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 "dsputil.h"
29 #include "bytestream.h"
30 #include "internal.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 
44 };
45 
46 typedef struct PGSSubPictureReference {
47  int x;
48  int y;
52 
53 typedef struct PGSSubPresentation {
54  int id_number;
57  int64_t pts;
59 
60 typedef struct PGSSubPicture {
61  int w;
62  int h;
65  unsigned int rle_remaining_len;
67 
68 typedef struct PGSSubContext {
69  AVClass *class;
71  uint32_t clut[256];
72  PGSSubPicture pictures[UINT16_MAX];
75 
77 {
78  avctx->pix_fmt = AV_PIX_FMT_PAL8;
79 
80  return 0;
81 }
82 
84 {
85  uint16_t picture;
86 
87  PGSSubContext *ctx = avctx->priv_data;
88 
90  ctx->presentation.object_count = 0;
91 
92  for (picture = 0; picture < UINT16_MAX; ++picture) {
93  av_freep(&ctx->pictures[picture].rle);
94  ctx->pictures[picture].rle_buffer_size = 0;
95  }
96 
97  return 0;
98 }
99 
100 /**
101  * Decode the RLE data.
102  *
103  * The subtitle is stored as a Run Length Encoded image.
104  *
105  * @param avctx contains the current codec context
106  * @param sub pointer to the processed subtitle data
107  * @param buf pointer to the RLE data to process
108  * @param buf_size size of the RLE data to process
109  */
110 static int decode_rle(AVCodecContext *avctx, AVSubtitle *sub, int rect,
111  const uint8_t *buf, unsigned int buf_size)
112 {
113  const uint8_t *rle_bitmap_end;
114  int pixel_count, line_count;
115 
116  rle_bitmap_end = buf + buf_size;
117 
118  sub->rects[rect]->pict.data[0] = av_malloc(sub->rects[rect]->w * sub->rects[rect]->h);
119 
120  if (!sub->rects[rect]->pict.data[0])
121  return -1;
122 
123  pixel_count = 0;
124  line_count = 0;
125 
126  while (buf < rle_bitmap_end && line_count < sub->rects[rect]->h) {
128  int run;
129 
130  color = bytestream_get_byte(&buf);
131  run = 1;
132 
133  if (color == 0x00) {
134  flags = bytestream_get_byte(&buf);
135  run = flags & 0x3f;
136  if (flags & 0x40)
137  run = (run << 8) + bytestream_get_byte(&buf);
138  color = flags & 0x80 ? bytestream_get_byte(&buf) : 0;
139  }
140 
141  if (run > 0 && pixel_count + run <= sub->rects[rect]->w * sub->rects[rect]->h) {
142  memset(sub->rects[rect]->pict.data[0] + pixel_count, color, run);
143  pixel_count += run;
144  } else if (!run) {
145  /*
146  * New Line. Check if correct pixels decoded, if not display warning
147  * and adjust bitmap pointer to correct new line position.
148  */
149  if (pixel_count % sub->rects[rect]->w > 0)
150  av_log(avctx, AV_LOG_ERROR, "Decoded %d pixels, when line should be %d pixels\n",
151  pixel_count % sub->rects[rect]->w, sub->rects[rect]->w);
152  line_count++;
153  }
154  }
155 
156  if (pixel_count < sub->rects[rect]->w * sub->rects[rect]->h) {
157  av_log(avctx, AV_LOG_ERROR, "Insufficient RLE data for subtitle\n");
158  return -1;
159  }
160 
161  av_dlog(avctx, "Pixel Count = %d, Area = %d\n", pixel_count, sub->rects[rect]->w * sub->rects[rect]->h);
162 
163  return 0;
164 }
165 
166 /**
167  * Parse the picture segment packet.
168  *
169  * The picture segment contains details on the sequence id,
170  * width, height and Run Length Encoded (RLE) bitmap data.
171  *
172  * @param avctx contains the current codec context
173  * @param buf pointer to the packet to process
174  * @param buf_size size of packet to process
175  * @todo TODO: Enable support for RLE data over multiple packets
176  */
178  const uint8_t *buf, int buf_size)
179 {
180  PGSSubContext *ctx = avctx->priv_data;
181 
182  uint8_t sequence_desc;
183  unsigned int rle_bitmap_len, width, height;
184  uint16_t picture_id;
185 
186  if (buf_size <= 4)
187  return -1;
188  buf_size -= 4;
189 
190  picture_id = bytestream_get_be16(&buf);
191 
192  /* skip 1 unknown byte: Version Number */
193  buf++;
194 
195  /* Read the Sequence Description to determine if start of RLE data or appended to previous RLE */
196  sequence_desc = bytestream_get_byte(&buf);
197 
198  if (!(sequence_desc & 0x80)) {
199  /* Additional RLE data */
200  if (buf_size > ctx->pictures[picture_id].rle_remaining_len)
201  return -1;
202 
203  memcpy(ctx->pictures[picture_id].rle + ctx->pictures[picture_id].rle_data_len, buf, buf_size);
204  ctx->pictures[picture_id].rle_data_len += buf_size;
205  ctx->pictures[picture_id].rle_remaining_len -= buf_size;
206 
207  return 0;
208  }
209 
210  if (buf_size <= 7)
211  return -1;
212  buf_size -= 7;
213 
214  /* Decode rle bitmap length, stored size includes width/height data */
215  rle_bitmap_len = bytestream_get_be24(&buf) - 2*2;
216 
217  /* Get bitmap dimensions from data */
218  width = bytestream_get_be16(&buf);
219  height = bytestream_get_be16(&buf);
220 
221  /* Make sure the bitmap is not too large */
222  if (avctx->width < width || avctx->height < height) {
223  av_log(avctx, AV_LOG_ERROR, "Bitmap dimensions larger than video.\n");
224  return -1;
225  }
226 
227  if (buf_size > rle_bitmap_len) {
228  av_log(avctx, AV_LOG_ERROR, "too much RLE data\n");
229  return AVERROR_INVALIDDATA;
230  }
231 
232  ctx->pictures[picture_id].w = width;
233  ctx->pictures[picture_id].h = height;
234 
235  av_fast_padded_malloc(&ctx->pictures[picture_id].rle, &ctx->pictures[picture_id].rle_buffer_size, rle_bitmap_len);
236 
237  if (!ctx->pictures[picture_id].rle)
238  return -1;
239 
240  memcpy(ctx->pictures[picture_id].rle, buf, buf_size);
241  ctx->pictures[picture_id].rle_data_len = buf_size;
242  ctx->pictures[picture_id].rle_remaining_len = rle_bitmap_len - buf_size;
243 
244  return 0;
245 }
246 
247 /**
248  * Parse the palette segment packet.
249  *
250  * The palette segment contains details of the palette,
251  * a maximum of 256 colors can be defined.
252  *
253  * @param avctx contains the current codec context
254  * @param buf pointer to the packet to process
255  * @param buf_size size of packet to process
256  */
258  const uint8_t *buf, int buf_size)
259 {
260  PGSSubContext *ctx = avctx->priv_data;
261 
262  const uint8_t *buf_end = buf + buf_size;
263  const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
264  int color_id;
265  int y, cb, cr, alpha;
266  int r, g, b, r_add, g_add, b_add;
267 
268  /* Skip two null bytes */
269  buf += 2;
270 
271  while (buf < buf_end) {
272  color_id = bytestream_get_byte(&buf);
273  y = bytestream_get_byte(&buf);
274  cr = bytestream_get_byte(&buf);
275  cb = bytestream_get_byte(&buf);
276  alpha = bytestream_get_byte(&buf);
277 
278  YUV_TO_RGB1(cb, cr);
279  YUV_TO_RGB2(r, g, b, y);
280 
281  av_dlog(avctx, "Color %d := (%d,%d,%d,%d)\n", color_id, r, g, b, alpha);
282 
283  /* Store color in palette */
284  ctx->clut[color_id] = RGBA(r,g,b,alpha);
285  }
286 }
287 
288 /**
289  * Parse the presentation segment packet.
290  *
291  * The presentation segment contains details on the video
292  * width, video height, x & y subtitle position.
293  *
294  * @param avctx contains the current codec context
295  * @param buf pointer to the packet to process
296  * @param buf_size size of packet to process
297  * @todo TODO: Implement cropping
298  */
300  const uint8_t *buf, int buf_size,
301  int64_t pts)
302 {
303  PGSSubContext *ctx = avctx->priv_data;
304  int ret;
305 
306  int w = bytestream_get_be16(&buf);
307  int h = bytestream_get_be16(&buf);
308 
309  uint16_t object_index;
310 
311  ctx->presentation.pts = pts;
312 
313  av_dlog(avctx, "Video Dimensions %dx%d\n",
314  w, h);
315  ret = ff_set_dimensions(avctx, w, h);
316  if (ret < 0)
317  return ret;
318 
319  /* Skip 1 bytes of unknown, frame rate? */
320  buf++;
321 
322  ctx->presentation.id_number = bytestream_get_be16(&buf);
323 
324  /*
325  * Skip 3 bytes of unknown:
326  * state
327  * palette_update_flag (0x80),
328  * palette_id_to_use,
329  */
330  buf += 3;
331 
332  ctx->presentation.object_count = bytestream_get_byte(&buf);
333  if (!ctx->presentation.object_count)
334  return 0;
335 
336  /* Verify that enough bytes are remaining for all of the objects. */
337  buf_size -= 11;
338  if (buf_size < ctx->presentation.object_count * 8) {
339  ctx->presentation.object_count = 0;
340  return AVERROR_INVALIDDATA;
341  }
342 
345  if (!ctx->presentation.objects) {
346  ctx->presentation.object_count = 0;
347  return AVERROR(ENOMEM);
348  }
349 
350  for (object_index = 0; object_index < ctx->presentation.object_count; ++object_index) {
351  PGSSubPictureReference *reference = &ctx->presentation.objects[object_index];
352  reference->picture_id = bytestream_get_be16(&buf);
353 
354  /* Skip window_id_ref */
355  buf++;
356  /* composition_flag (0x80 - object cropped, 0x40 - object forced) */
357  reference->composition = bytestream_get_byte(&buf);
358 
359  reference->x = bytestream_get_be16(&buf);
360  reference->y = bytestream_get_be16(&buf);
361 
362  /* TODO If cropping, cropping_x, cropping_y, cropping_width, cropping_height (all 2 bytes).*/
363  av_dlog(avctx, "Subtitle Placement ID=%d, x=%d, y=%d\n", reference->picture_id, reference->x, reference->y);
364 
365  if (reference->x > avctx->width || reference->y > avctx->height) {
366  av_log(avctx, AV_LOG_ERROR, "Subtitle out of video bounds. x = %d, y = %d, video width = %d, video height = %d.\n",
367  reference->x, reference->y, avctx->width, avctx->height);
368  reference->x = 0;
369  reference->y = 0;
370  }
371  }
372 
373  return 0;
374 }
375 
376 /**
377  * Parse the display segment packet.
378  *
379  * The display segment controls the updating of the display.
380  *
381  * @param avctx contains the current codec context
382  * @param data pointer to the data pertaining the subtitle to display
383  * @param buf pointer to the packet to process
384  * @param buf_size size of packet to process
385  * @todo TODO: Fix start time, relies on correct PTS, currently too late
386  *
387  * @todo TODO: Fix end time, normally cleared by a second display
388  * @todo segment, which is currently ignored as it clears
389  * @todo the subtitle too early.
390  */
391 static int display_end_segment(AVCodecContext *avctx, void *data,
392  const uint8_t *buf, int buf_size)
393 {
394  AVSubtitle *sub = data;
395  PGSSubContext *ctx = avctx->priv_data;
396  int64_t pts;
397 
398  uint16_t rect;
399 
400  /*
401  * The end display time is a timeout value and is only reached
402  * if the next subtitle is later than timeout or subtitle has
403  * not been cleared by a subsequent empty display command.
404  */
405 
406  pts = ctx->presentation.pts != AV_NOPTS_VALUE ? ctx->presentation.pts : sub->pts;
407  memset(sub, 0, sizeof(*sub));
408  sub->pts = pts;
410 
411  // Blank if last object_count was 0.
412  if (!ctx->presentation.object_count)
413  return 1;
414 
415  sub->start_display_time = 0;
416  sub->end_display_time = 20000;
417  sub->format = 0;
418 
419  sub->num_rects = ctx->presentation.object_count;
420  sub->rects = av_mallocz(sizeof(*sub->rects) * sub->num_rects);
421 
422  for (rect = 0; rect < sub->num_rects; ++rect) {
423  uint16_t picture_id = ctx->presentation.objects[rect].picture_id;
424  sub->rects[rect] = av_mallocz(sizeof(*sub->rects[rect]));
425  sub->rects[rect]->x = ctx->presentation.objects[rect].x;
426  sub->rects[rect]->y = ctx->presentation.objects[rect].y;
427  sub->rects[rect]->w = ctx->pictures[picture_id].w;
428  sub->rects[rect]->h = ctx->pictures[picture_id].h;
429  sub->rects[rect]->type = SUBTITLE_BITMAP;
430 
431  /* Process bitmap */
432  sub->rects[rect]->pict.linesize[0] = ctx->pictures[picture_id].w;
433  if (ctx->pictures[picture_id].rle) {
434  if (ctx->pictures[picture_id].rle_remaining_len)
435  av_log(avctx, AV_LOG_ERROR, "RLE data length %u is %u bytes shorter than expected\n",
436  ctx->pictures[picture_id].rle_data_len, ctx->pictures[picture_id].rle_remaining_len);
437  if (decode_rle(avctx, sub, rect, ctx->pictures[picture_id].rle, ctx->pictures[picture_id].rle_data_len) < 0)
438  return 0;
439  }
440 
441  /* Allocate memory for colors */
442  sub->rects[rect]->nb_colors = 256;
443  sub->rects[rect]->pict.data[1] = av_mallocz(AVPALETTE_SIZE);
444 
445  /* Copy the forced flag */
446  sub->rects[rect]->flags = (ctx->presentation.objects[rect].composition & 0x40) != 0 ? AV_SUBTITLE_FLAG_FORCED : 0;
447 
448  if (!ctx->forced_subs_only || ctx->presentation.objects[rect].composition & 0x40)
449  memcpy(sub->rects[rect]->pict.data[1], ctx->clut, sub->rects[rect]->nb_colors * sizeof(uint32_t));
450  }
451 
452  return 1;
453 }
454 
455 static int decode(AVCodecContext *avctx, void *data, int *data_size,
456  AVPacket *avpkt)
457 {
458  const uint8_t *buf = avpkt->data;
459  int buf_size = avpkt->size;
460  AVSubtitle *sub = data;
461 
462  const uint8_t *buf_end;
463  uint8_t segment_type;
464  int segment_length;
465  int i, ret;
466 
467  av_dlog(avctx, "PGS sub packet:\n");
468 
469  for (i = 0; i < buf_size; i++) {
470  av_dlog(avctx, "%02x ", buf[i]);
471  if (i % 16 == 15)
472  av_dlog(avctx, "\n");
473  }
474 
475  if (i & 15)
476  av_dlog(avctx, "\n");
477 
478  *data_size = 0;
479 
480  /* Ensure that we have received at a least a segment code and segment length */
481  if (buf_size < 3)
482  return -1;
483 
484  buf_end = buf + buf_size;
485 
486  /* Step through buffer to identify segments */
487  while (buf < buf_end) {
488  segment_type = bytestream_get_byte(&buf);
489  segment_length = bytestream_get_be16(&buf);
490 
491  av_dlog(avctx, "Segment Length %d, Segment Type %x\n", segment_length, segment_type);
492 
493  if (segment_type != DISPLAY_SEGMENT && segment_length > buf_end - buf)
494  break;
495 
496  switch (segment_type) {
497  case PALETTE_SEGMENT:
498  parse_palette_segment(avctx, buf, segment_length);
499  break;
500  case PICTURE_SEGMENT:
501  parse_picture_segment(avctx, buf, segment_length);
502  break;
504  ret = parse_presentation_segment(avctx, buf, segment_length, sub->pts);
505  if (ret < 0)
506  return ret;
507  break;
508  case WINDOW_SEGMENT:
509  /*
510  * Window Segment Structure (No new information provided):
511  * 2 bytes: Unknown,
512  * 2 bytes: X position of subtitle,
513  * 2 bytes: Y position of subtitle,
514  * 2 bytes: Width of subtitle,
515  * 2 bytes: Height of subtitle.
516  */
517  break;
518  case DISPLAY_SEGMENT:
519  *data_size = display_end_segment(avctx, data, buf, segment_length);
520  break;
521  default:
522  av_log(avctx, AV_LOG_ERROR, "Unknown subtitle segment type 0x%x, length %d\n",
523  segment_type, segment_length);
524  break;
525  }
526 
527  buf += segment_length;
528  }
529 
530  return buf_size;
531 }
532 
533 #define OFFSET(x) offsetof(PGSSubContext, x)
534 #define SD AV_OPT_FLAG_SUBTITLE_PARAM | AV_OPT_FLAG_DECODING_PARAM
535 static const AVOption options[] = {
536  {"forced_subs_only", "Only show forced subtitles", OFFSET(forced_subs_only), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, SD},
537  { NULL },
538 };
539 
540 static const AVClass pgsdec_class = {
541  .class_name = "PGS subtitle decoder",
542  .item_name = av_default_item_name,
543  .option = options,
544  .version = LIBAVUTIL_VERSION_INT,
545 };
546 
548  .name = "pgssub",
549  .long_name = NULL_IF_CONFIG_SMALL("HDMV Presentation Graphic Stream subtitles"),
550  .type = AVMEDIA_TYPE_SUBTITLE,
552  .priv_data_size = sizeof(PGSSubContext),
553  .init = init_decoder,
554  .close = close_decoder,
555  .decode = decode,
556  .priv_class = &pgsdec_class,
557 };