2 * HEVC Annex B format parser
4 * Copyright (C) 2012 - 2013 Guillaume Martres
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "libavutil/common.h"
28 #include "h2645_parse.h"
31 #define START_CODE 0x000001 ///< start_code_prefix_one_3bytes
33 #define IS_IRAP_NAL(nal) (nal->type >= 16 && nal->type <= 23)
35 #define ADVANCED_PARSER CONFIG_HEVC_DECODER
37 typedef struct HEVCParserContext {
51 static int hevc_parse_slice_header(AVCodecParserContext *s, H2645NAL *nal,
52 AVCodecContext *avctx)
54 HEVCParserContext *ctx = s->priv_data;
55 GetBitContext *gb = &nal->gb;
61 get_bits1(gb); // first slice in pic
63 get_bits1(gb); // no output of prior pics
65 pps_id = get_ue_golomb_long(gb);
66 if (pps_id >= HEVC_MAX_PPS_COUNT || !ctx->ps.pps_list[pps_id]) {
67 av_log(avctx, AV_LOG_ERROR, "PPS id out of range: %d\n", pps_id);
68 return AVERROR_INVALIDDATA;
70 pps = (HEVCPPS*)ctx->ps.pps_list[pps_id]->data;
71 sps = (HEVCSPS*)ctx->ps.sps_list[pps->sps_id]->data;
73 /* export the stream parameters */
74 s->coded_width = sps->width;
75 s->coded_height = sps->height;
76 s->width = sps->output_width;
77 s->height = sps->output_height;
78 s->format = sps->pix_fmt;
79 avctx->profile = sps->ptl.general_ptl.profile_idc;
80 avctx->level = sps->ptl.general_ptl.level_idc;
82 /* ignore the rest for now*/
87 static int parse_nal_units(AVCodecParserContext *s, const uint8_t *buf,
88 int buf_size, AVCodecContext *avctx)
90 HEVCParserContext *ctx = s->priv_data;
93 ret = ff_h2645_packet_split(&ctx->pkt, buf, buf_size, avctx, 0, 0,
98 for (i = 0; i < ctx->pkt.nb_nals; i++) {
99 H2645NAL *nal = &ctx->pkt.nals[i];
101 /* ignore everything except parameter sets and VCL NALUs */
103 case HEVC_NAL_VPS: ff_hevc_decode_nal_vps(&nal->gb, avctx, &ctx->ps); break;
104 case HEVC_NAL_SPS: ff_hevc_decode_nal_sps(&nal->gb, avctx, &ctx->ps, 1); break;
105 case HEVC_NAL_PPS: ff_hevc_decode_nal_pps(&nal->gb, avctx, &ctx->ps); break;
106 case HEVC_NAL_TRAIL_R:
107 case HEVC_NAL_TRAIL_N:
110 case HEVC_NAL_STSA_N:
111 case HEVC_NAL_STSA_R:
112 case HEVC_NAL_BLA_W_LP:
113 case HEVC_NAL_BLA_W_RADL:
114 case HEVC_NAL_BLA_N_LP:
115 case HEVC_NAL_IDR_W_RADL:
116 case HEVC_NAL_IDR_N_LP:
117 case HEVC_NAL_CRA_NUT:
118 case HEVC_NAL_RADL_N:
119 case HEVC_NAL_RADL_R:
120 case HEVC_NAL_RASL_N:
121 case HEVC_NAL_RASL_R:
122 if (buf == avctx->extradata) {
123 av_log(avctx, AV_LOG_ERROR, "Invalid NAL unit: %d\n", nal->type);
124 return AVERROR_INVALIDDATA;
126 hevc_parse_slice_header(s, nal, avctx);
136 * Find the end of the current frame in the bitstream.
137 * @return the position of the first byte of the next frame, or END_NOT_FOUND
139 static int hevc_find_frame_end(AVCodecParserContext *s, const uint8_t *buf,
143 ParseContext *pc = s->priv_data;
145 for (i = 0; i < buf_size; i++) {
148 pc->state64 = (pc->state64 << 8) | buf[i];
150 if (((pc->state64 >> 3 * 8) & 0xFFFFFF) != START_CODE)
153 nut = (pc->state64 >> 2 * 8 + 1) & 0x3F;
154 // Beginning of access unit
155 if ((nut >= HEVC_NAL_VPS && nut <= HEVC_NAL_AUD) || nut == HEVC_NAL_SEI_PREFIX ||
156 (nut >= 41 && nut <= 44) || (nut >= 48 && nut <= 55)) {
157 if (pc->frame_start_found) {
158 pc->frame_start_found = 0;
161 } else if (nut <= HEVC_NAL_RASL_R ||
162 (nut >= HEVC_NAL_BLA_W_LP && nut <= HEVC_NAL_CRA_NUT)) {
163 int first_slice_segment_in_pic_flag = buf[i] >> 7;
164 if (first_slice_segment_in_pic_flag) {
165 if (!pc->frame_start_found) {
166 pc->frame_start_found = 1;
167 } else { // First slice of next frame found
168 pc->frame_start_found = 0;
175 return END_NOT_FOUND;
180 * Parse NAL units of found picture and decode some basic information.
182 * @param s parser context.
183 * @param avctx codec context.
184 * @param buf buffer with field/frame data.
185 * @param buf_size size of the buffer.
187 static inline int parse_nal_units(AVCodecParserContext *s, const uint8_t *buf,
188 int buf_size, AVCodecContext *avctx)
190 HEVCParserContext *ctx = s->priv_data;
191 HEVCContext *h = &ctx->h;
193 SliceHeader *sh = &h->sh;
194 HEVCParamSets *ps = &h->ps;
195 H2645Packet *pkt = &ctx->pkt;
196 const uint8_t *buf_end = buf + buf_size;
199 int is_global = buf == avctx->extradata;
202 h->HEVClc = av_mallocz(sizeof(HEVCLocalContext));
204 return AVERROR(ENOMEM);
208 /* set some sane default values */
209 s->pict_type = AV_PICTURE_TYPE_I;
211 s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
215 ff_hevc_reset_sei(h);
220 if (pkt->nals_allocated < 1) {
221 H2645NAL *tmp = av_realloc_array(pkt->nals, 1, sizeof(*tmp));
223 return AVERROR(ENOMEM);
225 memset(pkt->nals, 0, sizeof(*tmp));
226 pkt->nals_allocated = 1;
232 int src_length, consumed;
234 int num = 0, den = 0;
235 buf = avpriv_find_start_code(buf, buf_end, &state);
236 if (--buf + 2 >= buf_end)
238 src_length = buf_end - buf;
240 h->nal_unit_type = (*buf >> 1) & 0x3f;
241 h->temporal_id = (*(buf + 1) & 0x07) - 1;
242 if (h->nal_unit_type <= HEVC_NAL_CRA_NUT) {
243 // Do not walk the whole buffer just to decode slice segment header
248 consumed = ff_h2645_extract_rbsp(buf, src_length, nal, 1);
252 ret = init_get_bits8(gb, nal->data + 2, nal->size);
256 switch (h->nal_unit_type) {
258 ff_hevc_decode_nal_vps(gb, avctx, ps);
261 ff_hevc_decode_nal_sps(gb, avctx, ps, 1);
264 ff_hevc_decode_nal_pps(gb, avctx, ps);
266 case HEVC_NAL_SEI_PREFIX:
267 case HEVC_NAL_SEI_SUFFIX:
268 ff_hevc_decode_nal_sei(h);
270 case HEVC_NAL_TRAIL_N:
271 case HEVC_NAL_TRAIL_R:
274 case HEVC_NAL_STSA_N:
275 case HEVC_NAL_STSA_R:
276 case HEVC_NAL_RADL_N:
277 case HEVC_NAL_RADL_R:
278 case HEVC_NAL_RASL_N:
279 case HEVC_NAL_RASL_R:
280 case HEVC_NAL_BLA_W_LP:
281 case HEVC_NAL_BLA_W_RADL:
282 case HEVC_NAL_BLA_N_LP:
283 case HEVC_NAL_IDR_W_RADL:
284 case HEVC_NAL_IDR_N_LP:
285 case HEVC_NAL_CRA_NUT:
288 av_log(avctx, AV_LOG_ERROR, "Invalid NAL unit: %d\n", h->nal_unit_type);
289 return AVERROR_INVALIDDATA;
292 sh->first_slice_in_pic_flag = get_bits1(gb);
293 s->picture_structure = h->picture_struct;
294 s->field_order = h->picture_struct;
298 sh->no_output_of_prior_pics_flag = get_bits1(gb);
301 sh->pps_id = get_ue_golomb(gb);
302 if (sh->pps_id >= HEVC_MAX_PPS_COUNT || !ps->pps_list[sh->pps_id]) {
303 av_log(avctx, AV_LOG_ERROR, "PPS id out of range: %d\n", sh->pps_id);
304 return AVERROR_INVALIDDATA;
306 ps->pps = (HEVCPPS*)ps->pps_list[sh->pps_id]->data;
308 if (ps->pps->sps_id >= HEVC_MAX_SPS_COUNT || !ps->sps_list[ps->pps->sps_id]) {
309 av_log(avctx, AV_LOG_ERROR, "SPS id out of range: %d\n", ps->pps->sps_id);
310 return AVERROR_INVALIDDATA;
312 if (ps->sps != (HEVCSPS*)ps->sps_list[ps->pps->sps_id]->data) {
313 ps->sps = (HEVCSPS*)ps->sps_list[ps->pps->sps_id]->data;
314 ps->vps = (HEVCVPS*)ps->vps_list[ps->sps->vps_id]->data;
317 s->coded_width = ps->sps->width;
318 s->coded_height = ps->sps->height;
319 s->width = ps->sps->output_width;
320 s->height = ps->sps->output_height;
321 s->format = ps->sps->pix_fmt;
322 avctx->profile = ps->sps->ptl.general_ptl.profile_idc;
323 avctx->level = ps->sps->ptl.general_ptl.level_idc;
325 if (ps->vps->vps_timing_info_present_flag) {
326 num = ps->vps->vps_num_units_in_tick;
327 den = ps->vps->vps_time_scale;
328 } else if (ps->sps->vui.vui_timing_info_present_flag) {
329 num = ps->sps->vui.vui_num_units_in_tick;
330 den = ps->sps->vui.vui_time_scale;
333 if (num != 0 && den != 0)
334 av_reduce(&avctx->framerate.den, &avctx->framerate.num,
337 if (!sh->first_slice_in_pic_flag) {
338 int slice_address_length;
340 if (ps->pps->dependent_slice_segments_enabled_flag)
341 sh->dependent_slice_segment_flag = get_bits1(gb);
343 sh->dependent_slice_segment_flag = 0;
345 slice_address_length = av_ceil_log2_c(ps->sps->ctb_width *
346 ps->sps->ctb_height);
347 sh->slice_segment_addr = get_bitsz(gb, slice_address_length);
348 if (sh->slice_segment_addr >= ps->sps->ctb_width * ps->sps->ctb_height) {
349 av_log(avctx, AV_LOG_ERROR, "Invalid slice segment address: %u.\n",
350 sh->slice_segment_addr);
351 return AVERROR_INVALIDDATA;
354 sh->dependent_slice_segment_flag = 0;
356 if (sh->dependent_slice_segment_flag)
359 for (i = 0; i < ps->pps->num_extra_slice_header_bits; i++)
360 skip_bits(gb, 1); // slice_reserved_undetermined_flag[]
362 sh->slice_type = get_ue_golomb(gb);
363 if (!(sh->slice_type == HEVC_SLICE_I || sh->slice_type == HEVC_SLICE_P ||
364 sh->slice_type == HEVC_SLICE_B)) {
365 av_log(avctx, AV_LOG_ERROR, "Unknown slice type: %d.\n",
367 return AVERROR_INVALIDDATA;
369 s->pict_type = sh->slice_type == HEVC_SLICE_B ? AV_PICTURE_TYPE_B :
370 sh->slice_type == HEVC_SLICE_P ? AV_PICTURE_TYPE_P :
373 if (ps->pps->output_flag_present_flag)
374 sh->pic_output_flag = get_bits1(gb);
376 if (ps->sps->separate_colour_plane_flag)
377 sh->colour_plane_id = get_bits(gb, 2);
380 sh->pic_order_cnt_lsb = get_bits(gb, ps->sps->log2_max_poc_lsb);
381 s->output_picture_number = h->poc = ff_hevc_compute_poc(h, sh->pic_order_cnt_lsb);
383 s->output_picture_number = h->poc = 0;
385 if (h->temporal_id == 0 &&
386 h->nal_unit_type != HEVC_NAL_TRAIL_N &&
387 h->nal_unit_type != HEVC_NAL_TSA_N &&
388 h->nal_unit_type != HEVC_NAL_STSA_N &&
389 h->nal_unit_type != HEVC_NAL_RADL_N &&
390 h->nal_unit_type != HEVC_NAL_RASL_N &&
391 h->nal_unit_type != HEVC_NAL_RADL_R &&
392 h->nal_unit_type != HEVC_NAL_RASL_R)
395 return 0; /* no need to evaluate the rest */
399 /* didn't find a picture! */
401 av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit\n");
406 static int hevc_parse(AVCodecParserContext *s,
407 AVCodecContext *avctx,
408 const uint8_t **poutbuf, int *poutbuf_size,
409 const uint8_t *buf, int buf_size)
412 HEVCParserContext *ctx = s->priv_data;
413 ParseContext *pc = &ctx->pc;
415 if (avctx->extradata && !ctx->parsed_extradata) {
416 parse_nal_units(s, avctx->extradata, avctx->extradata_size, avctx);
417 ctx->parsed_extradata = 1;
420 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
423 next = hevc_find_frame_end(s, buf, buf_size);
424 if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
431 parse_nal_units(s, buf, buf_size, avctx);
434 *poutbuf_size = buf_size;
438 // Split after the parameter sets at the beginning of the stream if they exist.
439 static int hevc_split(AVCodecContext *avctx, const uint8_t *buf, int buf_size)
441 const uint8_t *ptr = buf, *end = buf + buf_size;
449 ptr = avpriv_find_start_code(ptr, end, &state);
450 if ((state >> 8) != START_CODE)
452 nut = (state >> 1) & 0x3F;
453 if (nut == HEVC_NAL_VPS)
455 else if (nut == HEVC_NAL_SPS)
457 else if (nut == HEVC_NAL_PPS)
459 else if ((nut != HEVC_NAL_SEI_PREFIX || has_pps) &&
460 nut != HEVC_NAL_AUD) {
461 if (has_vps && has_sps) {
462 while (ptr - 4 > buf && ptr[-5] == 0)
464 return ptr - 4 - buf;
471 static void hevc_parser_close(AVCodecParserContext *s)
473 HEVCParserContext *ctx = s->priv_data;
477 HEVCContext *h = &ctx->h;
479 for (i = 0; i < FF_ARRAY_ELEMS(h->ps.vps_list); i++)
480 av_buffer_unref(&h->ps.vps_list[i]);
481 for (i = 0; i < FF_ARRAY_ELEMS(h->ps.sps_list); i++)
482 av_buffer_unref(&h->ps.sps_list[i]);
483 for (i = 0; i < FF_ARRAY_ELEMS(h->ps.pps_list); i++)
484 av_buffer_unref(&h->ps.pps_list[i]);
488 av_freep(&h->HEVClc);
491 for (i = 0; i < FF_ARRAY_ELEMS(ctx->ps.vps_list); i++)
492 av_buffer_unref(&ctx->ps.vps_list[i]);
493 for (i = 0; i < FF_ARRAY_ELEMS(ctx->ps.sps_list); i++)
494 av_buffer_unref(&ctx->ps.sps_list[i]);
495 for (i = 0; i < FF_ARRAY_ELEMS(ctx->ps.pps_list); i++)
496 av_buffer_unref(&ctx->ps.pps_list[i]);
500 ff_h2645_packet_uninit(&ctx->pkt);
502 av_freep(&ctx->pc.buffer);
505 AVCodecParser ff_hevc_parser = {
506 .codec_ids = { AV_CODEC_ID_HEVC },
507 .priv_data_size = sizeof(HEVCParserContext),
508 .parser_parse = hevc_parse,
509 .parser_close = hevc_parser_close,