3 * Copyright (c) 2009 Stephen Backway
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * PGS subtitle decoder
28 #include "bytestream.h"
32 #include "libavutil/colorspace.h"
33 #include "libavutil/imgutils.h"
35 #define RGBA(r,g,b,a) (((a) << 24) | ((r) << 16) | ((g) << 8) | (b))
36 #define MAX_EPOCH_PALETTES 8 // Max 8 allowed per PGS epoch
37 #define MAX_EPOCH_OBJECTS 64 // Max 64 allowed per PGS epoch
38 #define MAX_OBJECT_REFS 2 // Max objects per display set
41 PALETTE_SEGMENT = 0x14,
42 OBJECT_SEGMENT = 0x15,
43 PRESENTATION_SEGMENT = 0x16,
44 WINDOW_SEGMENT = 0x17,
45 DISPLAY_SEGMENT = 0x80,
48 typedef struct PGSSubObjectRef {
51 uint8_t composition_flag;
60 typedef struct PGSSubPresentation {
64 PGSSubObjectRef objects[MAX_OBJECT_REFS];
68 typedef struct PGSSubObject {
73 unsigned int rle_buffer_size, rle_data_len;
74 unsigned int rle_remaining_len;
77 typedef struct PGSSubObjects {
79 PGSSubObject object[MAX_EPOCH_OBJECTS];
82 typedef struct PGSSubPalette {
87 typedef struct PGSSubPalettes {
89 PGSSubPalette palette[MAX_EPOCH_PALETTES];
92 typedef struct PGSSubContext {
93 PGSSubPresentation presentation;
94 PGSSubPalettes palettes;
95 PGSSubObjects objects;
98 static void flush_cache(AVCodecContext *avctx)
100 PGSSubContext *ctx = avctx->priv_data;
103 for (i = 0; i < ctx->objects.count; i++) {
104 av_freep(&ctx->objects.object[i].rle);
105 ctx->objects.object[i].rle_buffer_size = 0;
106 ctx->objects.object[i].rle_remaining_len = 0;
108 ctx->objects.count = 0;
109 ctx->palettes.count = 0;
112 static PGSSubObject * find_object(int id, PGSSubObjects *objects)
116 for (i = 0; i < objects->count; i++) {
117 if (objects->object[i].id == id)
118 return &objects->object[i];
123 static PGSSubPalette * find_palette(int id, PGSSubPalettes *palettes)
127 for (i = 0; i < palettes->count; i++) {
128 if (palettes->palette[i].id == id)
129 return &palettes->palette[i];
134 static av_cold int init_decoder(AVCodecContext *avctx)
136 avctx->pix_fmt = AV_PIX_FMT_PAL8;
141 static av_cold int close_decoder(AVCodecContext *avctx)
149 * Decode the RLE data.
151 * The subtitle is stored as an Run Length Encoded image.
153 * @param avctx contains the current codec context
154 * @param sub pointer to the processed subtitle data
155 * @param buf pointer to the RLE data to process
156 * @param buf_size size of the RLE data to process
158 static int decode_rle(AVCodecContext *avctx, AVSubtitleRect *rect,
159 const uint8_t *buf, unsigned int buf_size)
161 const uint8_t *rle_bitmap_end;
162 int pixel_count, line_count;
164 rle_bitmap_end = buf + buf_size;
166 rect->pict.data[0] = av_malloc(rect->w * rect->h);
168 if (!rect->pict.data[0])
169 return AVERROR(ENOMEM);
174 while (buf < rle_bitmap_end && line_count < rect->h) {
175 uint8_t flags, color;
178 color = bytestream_get_byte(&buf);
182 flags = bytestream_get_byte(&buf);
185 run = (run << 8) + bytestream_get_byte(&buf);
186 color = flags & 0x80 ? bytestream_get_byte(&buf) : 0;
189 if (run > 0 && pixel_count + run <= rect->w * rect->h) {
190 memset(rect->pict.data[0] + pixel_count, color, run);
194 * New Line. Check if correct pixels decoded, if not display warning
195 * and adjust bitmap pointer to correct new line position.
197 if (pixel_count % rect->w > 0) {
198 av_log(avctx, AV_LOG_ERROR, "Decoded %d pixels, when line should be %d pixels\n",
199 pixel_count % rect->w, rect->w);
200 if (avctx->err_recognition & AV_EF_EXPLODE) {
201 return AVERROR_INVALIDDATA;
208 if (pixel_count < rect->w * rect->h) {
209 av_log(avctx, AV_LOG_ERROR, "Insufficient RLE data for subtitle\n");
210 return AVERROR_INVALIDDATA;
213 av_dlog(avctx, "Pixel Count = %d, Area = %d\n", pixel_count, rect->w * rect->h);
219 * Parse the picture segment packet.
221 * The picture segment contains details on the sequence id,
222 * width, height and Run Length Encoded (RLE) bitmap data.
224 * @param avctx contains the current codec context
225 * @param buf pointer to the packet to process
226 * @param buf_size size of packet to process
228 static int parse_object_segment(AVCodecContext *avctx,
229 const uint8_t *buf, int buf_size)
231 PGSSubContext *ctx = avctx->priv_data;
232 PGSSubObject *object;
234 uint8_t sequence_desc;
235 unsigned int rle_bitmap_len, width, height;
239 return AVERROR_INVALIDDATA;
242 id = bytestream_get_be16(&buf);
243 object = find_object(id, &ctx->objects);
245 if (ctx->objects.count >= MAX_EPOCH_OBJECTS) {
246 av_log(avctx, AV_LOG_ERROR, "Too many objects in epoch\n");
247 return AVERROR_INVALIDDATA;
249 object = &ctx->objects.object[ctx->objects.count++];
253 /* skip object version number */
256 /* Read the Sequence Description to determine if start of RLE data or appended to previous RLE */
257 sequence_desc = bytestream_get_byte(&buf);
259 if (!(sequence_desc & 0x80)) {
260 /* Additional RLE data */
261 if (buf_size > object->rle_remaining_len)
262 return AVERROR_INVALIDDATA;
264 memcpy(object->rle + object->rle_data_len, buf, buf_size);
265 object->rle_data_len += buf_size;
266 object->rle_remaining_len -= buf_size;
272 return AVERROR_INVALIDDATA;
275 /* Decode rle bitmap length, stored size includes width/height data */
276 rle_bitmap_len = bytestream_get_be24(&buf) - 2*2;
278 /* Get bitmap dimensions from data */
279 width = bytestream_get_be16(&buf);
280 height = bytestream_get_be16(&buf);
282 /* Make sure the bitmap is not too large */
283 if (avctx->width < width || avctx->height < height) {
284 av_log(avctx, AV_LOG_ERROR, "Bitmap dimensions larger than video.\n");
285 return AVERROR_INVALIDDATA;
291 av_fast_malloc(&object->rle, &object->rle_buffer_size, rle_bitmap_len);
294 return AVERROR(ENOMEM);
296 memcpy(object->rle, buf, buf_size);
297 object->rle_data_len = buf_size;
298 object->rle_remaining_len = rle_bitmap_len - buf_size;
304 * Parse the palette segment packet.
306 * The palette segment contains details of the palette,
307 * a maximum of 256 colors can be defined.
309 * @param avctx contains the current codec context
310 * @param buf pointer to the packet to process
311 * @param buf_size size of packet to process
313 static int parse_palette_segment(AVCodecContext *avctx,
314 const uint8_t *buf, int buf_size)
316 PGSSubContext *ctx = avctx->priv_data;
317 PGSSubPalette *palette;
319 const uint8_t *buf_end = buf + buf_size;
320 const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
322 int y, cb, cr, alpha;
323 int r, g, b, r_add, g_add, b_add;
326 id = bytestream_get_byte(&buf);
327 palette = find_palette(id, &ctx->palettes);
329 if (ctx->palettes.count >= MAX_EPOCH_PALETTES) {
330 av_log(avctx, AV_LOG_ERROR, "Too many palettes in epoch\n");
331 return AVERROR_INVALIDDATA;
333 palette = &ctx->palettes.palette[ctx->palettes.count++];
337 /* Skip palette version */
340 while (buf < buf_end) {
341 color_id = bytestream_get_byte(&buf);
342 y = bytestream_get_byte(&buf);
343 cr = bytestream_get_byte(&buf);
344 cb = bytestream_get_byte(&buf);
345 alpha = bytestream_get_byte(&buf);
348 YUV_TO_RGB2(r, g, b, y);
350 av_dlog(avctx, "Color %d := (%d,%d,%d,%d)\n", color_id, r, g, b, alpha);
352 /* Store color in palette */
353 palette->clut[color_id] = RGBA(r,g,b,alpha);
359 * Parse the presentation segment packet.
361 * The presentation segment contains details on the video
362 * width, video height, x & y subtitle position.
364 * @param avctx contains the current codec context
365 * @param buf pointer to the packet to process
366 * @param buf_size size of packet to process
367 * @todo TODO: Implement cropping
369 static int parse_presentation_segment(AVCodecContext *avctx,
370 const uint8_t *buf, int buf_size,
373 PGSSubContext *ctx = avctx->priv_data;
378 int w = bytestream_get_be16(&buf);
379 int h = bytestream_get_be16(&buf);
381 ctx->presentation.pts = pts;
383 av_dlog(avctx, "Video Dimensions %dx%d\n",
385 ret = ff_set_dimensions(avctx, w, h);
389 /* Skip 1 bytes of unknown, frame rate */
392 // Composition descriptor
393 ctx->presentation.id_number = bytestream_get_be16(&buf);
395 * state is a 2 bit field that defines pgs epoch boundaries
396 * 00 - Normal, previously defined objects and palettes are still valid
397 * 01 - Acquisition point, previous objects and palettes can be released
398 * 10 - Epoch start, previous objects and palettes can be released
399 * 11 - Epoch continue, previous objects and palettes can be released
401 * reserved 6 bits discarded
403 state = bytestream_get_byte(&buf) >> 6;
409 * skip palette_update_flag (0x80),
412 ctx->presentation.palette_id = bytestream_get_byte(&buf);
413 ctx->presentation.object_count = bytestream_get_byte(&buf);
414 if (ctx->presentation.object_count > MAX_OBJECT_REFS) {
415 av_log(avctx, AV_LOG_ERROR,
416 "Invalid number of presentation objects %d\n",
417 ctx->presentation.object_count);
418 ctx->presentation.object_count = 2;
419 if (avctx->err_recognition & AV_EF_EXPLODE) {
420 return AVERROR_INVALIDDATA;
424 for (i = 0; i < ctx->presentation.object_count; i++)
426 ctx->presentation.objects[i].id = bytestream_get_be16(&buf);
427 ctx->presentation.objects[i].window_id = bytestream_get_byte(&buf);
428 ctx->presentation.objects[i].composition_flag = bytestream_get_byte(&buf);
430 ctx->presentation.objects[i].x = bytestream_get_be16(&buf);
431 ctx->presentation.objects[i].y = bytestream_get_be16(&buf);
434 if (ctx->presentation.objects[i].composition_flag & 0x80) {
435 ctx->presentation.objects[i].crop_x = bytestream_get_be16(&buf);
436 ctx->presentation.objects[i].crop_y = bytestream_get_be16(&buf);
437 ctx->presentation.objects[i].crop_w = bytestream_get_be16(&buf);
438 ctx->presentation.objects[i].crop_h = bytestream_get_be16(&buf);
441 av_dlog(avctx, "Subtitle Placement x=%d, y=%d\n",
442 ctx->presentation.objects[i].x, ctx->presentation.objects[i].y);
444 if (ctx->presentation.objects[i].x > avctx->width ||
445 ctx->presentation.objects[i].y > avctx->height) {
446 av_log(avctx, AV_LOG_ERROR, "Subtitle out of video bounds. x = %d, y = %d, video width = %d, video height = %d.\n",
447 ctx->presentation.objects[i].x,
448 ctx->presentation.objects[i].y,
449 avctx->width, avctx->height);
450 ctx->presentation.objects[i].x = 0;
451 ctx->presentation.objects[i].y = 0;
452 if (avctx->err_recognition & AV_EF_EXPLODE) {
453 return AVERROR_INVALIDDATA;
462 * Parse the display segment packet.
464 * The display segment controls the updating of the display.
466 * @param avctx contains the current codec context
467 * @param data pointer to the data pertaining the subtitle to display
468 * @param buf pointer to the packet to process
469 * @param buf_size size of packet to process
471 static int display_end_segment(AVCodecContext *avctx, void *data,
472 const uint8_t *buf, int buf_size)
474 AVSubtitle *sub = data;
475 PGSSubContext *ctx = avctx->priv_data;
476 PGSSubPalette *palette;
479 memset(sub, 0, sizeof(*sub));
480 sub->pts = ctx->presentation.pts;
481 sub->start_display_time = 0;
482 // There is no explicit end time for PGS subtitles. The end time
483 // is defined by the start of the next sub which may contain no
484 // objects (i.e. clears the previous sub)
485 sub->end_display_time = UINT32_MAX;
488 // Blank if last object_count was 0.
489 if (!ctx->presentation.object_count)
491 sub->rects = av_mallocz(sizeof(*sub->rects) * ctx->presentation.object_count);
493 return AVERROR(ENOMEM);
495 palette = find_palette(ctx->presentation.palette_id, &ctx->palettes);
497 // Missing palette. Should only happen with damaged streams.
498 av_log(avctx, AV_LOG_ERROR, "Invalid palette id %d\n",
499 ctx->presentation.palette_id);
500 avsubtitle_free(sub);
501 return AVERROR_INVALIDDATA;
503 for (i = 0; i < ctx->presentation.object_count; i++) {
504 PGSSubObject *object;
506 sub->rects[i] = av_mallocz(sizeof(*sub->rects[0]));
507 if (!sub->rects[i]) {
508 avsubtitle_free(sub);
509 return AVERROR(ENOMEM);
512 sub->rects[i]->type = SUBTITLE_BITMAP;
515 object = find_object(ctx->presentation.objects[i].id, &ctx->objects);
517 // Missing object. Should only happen with damaged streams.
518 av_log(avctx, AV_LOG_ERROR, "Invalid object id %d\n",
519 ctx->presentation.objects[i].id);
520 if (avctx->err_recognition & AV_EF_EXPLODE) {
521 avsubtitle_free(sub);
522 return AVERROR_INVALIDDATA;
524 // Leaves rect empty with 0 width and height.
527 if (ctx->presentation.objects[i].composition_flag & 0x40)
528 sub->rects[i]->flags |= AV_SUBTITLE_FLAG_FORCED;
530 sub->rects[i]->x = ctx->presentation.objects[i].x;
531 sub->rects[i]->y = ctx->presentation.objects[i].y;
532 sub->rects[i]->w = object->w;
533 sub->rects[i]->h = object->h;
535 sub->rects[i]->pict.linesize[0] = object->w;
538 if (object->rle_remaining_len) {
539 av_log(avctx, AV_LOG_ERROR, "RLE data length %u is %u bytes shorter than expected\n",
540 object->rle_data_len, object->rle_remaining_len);
541 if (avctx->err_recognition & AV_EF_EXPLODE) {
542 avsubtitle_free(sub);
543 return AVERROR_INVALIDDATA;
546 ret = decode_rle(avctx, sub->rects[i], object->rle, object->rle_data_len);
548 if ((avctx->err_recognition & AV_EF_EXPLODE) ||
549 ret == AVERROR(ENOMEM)) {
550 avsubtitle_free(sub);
553 sub->rects[i]->w = 0;
554 sub->rects[i]->h = 0;
558 /* Allocate memory for colors */
559 sub->rects[i]->nb_colors = 256;
560 sub->rects[i]->pict.data[1] = av_mallocz(AVPALETTE_SIZE);
561 if (!sub->rects[i]->pict.data[1]) {
562 avsubtitle_free(sub);
563 return AVERROR(ENOMEM);
566 memcpy(sub->rects[i]->pict.data[1], palette->clut, sub->rects[i]->nb_colors * sizeof(uint32_t));
572 static int decode(AVCodecContext *avctx, void *data, int *data_size,
575 const uint8_t *buf = avpkt->data;
576 int buf_size = avpkt->size;
578 const uint8_t *buf_end;
579 uint8_t segment_type;
583 av_dlog(avctx, "PGS sub packet:\n");
585 for (i = 0; i < buf_size; i++) {
586 av_dlog(avctx, "%02x ", buf[i]);
588 av_dlog(avctx, "\n");
592 av_dlog(avctx, "\n");
596 /* Ensure that we have received at a least a segment code and segment length */
600 buf_end = buf + buf_size;
602 /* Step through buffer to identify segments */
603 while (buf < buf_end) {
604 segment_type = bytestream_get_byte(&buf);
605 segment_length = bytestream_get_be16(&buf);
607 av_dlog(avctx, "Segment Length %d, Segment Type %x\n", segment_length, segment_type);
609 if (segment_type != DISPLAY_SEGMENT && segment_length > buf_end - buf)
613 switch (segment_type) {
614 case PALETTE_SEGMENT:
615 ret = parse_palette_segment(avctx, buf, segment_length);
618 ret = parse_object_segment(avctx, buf, segment_length);
620 case PRESENTATION_SEGMENT:
621 ret = parse_presentation_segment(avctx, buf, segment_length, avpkt->pts);
625 * Window Segment Structure (No new information provided):
627 * 2 bytes: X position of subtitle,
628 * 2 bytes: Y position of subtitle,
629 * 2 bytes: Width of subtitle,
630 * 2 bytes: Height of subtitle.
633 case DISPLAY_SEGMENT:
634 ret = display_end_segment(avctx, data, buf, segment_length);
639 av_log(avctx, AV_LOG_ERROR, "Unknown subtitle segment type 0x%x, length %d\n",
640 segment_type, segment_length);
641 ret = AVERROR_INVALIDDATA;
644 if (ret < 0 && (avctx->err_recognition & AV_EF_EXPLODE))
647 buf += segment_length;
653 AVCodec ff_pgssub_decoder = {
655 .long_name = NULL_IF_CONFIG_SMALL("HDMV Presentation Graphic Stream subtitles"),
656 .type = AVMEDIA_TYPE_SUBTITLE,
657 .id = AV_CODEC_ID_HDMV_PGS_SUBTITLE,
658 .priv_data_size = sizeof(PGSSubContext),
659 .init = init_decoder,
660 .close = close_decoder,