h264_parser.c 20.6 KB
Newer Older
1 2 3 4
/*
 * H.26L/H.264/AVC/JVT/14496-10/... parser
 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
 *
5
 * This file is part of Libav.
6
 *
7
 * Libav is free software; you can redistribute it and/or
8 9 10 11
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * 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 14 15 16 17
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * 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 20 21 22
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
23
 * @file
24
 * H.264 / AVC / MPEG-4 part10 parser.
25 26 27
 * @author Michael Niedermayer <michaelni@gmx.at>
 */

28 29 30 31 32 33 34 35 36
#include <assert.h>
#include <stdint.h>

#include "libavutil/avutil.h"
#include "libavutil/error.h"
#include "libavutil/log.h"
#include "libavutil/mem.h"
#include "libavutil/pixfmt.h"

37
#include "avcodec.h"
38
#include "get_bits.h"
39
#include "golomb.h"
40
#include "h264.h"
41
#include "h264_sei.h"
42
#include "h264_ps.h"
43
#include "h264data.h"
44
#include "internal.h"
45
#include "mpegutils.h"
46
#include "parser.h"
47

48 49
typedef struct H264ParseContext {
    ParseContext pc;
50
    H264ParamSets ps;
51
    H264DSPContext h264dsp;
52
    H264POCContext poc;
53
    H264SEIContext sei;
54 55
    int is_avc;
    int nal_length_size;
56
    int got_first;
57
    int picture_structure;
58
} H264ParseContext;
59

60 61

static int h264_find_frame_end(H264ParseContext *p, const uint8_t *buf,
62
                               int buf_size)
63 64 65
{
    int i;
    uint32_t state;
66
    ParseContext *pc = &p->pc;
67
//    mb_addr= pc->mb_addr - 1;
68 69 70
    state = pc->state;
    if (state > 13)
        state = 7;
71

72 73
    for (i = 0; i < buf_size; i++) {
        if (state == 7) {
74
            i += p->h264dsp.startcode_find_candidate(buf + i, buf_size - i);
75 76
            if (i < buf_size)
                state = 2;
77 78 79 80 81 82 83 84
        } else if (state <= 2) {
            if (buf[i] == 1)
                state ^= 5;            // 2->7, 1->4, 0->5
            else if (buf[i])
                state = 7;
            else
                state >>= 1;           // 2->1, 1->0, 0->0
        } else if (state <= 5) {
85
            int nalu_type = buf[i] & 0x1F;
86 87
            if (nalu_type == H264_NAL_SEI || nalu_type == H264_NAL_SPS ||
                nalu_type == H264_NAL_PPS || nalu_type == H264_NAL_AUD) {
88
                if (pc->frame_start_found) {
89
                    i++;
90
                    goto found;
91
                }
92 93
            } else if (nalu_type == H264_NAL_SLICE || nalu_type == H264_NAL_DPA ||
                       nalu_type == H264_NAL_IDR_SLICE) {
94 95
                if (pc->frame_start_found) {
                    state += 8;
96
                    continue;
97
                } else
98 99
                    pc->frame_start_found = 1;
            }
100 101
            state = 7;
        } else {
102
            // first_mb_in_slice is 0, probably the first nal of a new slice
103
            if (buf[i] & 0x80)
104
                goto found;
105
            state = 7;
106 107
        }
    }
108
    pc->state = state;
109
    return END_NOT_FOUND;
110 111

found:
112 113 114
    pc->state             = 7;
    pc->frame_start_found = 0;
    return i - (state & 5);
115 116
}

117 118
static int scan_mmco_reset(AVCodecParserContext *s, GetBitContext *gb,
                           AVCodecContext *avctx)
119
{
120 121
    H264PredWeightTable pwt;
    int slice_type_nos = s->pict_type & 3;
122
    H264ParseContext *p = s->priv_data;
123
    int list_count, ref_count[2];
124 125


126
    if (p->ps.pps->redundant_pic_cnt_present)
127
        get_ue_golomb(gb); // redundant_pic_count
128

129 130
    if (slice_type_nos == AV_PICTURE_TYPE_B)
        get_bits1(gb); // direct_spatial_mv_pred
131

132
    if (ff_h264_parse_ref_count(&list_count, ref_count, gb, p->ps.pps,
133
                                slice_type_nos, p->picture_structure) < 0)
134 135
        return AVERROR_INVALIDDATA;

136
    if (slice_type_nos != AV_PICTURE_TYPE_I) {
137
        int list;
138
        for (list = 0; list < list_count; list++) {
139
            if (get_bits1(gb)) {
140 141
                int index;
                for (index = 0; ; index++) {
142
                    unsigned int reordering_of_pic_nums_idc = get_ue_golomb_31(gb);
143 144

                    if (reordering_of_pic_nums_idc < 3)
145
                        get_ue_golomb(gb);
146
                    else if (reordering_of_pic_nums_idc > 3) {
147
                        av_log(avctx, AV_LOG_ERROR,
148 149 150 151 152 153
                               "illegal reordering_of_pic_nums_idc %d\n",
                               reordering_of_pic_nums_idc);
                        return AVERROR_INVALIDDATA;
                    } else
                        break;

154
                    if (index >= ref_count[list]) {
155
                        av_log(avctx, AV_LOG_ERROR,
156
                               "reference count %d overflow\n", index);
157 158 159 160 161 162 163
                        return AVERROR_INVALIDDATA;
                    }
                }
            }
        }
    }

164 165 166
    if ((p->ps.pps->weighted_pred && slice_type_nos == AV_PICTURE_TYPE_P) ||
        (p->ps.pps->weighted_bipred_idc == 1 && slice_type_nos == AV_PICTURE_TYPE_B))
        ff_h264_pred_weight_table(gb, p->ps.sps, ref_count, slice_type_nos,
167
                                  &pwt);
168

169
    if (get_bits1(gb)) { // adaptive_ref_pic_marking_mode_flag
170 171
        int i;
        for (i = 0; i < MAX_MMCO_COUNT; i++) {
172
            MMCOOpcode opcode = get_ue_golomb_31(gb);
173
            if (opcode > (unsigned) MMCO_LONG) {
174
                av_log(avctx, AV_LOG_ERROR,
175 176 177 178 179 180 181 182 183 184
                       "illegal memory management control operation %d\n",
                       opcode);
                return AVERROR_INVALIDDATA;
            }
            if (opcode == MMCO_END)
               return 0;
            else if (opcode == MMCO_RESET)
                return 1;

            if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG)
185
                get_ue_golomb(gb);
186 187
            if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED ||
                opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG)
188
                get_ue_golomb_31(gb);
189 190 191 192 193 194
        }
    }

    return 0;
}

195
/**
196 197 198 199 200 201 202 203 204 205 206
 * Parse NAL units of found picture and decode some basic information.
 *
 * @param s parser context.
 * @param avctx codec context.
 * @param buf buffer with field/frame data.
 * @param buf_size size of the buffer.
 */
static inline int parse_nal_units(AVCodecParserContext *s,
                                  AVCodecContext *avctx,
                                  const uint8_t *buf, int buf_size)
{
207
    H264ParseContext *p = s->priv_data;
208
    const uint8_t *buf_end = buf + buf_size;
209 210 211

    H2645NAL nal = { NULL };

212
    unsigned int pps_id;
213
    unsigned int slice_type;
214
    int state = -1, got_reset = 0;
215
    int field_poc[2];
216
    int ret;
217 218

    /* set some sane default values */
219 220
    s->pict_type         = AV_PICTURE_TYPE_I;
    s->key_frame         = 0;
221
    s->picture_structure = AV_PICTURE_STRUCTURE_UNKNOWN;
222

223
    ff_h264_sei_uninit(&p->sei);
224

225 226 227
    if (!buf_size)
        return 0;

228
    for (;;) {
229
        const SPS *sps;
230
        int src_length, consumed;
231
        buf = avpriv_find_start_code(buf, buf_end, &state);
232
        if (buf >= buf_end)
233 234 235 236
            break;
        --buf;
        src_length = buf_end - buf;
        switch (state & 0x1f) {
237 238
        case H264_NAL_SLICE:
        case H264_NAL_IDR_SLICE:
239
            // Do not walk the whole buffer just to decode slice header
240
            if ((state & 0x1f) == H264_NAL_IDR_SLICE || ((state >> 5) & 0x3) == 0) {
241 242 243 244 245 246 247 248 249
                /* IDR or disposable slice
                 * No need to decode many bytes because MMCOs shall not be present. */
                if (src_length > 60)
                    src_length = 60;
            } else {
                /* To decode up to MMCOs */
                if (src_length > 1000)
                    src_length = 1000;
            }
250 251
            break;
        }
252 253 254

        consumed = ff_h2645_extract_rbsp(buf, src_length, &nal);
        if (consumed < 0)
255 256
            break;

257 258 259 260 261 262 263
        ret = init_get_bits(&nal.gb, nal.data, nal.size * 8);
        if (ret < 0)
            goto fail;
        get_bits1(&nal.gb);
        nal.ref_idc = get_bits(&nal.gb, 2);
        nal.type    = get_bits(&nal.gb, 5);

264
        switch (nal.type) {
265
        case H264_NAL_SPS:
266
            ff_h264_decode_seq_parameter_set(&nal.gb, avctx, &p->ps);
267
            break;
268
        case H264_NAL_PPS:
269 270
            ff_h264_decode_picture_parameter_set(&nal.gb, avctx, &p->ps,
                                                 nal.size_bits);
271
            break;
272
        case H264_NAL_SEI:
273
            ff_h264_sei_decode(&p->sei, &nal.gb, &p->ps, avctx);
274
            break;
275
        case H264_NAL_IDR_SLICE:
276
            s->key_frame = 1;
277

278 279 280 281
            p->poc.prev_frame_num        = 0;
            p->poc.prev_frame_num_offset = 0;
            p->poc.prev_poc_msb          =
            p->poc.prev_poc_lsb          = 0;
282
        /* fall through */
283
        case H264_NAL_SLICE:
284 285
            get_ue_golomb(&nal.gb);  // skip first_mb_in_slice
            slice_type   = get_ue_golomb_31(&nal.gb);
286
            s->pict_type = ff_h264_golomb_to_pict_type[slice_type % 5];
287
            if (p->sei.recovery_point.recovery_frame_cnt >= 0) {
288 289 290
                /* key frame, since recovery_frame_cnt is set */
                s->key_frame = 1;
            }
291
            pps_id = get_ue_golomb(&nal.gb);
292
            if (pps_id >= MAX_PPS_COUNT) {
293
                av_log(avctx, AV_LOG_ERROR,
294
                       "pps_id %u out of range\n", pps_id);
295
                goto fail;
296
            }
297
            if (!p->ps.pps_list[pps_id]) {
298
                av_log(avctx, AV_LOG_ERROR,
299
                       "non-existing PPS %u referenced\n", pps_id);
300
                goto fail;
301
            }
302 303
            p->ps.pps = (const PPS*)p->ps.pps_list[pps_id]->data;
            if (!p->ps.sps_list[p->ps.pps->sps_id]) {
304
                av_log(avctx, AV_LOG_ERROR,
305
                       "non-existing SPS %u referenced\n", p->ps.pps->sps_id);
306
                goto fail;
307
            }
308
            p->ps.sps = (SPS*)p->ps.sps_list[p->ps.pps->sps_id]->data;
309

310 311
            sps = p->ps.sps;

312
            p->poc.frame_num = get_bits(&nal.gb, sps->log2_max_frame_num);
313 314 315 316 317

            s->coded_width  = 16 * sps->mb_width;
            s->coded_height = 16 * sps->mb_height;
            s->width        = s->coded_width  - (sps->crop_right + sps->crop_left);
            s->height       = s->coded_height - (sps->crop_top   + sps->crop_bottom);
318 319 320 321 322
            if (s->width <= 0 || s->height <= 0) {
                s->width  = s->coded_width;
                s->height = s->coded_height;
            }

323
            switch (sps->bit_depth_luma) {
324
            case 9:
325 326 327
                if (sps->chroma_format_idc == 3)      s->format = AV_PIX_FMT_YUV444P9;
                else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P9;
                else                                  s->format = AV_PIX_FMT_YUV420P9;
328 329
                break;
            case 10:
330 331 332
                if (sps->chroma_format_idc == 3)      s->format = AV_PIX_FMT_YUV444P10;
                else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P10;
                else                                  s->format = AV_PIX_FMT_YUV420P10;
333 334
                break;
            case 8:
335 336 337
                if (sps->chroma_format_idc == 3)      s->format = AV_PIX_FMT_YUV444P;
                else if (sps->chroma_format_idc == 2) s->format = AV_PIX_FMT_YUV422P;
                else                                  s->format = AV_PIX_FMT_YUV420P;
338 339 340 341 342
                break;
            default:
                s->format = AV_PIX_FMT_NONE;
            }

343 344
            avctx->profile = ff_h264_get_profile(sps);
            avctx->level   = sps->level_idc;
345

346
            if (sps->frame_mbs_only_flag) {
347
                p->picture_structure = PICT_FRAME;
348
            } else {
349
                if (get_bits1(&nal.gb)) { // field_pic_flag
350
                    p->picture_structure = PICT_TOP_FIELD + get_bits1(&nal.gb); // bottom_field_flag
351
                } else {
352
                    p->picture_structure = PICT_FRAME;
353 354 355
                }
            }

356
            if (nal.type == H264_NAL_IDR_SLICE)
357
                get_ue_golomb(&nal.gb); /* idr_pic_id */
358
            if (sps->poc_type == 0) {
359
                p->poc.poc_lsb = get_bits(&nal.gb, sps->log2_max_poc_lsb);
360

361
                if (p->ps.pps->pic_order_present == 1 &&
362
                    p->picture_structure == PICT_FRAME)
363
                    p->poc.delta_poc_bottom = get_se_golomb(&nal.gb);
364 365
            }

366 367
            if (sps->poc_type == 1 &&
                !sps->delta_pic_order_always_zero_flag) {
368
                p->poc.delta_poc[0] = get_se_golomb(&nal.gb);
369

370
                if (p->ps.pps->pic_order_present == 1 &&
371
                    p->picture_structure == PICT_FRAME)
372
                    p->poc.delta_poc[1] = get_se_golomb(&nal.gb);
373 374
            }

375 376
            /* Decode POC of this picture.
             * The prev_ values needed for decoding POC of the next picture are not set here. */
377
            field_poc[0] = field_poc[1] = INT_MAX;
378
            ff_h264_init_poc(field_poc, &s->output_picture_number, sps,
379
                             &p->poc, p->picture_structure, nal.ref_idc);
380

381 382 383 384
            /* Continue parsing to check if MMCO_RESET is present.
             * FIXME: MMCO_RESET could appear in non-first slice.
             *        Maybe, we should parse all undisposable non-IDR slice of this
             *        picture until encountering MMCO_RESET in a slice of it. */
385
            if (nal.ref_idc && nal.type != H264_NAL_IDR_SLICE) {
386
                got_reset = scan_mmco_reset(s, &nal.gb, avctx);
387
                if (got_reset < 0)
388
                    goto fail;
389 390
            }

391
            /* Set up the prev_ values for decoding POC of the next picture. */
392 393
            p->poc.prev_frame_num        = got_reset ? 0 : p->poc.frame_num;
            p->poc.prev_frame_num_offset = got_reset ? 0 : p->poc.frame_num_offset;
394
            if (nal.ref_idc != 0) {
395
                if (!got_reset) {
396 397
                    p->poc.prev_poc_msb = p->poc.poc_msb;
                    p->poc.prev_poc_lsb = p->poc.poc_lsb;
398
                } else {
399 400
                    p->poc.prev_poc_msb = 0;
                    p->poc.prev_poc_lsb =
401
                        p->picture_structure == PICT_BOTTOM_FIELD ? 0 : field_poc[0];
402
                }
403 404
            }

405
            if (sps->pic_struct_present_flag) {
406
                switch (p->sei.picture_timing.pic_struct) {
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
                case SEI_PIC_STRUCT_TOP_FIELD:
                case SEI_PIC_STRUCT_BOTTOM_FIELD:
                    s->repeat_pict = 0;
                    break;
                case SEI_PIC_STRUCT_FRAME:
                case SEI_PIC_STRUCT_TOP_BOTTOM:
                case SEI_PIC_STRUCT_BOTTOM_TOP:
                    s->repeat_pict = 1;
                    break;
                case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
                case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
                    s->repeat_pict = 2;
                    break;
                case SEI_PIC_STRUCT_FRAME_DOUBLING:
                    s->repeat_pict = 3;
                    break;
                case SEI_PIC_STRUCT_FRAME_TRIPLING:
                    s->repeat_pict = 5;
                    break;
                default:
427
                    s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
428
                    break;
429 430
                }
            } else {
431
                s->repeat_pict = p->picture_structure == PICT_FRAME ? 1 : 0;
432 433
            }

434
            if (p->picture_structure == PICT_FRAME) {
435
                s->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
436
                if (sps->pic_struct_present_flag) {
437
                    switch (p->sei.picture_timing.pic_struct) {
438 439 440 441 442 443 444 445 446 447 448
                    case SEI_PIC_STRUCT_TOP_BOTTOM:
                    case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
                        s->field_order = AV_FIELD_TT;
                        break;
                    case SEI_PIC_STRUCT_BOTTOM_TOP:
                    case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
                        s->field_order = AV_FIELD_BB;
                        break;
                    default:
                        s->field_order = AV_FIELD_PROGRESSIVE;
                        break;
449 450 451 452 453 454 455 456 457 458
                    }
                } else {
                    if (field_poc[0] < field_poc[1])
                        s->field_order = AV_FIELD_TT;
                    else if (field_poc[0] > field_poc[1])
                        s->field_order = AV_FIELD_BB;
                    else
                        s->field_order = AV_FIELD_PROGRESSIVE;
                }
            } else {
459
                if (p->picture_structure == PICT_TOP_FIELD)
460 461 462 463 464 465
                    s->picture_structure = AV_PICTURE_STRUCTURE_TOP_FIELD;
                else
                    s->picture_structure = AV_PICTURE_STRUCTURE_BOTTOM_FIELD;
                s->field_order = AV_FIELD_UNKNOWN;
            }

466
            av_freep(&nal.rbsp_buffer);
467 468 469 470 471
            return 0; /* no need to evaluate the rest */
        }
        buf += consumed;
    }
    /* didn't find a picture! */
472
    av_log(avctx, AV_LOG_ERROR, "missing picture in access unit\n");
473 474
fail:
    av_freep(&nal.rbsp_buffer);
475 476 477
    return -1;
}

478 479 480 481 482
static int h264_parse(AVCodecParserContext *s,
                      AVCodecContext *avctx,
                      const uint8_t **poutbuf, int *poutbuf_size,
                      const uint8_t *buf, int buf_size)
{
483 484
    H264ParseContext *p = s->priv_data;
    ParseContext *pc = &p->pc;
485 486
    int next;

487 488
    if (!p->got_first) {
        p->got_first = 1;
489
        if (avctx->extradata_size) {
490 491 492
            ff_h264_decode_extradata(avctx->extradata, avctx->extradata_size,
                                     &p->ps, &p->is_avc, &p->nal_length_size,
                                     avctx->err_recognition, avctx);
493 494 495
        }
    }

496 497 498
    if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
        next = buf_size;
    } else {
499
        next = h264_find_frame_end(p, buf, buf_size);
500 501

        if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
502
            *poutbuf      = NULL;
503 504 505 506
            *poutbuf_size = 0;
            return buf_size;
        }

507 508
        if (next < 0 && next != END_NOT_FOUND) {
            assert(pc->last_index + next >= 0);
509
            h264_find_frame_end(p, &pc->buffer[pc->last_index + next], -next); // update state
510
        }
511
    }
512

513
    parse_nal_units(s, avctx, buf, buf_size);
514

515 516 517 518
    if (p->sei.picture_timing.cpb_removal_delay >= 0) {
        s->dts_sync_point    = p->sei.buffering_period.present;
        s->dts_ref_dts_delta = p->sei.picture_timing.cpb_removal_delay;
        s->pts_dts_delta     = p->sei.picture_timing.dpb_output_delay;
519 520 521 522 523 524 525 526 527
    } else {
        s->dts_sync_point    = INT_MIN;
        s->dts_ref_dts_delta = INT_MIN;
        s->pts_dts_delta     = INT_MIN;
    }

    if (s->flags & PARSER_FLAG_ONCE) {
        s->flags &= PARSER_FLAG_COMPLETE_FRAMES;
    }
528

529
    *poutbuf      = buf;
530 531 532 533 534 535 536 537 538
    *poutbuf_size = buf_size;
    return next;
}

static int h264_split(AVCodecContext *avctx,
                      const uint8_t *buf, int buf_size)
{
    int i;
    uint32_t state = -1;
539 540 541 542 543 544 545 546 547 548
    int has_sps    = 0;

    for (i = 0; i <= buf_size; i++) {
        if ((state & 0xFFFFFF1F) == 0x107)
            has_sps = 1;
        /*  if((state&0xFFFFFF1F) == 0x101 ||
         *     (state&0xFFFFFF1F) == 0x102 ||
         *     (state&0xFFFFFF1F) == 0x105) {
         *  }
         */
549 550 551 552
        if ((state & 0xFFFFFF00) == 0x100 && (state & 0xFFFFFF1F) != 0x106 &&
            (state & 0xFFFFFF1F) != 0x107 && (state & 0xFFFFFF1F) != 0x108 &&
            (state & 0xFFFFFF1F) != 0x109 && (state & 0xFFFFFF1F) != 0x10d &&
            (state & 0xFFFFFF1F) != 0x10f) {
553 554 555 556
            if (has_sps) {
                while (i > 4 && buf[i - 5] == 0)
                    i--;
                return i - 4;
557 558
            }
        }
559 560
        if (i < buf_size)
            state = (state << 8) | buf[i];
561 562 563 564
    }
    return 0;
}

565
static void h264_close(AVCodecParserContext *s)
566
{
567 568
    H264ParseContext *p = s->priv_data;
    ParseContext *pc = &p->pc;
569
    int i;
570 571

    av_free(pc->buffer);
572

573 574
    ff_h264_sei_uninit(&p->sei);

575 576 577 578 579
    for (i = 0; i < FF_ARRAY_ELEMS(p->ps.sps_list); i++)
        av_buffer_unref(&p->ps.sps_list[i]);

    for (i = 0; i < FF_ARRAY_ELEMS(p->ps.pps_list); i++)
        av_buffer_unref(&p->ps.pps_list[i]);
580 581
}

582
static av_cold int init(AVCodecParserContext *s)
583
{
584
    H264ParseContext *p = s->priv_data;
585

586
    ff_h264dsp_init(&p->h264dsp, 8, 1);
587 588
    return 0;
}
589

590
AVCodecParser ff_h264_parser = {
591
    .codec_ids      = { AV_CODEC_ID_H264 },
592
    .priv_data_size = sizeof(H264ParseContext),
593 594
    .parser_init    = init,
    .parser_parse   = h264_parse,
595
    .parser_close   = h264_close,
596
    .split          = h264_split,
597
};