  uint32_t clut[256];
87 
88 typedef struct PGSSubPalettes {
89  int count;
92 
93 typedef struct PGSSubContext {
94  AVClass *class;
100 
101 static void flush_cache(AVCodecContext *avctx)
102 {
103  PGSSubContext *ctx = avctx->priv_data;
104  int i;
105 
106  for (i = 0; i < ctx->objects.count; i++) {
107  av_freep(&ctx->objects.object[i].rle);
108  ctx->objects.object[i].rle_buffer_size = 0;
109  ctx->objects.object[i].rle_remaining_len = 0;
110  }
111  ctx->objects.count = 0;
112  ctx->palettes.count = 0;
113 }
114 
115 static PGSSubObject * find_object(int id, PGSSubObjects *objects)
116 {
117  int i;
118 
119  for (i = 0; i < objects->count; i++) {
120  if (objects->object[i].id == id)
121  return &objects->object[i];
122  }
123  return NULL;
124 }
125 
126 static PGSSubPalette * find_palette(int id, PGSSubPalettes *palettes)
127 {
128  int i;
129 
130  for (i = 0; i < palettes->count; i++) {
131  if (palettes->palette[i].id == id)
132  return &palettes->palette[i];
133  }
134  return NULL;
135 }
136 
138 {
139  avctx->pix_fmt = AV_PIX_FMT_PAL8;
140 
141  return 0;
142 }
143 
145 {
146  flush_cache(avctx);
147 
148  return 0;
149 }
150 
151 /**
152  * Decode the RLE data.
153  *
154  * The subtitle is stored as a Run Length Encoded image.
155  *
156  * @param avctx contains the current codec context
157  * @param sub pointer to the processed subtitle data
158  * @param buf pointer to the RLE data to process
159  * @param buf_size size of the RLE data to process
160  */
162  const uint8_t *buf, unsigned int buf_size)
163 {
164  const uint8_t *rle_bitmap_end;
165  int pixel_count, line_count;
166 
167  rle_bitmap_end = buf + buf_size;
168 
169  rect->pict.data[0] = av_malloc(rect->w * rect->h);
170 
171  if (!rect->pict.data[0])
172  return AVERROR(ENOMEM);
173 
174  pixel_count = 0;
175  line_count = 0;
176 
177  while (buf < rle_bitmap_end && line_count < rect->h) {
179  int run;
180 
181  color = bytestream_get_byte(&buf);
182  run = 1;
183 
184  if (color == 0x00) {
185  flags = bytestream_get_byte(&buf);
186  run = flags & 0x3f;
187  if (flags & 0x40)
188  run = (run << 8) + bytestream_get_byte(&buf);
189  color = flags & 0x80 ? bytestream_get_byte(&buf) : 0;
190  }
191 
192  if (run > 0 && pixel_count + run <= rect->w * rect->h) {
193  memset(rect->pict.data[0] + pixel_count, color, run);
194  pixel_count += run;
195  } else if (!run) {
196  /*
197  * New Line. Check if correct pixels decoded, if not display warning
198  * and adjust bitmap pointer to correct new line position.
199  */
200  if (pixel_count % rect->w > 0) {
201  av_log(avctx, AV_LOG_ERROR, "Decoded %d pixels, when line should be %d pixels\n",
202  pixel_count % rect->w, rect->w);
203  if (avctx->err_recognition & AV_EF_EXPLODE) {
204  return AVERROR_INVALIDDATA;
205  }
206  }
207  line_count++;
208  }
209  }
210 
211  if (pixel_count < rect->w * rect->h) {
212  av_log(avctx, AV_LOG_ERROR, "Insufficient RLE data for subtitle\n");
213  return AVERROR_INVALIDDATA;
214  }
215 
216  av_dlog(avctx, "Pixel Count = %d, Area = %d\n", pixel_count, rect->w * rect->h);
217 
218  return 0;
219 }
220 
221 /**
222  * Parse the picture segment packet.
223  *
224  * The picture segment contains details on the sequence id,
225  * width, height and Run Length Encoded (RLE) bitmap data.
226  *
227  * @param avctx contains the current codec context
228  * @param buf pointer to the packet to process
229  * @param buf_size size of packet to process
230  */
232  const uint8_t *buf, int buf_size)
233 {
234  PGSSubContext *ctx = avctx->priv_data;
235  PGSSubObject *object;
236 
237  uint8_t sequence_desc;
238  unsigned int rle_bitmap_len, width, height;
239  int id;
240 
241  if (buf_size <= 4)
242  return AVERROR_INVALIDDATA;
243  buf_size -= 4;
244 
245  id = bytestream_get_be16(&buf);
246  object = find_object(id, &ctx->objects);
247  if (!object) {
248  if (ctx->objects.count >= MAX_EPOCH_OBJECTS) {
249  av_log(avctx, AV_LOG_ERROR, "Too many objects in epoch\n");
250  return AVERROR_INVALIDDATA;
251  }
252  object = &ctx->objects.object[ctx->objects.count++];
253  object->id = id;
254  }
255 
256  /* skip object version number */
257  buf += 1;
258 
259  /* Read the Sequence Description to determine if start of RLE data or appended to previous RLE */
260  sequence_desc = bytestream_get_byte(&buf);
261 
262  if (!(sequence_desc & 0x80)) {
263  /* Additional RLE data */
264  if (buf_size > object->rle_remaining_len)
265  return AVERROR_INVALIDDATA;
266 
267  memcpy(object->rle + object->rle_data_len, buf, buf_size);
268  object->rle_data_len += buf_size;
269  object->rle_remaining_len -= buf_size;
270 
271  return 0;
272  }
273 
274  if (buf_size <= 7)
275  return AVERROR_INVALIDDATA;
276  buf_size -= 7;
277 
278  /* Decode rle bitmap length, stored size includes width/height data */
279  rle_bitmap_len = bytestream_get_be24(&buf) - 2*2;
280 
281  if (buf_size > rle_bitmap_len) {
282  av_log(avctx, AV_LOG_ERROR,
283  "Buffer dimension %d larger than the expected RLE data %d\n",
284  buf_size, rle_bitmap_len);
285  return AVERROR_INVALIDDATA;
286  }
287 
288  /* Get bitmap dimensions from data */
289  width = bytestream_get_be16(&buf);
290  height = bytestream_get_be16(&buf);
291 
292  /* Make sure the bitmap is not too large */
293  if (avctx->width < width || avctx->height < height) {
294  av_log(avctx, AV_LOG_ERROR, "Bitmap dimensions larger than video.\n");
295  return AVERROR_INVALIDDATA;
296  }
297 
298  object->w = width;
299  object->h = height;
300 
301  av_fast_padded_malloc(&object->rle, &object->rle_buffer_size, rle_bitmap_len);
302 
303  if (!object->rle)
304  return AVERROR(ENOMEM);
305 
306  memcpy(object->rle, buf, buf_size);
307  object->rle_data_len = buf_size;
308  object->rle_remaining_len = rle_bitmap_len - buf_size;
309 
310  return 0;
311 }
312 
313 /**
314  * Parse the palette segment packet.
315  *
316  * The palette segment contains details of the palette,
317  * a maximum of 256 colors can be defined.
318  *
319  * @param avctx contains the current codec context
320  * @param buf pointer to the packet to process
321  * @param buf_size size of packet to process
322  */
324  const uint8_t *buf, int buf_size)
325 {
326  PGSSubContext *ctx = avctx->priv_data;
328 
329  const uint8_t *buf_end = buf + buf_size;
330  const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
331  int color_id;
332  int y, cb, cr, alpha;
333  int r, g, b, r_add, g_add, b_add;
334  int id;
335 
336  id = bytestream_get_byte(&buf);
337  palette = find_palette(id, &ctx->palettes);
338  if (!palette) {
339  if (ctx->palettes.count >= MAX_EPOCH_PALETTES) {
340  av_log(avctx, AV_LOG_ERROR, "Too many palettes in epoch\n");
341  return AVERROR_INVALIDDATA;
342  }
343  palette = &ctx->palettes.palette[ctx->palettes.count++];
344  palette->id = id;
345  }
346 
347  /* Skip palette version */
348  buf += 1;
349 
350  while (buf < buf_end) {
351  color_id = bytestream_get_byte(&buf);
352  y = bytestream_get_byte(&buf);
353  cr = bytestream_get_byte(&buf);
354  cb = bytestream_get_byte(&buf);
355  alpha = bytestream_get_byte(&buf);
356 
357  YUV_TO_RGB1(cb, cr);
358  YUV_TO_RGB2(r, g, b, y);
359 
360  av_dlog(avctx, "Color %d := (%d,%d,%d,%d)\n", color_id, r, g, b, alpha);
361 
362  /* Store color in palette */
363  palette->clut[color_id] = RGBA(r,g,b,alpha);
364  }
365  return 0;
366 }
367 
368 /**
369  * Parse the presentation segment packet.
370  *
371  * The presentation segment contains details on the video
372  * width, video height, x & y subtitle position.
373  *
374  * @param avctx contains the current codec context
375  * @param buf pointer to the packet to process
376  * @param buf_size size of packet to process
377  * @todo TODO: Implement cropping
378  */
380  const uint8_t *buf, int buf_size,
381  int64_t pts)
382 {
383  PGSSubContext *ctx = avctx->priv_data;
384  int i, state, ret;
385  const uint8_t *buf_end = buf + buf_size;
386 
387  // Video descriptor
388  int w = bytestream_get_be16(&buf);
389  int h = bytestream_get_be16(&buf);
390 
391  ctx->presentation.pts = pts;
392 
393  av_dlog(avctx, "Video Dimensions %dx%d\n",
394  w, h);
395  ret = ff_set_dimensions(avctx, w, h);
396  if (ret < 0)
397  return ret;
398 
399  /* Skip 1 bytes of unknown, frame rate */
400  buf++;
401 
402  // Composition descriptor
403  ctx->presentation.id_number = bytestream_get_be16(&buf);
404  /*
405  * state is a 2 bit field that defines pgs epoch boundaries
406  * 00 - Normal, previously defined objects and palettes are still valid
407  * 01 - Acquisition point, previous objects and palettes can be released
408  * 10 - Epoch start, previous objects and palettes can be released
409  * 11 - Epoch continue, previous objects and palettes can be released
410  *
411  * reserved 6 bits discarded
412  */
413  state = bytestream_get_byte(&buf) >> 6;
414  if (state != 0) {
415  flush_cache(avctx);
416  }
417 
418  /*
419  * skip palette_update_flag (0x80),
420  */
421  buf += 1;
422  ctx->presentation.palette_id = bytestream_get_byte(&buf);
423  ctx->presentation.object_count = bytestream_get_byte(&buf);
425  av_log(avctx, AV_LOG_ERROR,
426  "Invalid number of presentation objects %d\n",
428  ctx->presentation.object_count = 2;
429  if (avctx->err_recognition & AV_EF_EXPLODE) {
430  return AVERROR_INVALIDDATA;
431  }
432  }
433 
434 
435  for (i = 0; i < ctx->presentation.object_count; i++)
436  {
437 
438  if (buf_end - buf < 8) {
439  av_log(avctx, AV_LOG_ERROR, "Insufficent space for object\n");
440  ctx->presentation.object_count = i;
441  return AVERROR_INVALIDDATA;
442  }
443 
444  ctx->presentation.objects[i].id = bytestream_get_be16(&buf);
445  ctx->presentation.objects[i].window_id = bytestream_get_byte(&buf);
446  ctx->presentation.objects[i].composition_flag = bytestream_get_byte(&buf);
447 
448  ctx->presentation.objects[i].x = bytestream_get_be16(&buf);
449  ctx->presentation.objects[i].y = bytestream_get_be16(&buf);
450 
451  // If cropping
452  if (ctx->presentation.objects[i].composition_flag & 0x80) {
453  ctx->presentation.objects[i].crop_x = bytestream_get_be16(&buf);
454  ctx->presentation.objects[i].crop_y = bytestream_get_be16(&buf);
455  ctx->presentation.objects[i].crop_w = bytestream_get_be16(&buf);
456  ctx->presentation.objects[i].crop_h = bytestream_get_be16(&buf);
457  }
458 
459  av_dlog(avctx, "Subtitle Placement x=%d, y=%d\n",
460  ctx->presentation.objects[i].x, ctx->presentation.objects[i].y);
461 
462  if (ctx->presentation.objects[i].x > avctx->width ||
463  ctx->presentation.objects[i].y > avctx->height) {
464  av_log(avctx, AV_LOG_ERROR, "Subtitle out of video bounds. x = %d, y = %d, video width = %d, video height = %d.\n",
465  ctx->presentation.objects[i].x,
466  ctx->presentation.objects[i].y,
467  avctx->width, avctx->height);
468  ctx->presentation.objects[i].x = 0;
469  ctx->presentation.objects[i].y = 0;
470  if (avctx->err_recognition & AV_EF_EXPLODE) {
471  return AVERROR_INVALIDDATA;
472  }
473  }
474  }
475 
476  return 0;
477 }
478 
479 /**
480  * Parse the display segment packet.
481  *
482  * The display segment controls the updating of the display.
483  *
484  * @param avctx contains the current codec context
485  * @param data pointer to the data pertaining the subtitle to display
486  * @param buf pointer to the packet to process
487  * @param buf_size size of packet to process
488  */
489 static int display_end_segment(AVCodecContext *avctx, void *data,
490  const uint8_t *buf, int buf_size)
491 {
492  AVSubtitle *sub = data;
493  PGSSubContext *ctx = avctx->priv_data;
494  int64_t pts;
496  int i, ret;
497 
498  pts = ctx->presentation.pts != AV_NOPTS_VALUE ? ctx->presentation.pts : sub->pts;
499  memset(sub, 0, sizeof(*sub));
500  sub->pts = pts;
502  sub->start_display_time = 0;
503  // There is no explicit end time for PGS subtitles. The end time
504  // is defined by the start of the next sub which may contain no
505  // objects (i.e. clears the previous sub)
506  sub->end_display_time = UINT32_MAX;
507  sub->format = 0;
508 
509  // Blank if last object_count was 0.
510  if (!ctx->presentation.object_count)
511  return 1;
512  sub->rects = av_mallocz(sizeof(*sub->rects) * ctx->presentation.object_count);
513  if (!sub->rects) {
514  return AVERROR(ENOMEM);
515  }
516  palette = find_palette(ctx->presentation.palette_id, &ctx->palettes);
517  if (!palette) {
518  // Missing palette. Should only happen with damaged streams.
519  av_log(avctx, AV_LOG_ERROR, "Invalid palette id %d\n",
520  ctx->presentation.palette_id);
521  avsubtitle_free(sub);
522  return AVERROR_INVALIDDATA;
523  }
524  for (i = 0; i < ctx->presentation.object_count; i++) {
525  PGSSubObject *object;
526 
527  sub->rects[i] = av_mallocz(sizeof(*sub->rects[0]));
528  if (!sub->rects[i]) {
529  avsubtitle_free(sub);
530  return AVERROR(ENOMEM);
531  }
532  sub->num_rects++;
533  sub->rects[i]->type = SUBTITLE_BITMAP;
534 
535  /* Process bitmap */
536  object = find_object(ctx->presentation.objects[i].id, &ctx->objects);
537  if (!object) {
538  // Missing object. Should only happen with damaged streams.
539  av_log(avctx, AV_LOG_ERROR, "Invalid object id %d\n",
540  ctx->presentation.objects[i].id);
541  if (avctx->err_recognition & AV_EF_EXPLODE) {
542  avsubtitle_free(sub);
543  return AVERROR_INVALIDDATA;
544  }
545  // Leaves rect empty with 0 width and height.
546  continue;
547  }
548  if (ctx->presentation.objects[i].composition_flag & 0x40)
549  sub->rects[i]->flags |= AV_SUBTITLE_FLAG_FORCED;
550 
551  sub->rects[i]->x = ctx->presentation.objects[i].x;
552  sub->rects[i]->y = ctx->presentation.objects[i].y;
553  sub->rects[i]->w = object->w;
554  sub->rects[i]->h = object->h;
555 
556  sub->rects[i]->pict.linesize[0] = object->w;
557 
558  if (object->rle) {
559  if (object->rle_remaining_len) {
560  av_log(avctx, AV_LOG_ERROR, "RLE data length %u is %u bytes shorter than expected\n",
561  object->rle_data_len, object->rle_remaining_len);
562  if (avctx->err_recognition & AV_EF_EXPLODE) {
563  avsubtitle_free(sub);
564  return AVERROR_INVALIDDATA;
565  }
566  }
567  ret = decode_rle(avctx, sub->rects[i], object->rle, object->rle_data_len);
568  if (ret < 0) {
569  if ((avctx->err_recognition & AV_EF_EXPLODE) ||
570  ret == AVERROR(ENOMEM)) {
571  avsubtitle_free(sub);
572  return ret;
573  }
574  sub->rects[i]->w = 0;
575  sub->rects[i]->h = 0;
576  continue;
577  }
578  }
579  /* Allocate memory for colors */
580  sub->rects[i]->nb_colors = 256;
581  sub->rects[i]->pict.data[1] = av_mallocz(AVPALETTE_SIZE);
582  if (!sub->rects[i]->pict.data[1]) {
583  avsubtitle_free(sub);
584  return AVERROR(ENOMEM);
585  }
586 
587  if (!ctx->forced_subs_only || ctx->presentation.objects[i].composition_flag & 0x40)
588  memcpy(sub->rects[i]->pict.data[1], palette->clut, sub->rects[i]->nb_colors * sizeof(uint32_t));
589 
590  }
591  return 1;
592 }
593 
594 static int decode(AVCodecContext *avctx, void *data, int *data_size,
595  AVPacket *avpkt)
596 {
597  const uint8_t *buf = avpkt->data;
598  int buf_size = avpkt->size;
599 
600  const uint8_t *buf_end;
601  uint8_t segment_type;
602  int segment_length;
603  int i, ret;
604 
605  av_dlog(avctx, "PGS sub packet:\n");
606 
607  for (i = 0; i < buf_size; i++) {
608  av_dlog(avctx, "%02x ", buf[i]);
609  if (i % 16 == 15)
610  av_dlog(avctx, "\n");
611  }
612 
613  if (i & 15)
614  av_dlog(avctx, "\n");
615 
616  *data_size = 0;
617 
618  /* Ensure that we have received at a least a segment code and segment length */
619  if (buf_size < 3)
620  return -1;
621 
622  buf_end = buf + buf_size;
623 
624  /* Step through buffer to identify segments */
625  while (buf < buf_end) {
626  segment_type = bytestream_get_byte(&buf);
627  segment_length = bytestream_get_be16(&buf);
628 
629  av_dlog(avctx, "Segment Length %d, Segment Type %x\n", segment_length, segment_type);
630 
631  if (segment_type != DISPLAY_SEGMENT && segment_length > buf_end - buf)
632  break;
633 
634  ret = 0;
635  switch (segment_type) {
636  case PALETTE_SEGMENT:
637  ret = parse_palette_segment(avctx, buf, segment_length);
638  break;
639  case OBJECT_SEGMENT:
640  ret = parse_object_segment(avctx, buf, segment_length);
641  break;
643  ret = parse_presentation_segment(avctx, buf, segment_length, ((AVSubtitle*)(data))->pts);
644  break;
645  case WINDOW_SEGMENT:
646  /*
647  * Window Segment Structure (No new information provided):
648  * 2 bytes: Unknown,
649  * 2 bytes: X position of subtitle,
650  * 2 bytes: Y position of subtitle,
651  * 2 bytes: Width of subtitle,
652  * 2 bytes: Height of subtitle.
653  */
654  break;
655  case DISPLAY_SEGMENT:
656  ret = display_end_segment(avctx, data, buf, segment_length);
657  if (ret >= 0)
658  *data_size = ret;
659  break;
660  default:
661  av_log(avctx, AV_LOG_ERROR, "Unknown subtitle segment type 0x%x, length %d\n",
662  segment_type, segment_length);
663  ret = AVERROR_INVALIDDATA;
664  break;
665  }
666  if (ret < 0 && (avctx->err_recognition & AV_EF_EXPLODE))
667  return ret;
668 
669  buf += segment_length;
670  }
671 
672  return buf_size;
673 }
674 
675 #define OFFSET(x) offsetof(PGSSubContext, x)
676 #define SD AV_OPT_FLAG_SUBTITLE_PARAM | AV_OPT_FLAG_DECODING_PARAM
677 static const AVOption options[] = {
678  {"forced_subs_only", "Only show forced subtitles", OFFSET(forced_subs_only), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, SD},
679  { NULL },
680 };
681 
682 static const AVClass pgsdec_class = {
683  .class_name = "PGS subtitle decoder",
684  .item_name = av_default_item_name,
685  .option = options,
686  .version = LIBAVUTIL_VERSION_INT,
687 };
688 
690  .name = "pgssub",
691  .long_name = NULL_IF_CONFIG_SMALL("HDMV Presentation Graphic Stream subtitles"),
692  .type = AVMEDIA_TYPE_SUBTITLE,
694  .priv_data_size = sizeof(PGSSubContext),
695  .init = init_decoder,
696  .close = close_decoder,
697  .decode = decode,
698  .priv_class = &pgsdec_class,
699 };