| 1 | /* |
| 2 | * HEVC Annex B format parser |
| 3 | * |
| 4 | * Copyright (C) 2012 - 2013 Guillaume Martres |
| 5 | * |
| 6 | * This file is part of FFmpeg. |
| 7 | * |
| 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. |
| 12 | * |
| 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. |
| 17 | * |
| 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 |
| 21 | */ |
| 22 | |
| 23 | #include "libavutil/common.h" |
| 24 | |
| 25 | #include "parser.h" |
| 26 | #include "hevc.h" |
| 27 | #include "golomb.h" |
| 28 | |
| 29 | #define START_CODE 0x000001 ///< start_code_prefix_one_3bytes |
| 30 | |
| 31 | typedef struct HEVCParseContext { |
| 32 | HEVCContext h; |
| 33 | ParseContext pc; |
| 34 | } HEVCParseContext; |
| 35 | |
| 36 | /** |
| 37 | * Find the end of the current frame in the bitstream. |
| 38 | * @return the position of the first byte of the next frame, or END_NOT_FOUND |
| 39 | */ |
| 40 | static int hevc_find_frame_end(AVCodecParserContext *s, const uint8_t *buf, |
| 41 | int buf_size) |
| 42 | { |
| 43 | int i; |
| 44 | ParseContext *pc = &((HEVCParseContext *)s->priv_data)->pc; |
| 45 | |
| 46 | for (i = 0; i < buf_size; i++) { |
| 47 | int nut; |
| 48 | |
| 49 | pc->state64 = (pc->state64 << 8) | buf[i]; |
| 50 | |
| 51 | if (((pc->state64 >> 3 * 8) & 0xFFFFFF) != START_CODE) |
| 52 | continue; |
| 53 | |
| 54 | nut = (pc->state64 >> 2 * 8 + 1) & 0x3F; |
| 55 | // Beginning of access unit |
| 56 | if ((nut >= NAL_VPS && nut <= NAL_AUD) || nut == NAL_SEI_PREFIX || |
| 57 | (nut >= 41 && nut <= 44) || (nut >= 48 && nut <= 55)) { |
| 58 | if (pc->frame_start_found) { |
| 59 | pc->frame_start_found = 0; |
| 60 | return i - 5; |
| 61 | } |
| 62 | } else if (nut <= NAL_RASL_R || |
| 63 | (nut >= NAL_BLA_W_LP && nut <= NAL_CRA_NUT)) { |
| 64 | int first_slice_segment_in_pic_flag = buf[i] >> 7; |
| 65 | if (first_slice_segment_in_pic_flag) { |
| 66 | if (!pc->frame_start_found) { |
| 67 | pc->frame_start_found = 1; |
| 68 | } else { // First slice of next frame found |
| 69 | pc->frame_start_found = 0; |
| 70 | return i - 5; |
| 71 | } |
| 72 | } |
| 73 | } |
| 74 | } |
| 75 | |
| 76 | return END_NOT_FOUND; |
| 77 | } |
| 78 | |
| 79 | /** |
| 80 | * Parse NAL units of found picture and decode some basic information. |
| 81 | * |
| 82 | * @param s parser context. |
| 83 | * @param avctx codec context. |
| 84 | * @param buf buffer with field/frame data. |
| 85 | * @param buf_size size of the buffer. |
| 86 | */ |
| 87 | static inline int parse_nal_units(AVCodecParserContext *s, AVCodecContext *avctx, |
| 88 | const uint8_t *buf, int buf_size) |
| 89 | { |
| 90 | HEVCContext *h = &((HEVCParseContext *)s->priv_data)->h; |
| 91 | GetBitContext *gb = &h->HEVClc->gb; |
| 92 | SliceHeader *sh = &h->sh; |
| 93 | const uint8_t *buf_end = buf + buf_size; |
| 94 | int state = -1, i; |
| 95 | HEVCNAL *nal; |
| 96 | |
| 97 | /* set some sane default values */ |
| 98 | s->pict_type = AV_PICTURE_TYPE_I; |
| 99 | s->key_frame = 0; |
| 100 | s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN; |
| 101 | |
| 102 | h->avctx = avctx; |
| 103 | |
| 104 | if (!buf_size) |
| 105 | return 0; |
| 106 | |
| 107 | if (h->nals_allocated < 1) { |
| 108 | HEVCNAL *tmp = av_realloc_array(h->nals, 1, sizeof(*tmp)); |
| 109 | if (!tmp) |
| 110 | return AVERROR(ENOMEM); |
| 111 | h->nals = tmp; |
| 112 | memset(h->nals, 0, sizeof(*tmp)); |
| 113 | h->nals_allocated = 1; |
| 114 | } |
| 115 | |
| 116 | nal = &h->nals[0]; |
| 117 | |
| 118 | for (;;) { |
| 119 | int src_length, consumed; |
| 120 | buf = avpriv_find_start_code(buf, buf_end, &state); |
| 121 | if (--buf + 2 >= buf_end) |
| 122 | break; |
| 123 | src_length = buf_end - buf; |
| 124 | |
| 125 | h->nal_unit_type = (*buf >> 1) & 0x3f; |
| 126 | h->temporal_id = (*(buf + 1) & 0x07) - 1; |
| 127 | if (h->nal_unit_type <= NAL_CRA_NUT) { |
| 128 | // Do not walk the whole buffer just to decode slice segment header |
| 129 | if (src_length > 20) |
| 130 | src_length = 20; |
| 131 | } |
| 132 | |
| 133 | consumed = ff_hevc_extract_rbsp(h, buf, src_length, nal); |
| 134 | if (consumed < 0) |
| 135 | return consumed; |
| 136 | |
| 137 | init_get_bits8(gb, nal->data + 2, nal->size); |
| 138 | switch (h->nal_unit_type) { |
| 139 | case NAL_VPS: |
| 140 | ff_hevc_decode_nal_vps(h); |
| 141 | break; |
| 142 | case NAL_SPS: |
| 143 | ff_hevc_decode_nal_sps(h); |
| 144 | break; |
| 145 | case NAL_PPS: |
| 146 | ff_hevc_decode_nal_pps(h); |
| 147 | break; |
| 148 | case NAL_SEI_PREFIX: |
| 149 | case NAL_SEI_SUFFIX: |
| 150 | ff_hevc_decode_nal_sei(h); |
| 151 | break; |
| 152 | case NAL_TRAIL_N: |
| 153 | case NAL_TRAIL_R: |
| 154 | case NAL_TSA_N: |
| 155 | case NAL_TSA_R: |
| 156 | case NAL_STSA_N: |
| 157 | case NAL_STSA_R: |
| 158 | case NAL_RADL_N: |
| 159 | case NAL_RADL_R: |
| 160 | case NAL_RASL_N: |
| 161 | case NAL_RASL_R: |
| 162 | case NAL_BLA_W_LP: |
| 163 | case NAL_BLA_W_RADL: |
| 164 | case NAL_BLA_N_LP: |
| 165 | case NAL_IDR_W_RADL: |
| 166 | case NAL_IDR_N_LP: |
| 167 | case NAL_CRA_NUT: |
| 168 | sh->first_slice_in_pic_flag = get_bits1(gb); |
| 169 | s->picture_structure = h->picture_struct; |
| 170 | s->field_order = h->picture_struct; |
| 171 | |
| 172 | if (IS_IRAP(h)) { |
| 173 | s->key_frame = 1; |
| 174 | sh->no_output_of_prior_pics_flag = get_bits1(gb); |
| 175 | } |
| 176 | |
| 177 | sh->pps_id = get_ue_golomb(gb); |
| 178 | if (sh->pps_id >= MAX_PPS_COUNT || !h->pps_list[sh->pps_id]) { |
| 179 | av_log(h->avctx, AV_LOG_ERROR, "PPS id out of range: %d\n", sh->pps_id); |
| 180 | return AVERROR_INVALIDDATA; |
| 181 | } |
| 182 | h->pps = (HEVCPPS*)h->pps_list[sh->pps_id]->data; |
| 183 | |
| 184 | if (h->pps->sps_id >= MAX_SPS_COUNT || !h->sps_list[h->pps->sps_id]) { |
| 185 | av_log(h->avctx, AV_LOG_ERROR, "SPS id out of range: %d\n", h->pps->sps_id); |
| 186 | return AVERROR_INVALIDDATA; |
| 187 | } |
| 188 | if (h->sps != (HEVCSPS*)h->sps_list[h->pps->sps_id]->data) { |
| 189 | h->sps = (HEVCSPS*)h->sps_list[h->pps->sps_id]->data; |
| 190 | h->vps = (HEVCVPS*)h->vps_list[h->sps->vps_id]->data; |
| 191 | } |
| 192 | |
| 193 | if (!sh->first_slice_in_pic_flag) { |
| 194 | int slice_address_length; |
| 195 | |
| 196 | if (h->pps->dependent_slice_segments_enabled_flag) |
| 197 | sh->dependent_slice_segment_flag = get_bits1(gb); |
| 198 | else |
| 199 | sh->dependent_slice_segment_flag = 0; |
| 200 | |
| 201 | slice_address_length = av_ceil_log2_c(h->sps->ctb_width * |
| 202 | h->sps->ctb_height); |
| 203 | sh->slice_segment_addr = get_bits(gb, slice_address_length); |
| 204 | if (sh->slice_segment_addr >= h->sps->ctb_width * h->sps->ctb_height) { |
| 205 | av_log(h->avctx, AV_LOG_ERROR, "Invalid slice segment address: %u.\n", |
| 206 | sh->slice_segment_addr); |
| 207 | return AVERROR_INVALIDDATA; |
| 208 | } |
| 209 | } else |
| 210 | sh->dependent_slice_segment_flag = 0; |
| 211 | |
| 212 | if (sh->dependent_slice_segment_flag) |
| 213 | break; |
| 214 | |
| 215 | for (i = 0; i < h->pps->num_extra_slice_header_bits; i++) |
| 216 | skip_bits(gb, 1); // slice_reserved_undetermined_flag[] |
| 217 | |
| 218 | sh->slice_type = get_ue_golomb(gb); |
| 219 | if (!(sh->slice_type == I_SLICE || sh->slice_type == P_SLICE || |
| 220 | sh->slice_type == B_SLICE)) { |
| 221 | av_log(h->avctx, AV_LOG_ERROR, "Unknown slice type: %d.\n", |
| 222 | sh->slice_type); |
| 223 | return AVERROR_INVALIDDATA; |
| 224 | } |
| 225 | s->pict_type = sh->slice_type == B_SLICE ? AV_PICTURE_TYPE_B : |
| 226 | sh->slice_type == P_SLICE ? AV_PICTURE_TYPE_P : |
| 227 | AV_PICTURE_TYPE_I; |
| 228 | |
| 229 | if (h->pps->output_flag_present_flag) |
| 230 | sh->pic_output_flag = get_bits1(gb); |
| 231 | |
| 232 | if (h->sps->separate_colour_plane_flag) |
| 233 | sh->colour_plane_id = get_bits(gb, 2); |
| 234 | |
| 235 | if (!IS_IDR(h)) { |
| 236 | sh->pic_order_cnt_lsb = get_bits(gb, h->sps->log2_max_poc_lsb); |
| 237 | s->output_picture_number = h->poc = ff_hevc_compute_poc(h, sh->pic_order_cnt_lsb); |
| 238 | } else |
| 239 | s->output_picture_number = h->poc = 0; |
| 240 | |
| 241 | if (h->temporal_id == 0 && |
| 242 | h->nal_unit_type != NAL_TRAIL_N && |
| 243 | h->nal_unit_type != NAL_TSA_N && |
| 244 | h->nal_unit_type != NAL_STSA_N && |
| 245 | h->nal_unit_type != NAL_RADL_N && |
| 246 | h->nal_unit_type != NAL_RASL_N && |
| 247 | h->nal_unit_type != NAL_RADL_R && |
| 248 | h->nal_unit_type != NAL_RASL_R) |
| 249 | h->pocTid0 = h->poc; |
| 250 | |
| 251 | return 0; /* no need to evaluate the rest */ |
| 252 | } |
| 253 | buf += consumed; |
| 254 | } |
| 255 | /* didn't find a picture! */ |
| 256 | av_log(h->avctx, AV_LOG_ERROR, "missing picture in access unit\n"); |
| 257 | return -1; |
| 258 | } |
| 259 | |
| 260 | static int hevc_parse(AVCodecParserContext *s, |
| 261 | AVCodecContext *avctx, |
| 262 | const uint8_t **poutbuf, int *poutbuf_size, |
| 263 | const uint8_t *buf, int buf_size) |
| 264 | { |
| 265 | int next; |
| 266 | ParseContext *pc = &((HEVCParseContext *)s->priv_data)->pc; |
| 267 | |
| 268 | if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) { |
| 269 | next = buf_size; |
| 270 | } else { |
| 271 | next = hevc_find_frame_end(s, buf, buf_size); |
| 272 | if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) { |
| 273 | *poutbuf = NULL; |
| 274 | *poutbuf_size = 0; |
| 275 | return buf_size; |
| 276 | } |
| 277 | } |
| 278 | |
| 279 | parse_nal_units(s, avctx, buf, buf_size); |
| 280 | |
| 281 | *poutbuf = buf; |
| 282 | *poutbuf_size = buf_size; |
| 283 | return next; |
| 284 | } |
| 285 | |
| 286 | // Split after the parameter sets at the beginning of the stream if they exist. |
| 287 | static int hevc_split(AVCodecContext *avctx, const uint8_t *buf, int buf_size) |
| 288 | { |
| 289 | int i; |
| 290 | uint32_t state = -1; |
| 291 | int has_ps = 0; |
| 292 | |
| 293 | for (i = 0; i < buf_size; i++) { |
| 294 | state = (state << 8) | buf[i]; |
| 295 | if (((state >> 8) & 0xFFFFFF) == START_CODE) { |
| 296 | int nut = (state >> 1) & 0x3F; |
| 297 | if (nut >= NAL_VPS && nut <= NAL_PPS) |
| 298 | has_ps = 1; |
| 299 | else if (has_ps) |
| 300 | return i - 3; |
| 301 | else // no parameter set at the beginning of the stream |
| 302 | return 0; |
| 303 | } |
| 304 | } |
| 305 | return 0; |
| 306 | } |
| 307 | |
| 308 | static int hevc_init(AVCodecParserContext *s) |
| 309 | { |
| 310 | HEVCContext *h = &((HEVCParseContext *)s->priv_data)->h; |
| 311 | h->HEVClc = av_mallocz(sizeof(HEVCLocalContext)); |
| 312 | h->skipped_bytes_pos_size = INT_MAX; |
| 313 | |
| 314 | return 0; |
| 315 | } |
| 316 | |
| 317 | static void hevc_close(AVCodecParserContext *s) |
| 318 | { |
| 319 | int i; |
| 320 | HEVCContext *h = &((HEVCParseContext *)s->priv_data)->h; |
| 321 | ParseContext *pc = &((HEVCParseContext *)s->priv_data)->pc; |
| 322 | |
| 323 | av_freep(&h->skipped_bytes_pos); |
| 324 | av_freep(&h->HEVClc); |
| 325 | av_freep(&pc->buffer); |
| 326 | |
| 327 | for (i = 0; i < FF_ARRAY_ELEMS(h->vps_list); i++) |
| 328 | av_buffer_unref(&h->vps_list[i]); |
| 329 | for (i = 0; i < FF_ARRAY_ELEMS(h->sps_list); i++) |
| 330 | av_buffer_unref(&h->sps_list[i]); |
| 331 | for (i = 0; i < FF_ARRAY_ELEMS(h->pps_list); i++) |
| 332 | av_buffer_unref(&h->pps_list[i]); |
| 333 | |
| 334 | av_buffer_unref(&h->current_sps); |
| 335 | h->sps = NULL; |
| 336 | |
| 337 | for (i = 0; i < h->nals_allocated; i++) |
| 338 | av_freep(&h->nals[i].rbsp_buffer); |
| 339 | av_freep(&h->nals); |
| 340 | h->nals_allocated = 0; |
| 341 | } |
| 342 | |
| 343 | AVCodecParser ff_hevc_parser = { |
| 344 | .codec_ids = { AV_CODEC_ID_HEVC }, |
| 345 | .priv_data_size = sizeof(HEVCParseContext), |
| 346 | .parser_init = hevc_init, |
| 347 | .parser_parse = hevc_parse, |
| 348 | .parser_close = hevc_close, |
| 349 | .split = hevc_split, |
| 350 | }; |