git.sesse.net Git - vlc/blob - modules/packetizer/h264.c

   1 /*****************************************************************************
   2  * h264.c: h264/avc video packetizer
   3  *****************************************************************************
   4  * Copyright (C) 2001, 2002 VideoLAN
   5  * $Id$
   6  *
   7  * Authors: Laurent Aimar <fenrir@via.ecp.fr>
   8  *          Eric Petit <titer@videolan.org>
   9  *          Gildas Bazin <gbazin@videolan.org>
  10  *
  11  * This program is free software; you can redistribute it and/or modify
  12  * it under the terms of the GNU General Public License as published by
  13  * the Free Software Foundation; either version 2 of the License, or
  14  * (at your option) any later version.
  15  *
  16  * This program is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  19  * GNU General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU General Public License
  22  * along with this program; if not, write to the Free Software
  23  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
  24  *****************************************************************************/
  25
  26 /*****************************************************************************
  27  * Preamble
  28  *****************************************************************************/
  29 #include <stdlib.h>                                      /* malloc(), free() */
  30
  31 #include <vlc/vlc.h>
  32 #include <vlc/decoder.h>
  33 #include <vlc/sout.h>
  34
  35 #include "vlc_block_helper.h"
  36 #include "vlc_bits.h"
  37
  38 /*****************************************************************************
  39  * Module descriptor
  40  *****************************************************************************/
  41 static int  Open ( vlc_object_t * );
  42 static void Close( vlc_object_t * );
  43
  44 vlc_module_begin();
  45     set_description( _("H264 video packetizer") );
  46     set_capability( "packetizer", 50 );
  47     set_callbacks( Open, Close );
  48 vlc_module_end();
  49
  50
  51 /****************************************************************************
  52  * Local prototypes
  53  ****************************************************************************/
  54 static block_t *Packetize( decoder_t *, block_t ** );
  55 static block_t *PacketizeAVC1( decoder_t *, block_t ** );
  56
  57 struct decoder_sys_t
  58 {
  59     block_bytestream_t bytestream;
  60
  61     int     i_state;
  62     int     i_offset;
  63     uint8_t startcode[4];
  64
  65     vlc_bool_t b_slice;
  66     block_t    *p_frame;
  67
  68     int64_t      i_dts;
  69     int64_t      i_pts;
  70     unsigned int i_flags;
  71
  72     vlc_bool_t   b_sps;
  73
  74     /* avcC data */
  75     int i_avcC_length_size;
  76
  77     /* Useful values of the Sequence Parameter Set */
  78     int i_log2_max_frame_num;
  79     int b_frame_mbs_only;
  80
  81     /* Useful values of the Slice Header */
  82     int i_nal_type;
  83     int i_nal_ref_idc;
  84     int i_idr_pic_id;
  85     int i_frame_num;
  86 };
  87
  88 enum
  89 {
  90     STATE_NOSYNC,
  91     STATE_NEXT_SYNC,
  92 };
  93
  94 enum nal_unit_type_e
  95 {
  96     NAL_UNKNOWN = 0,
  97     NAL_SLICE   = 1,
  98     NAL_SLICE_DPA   = 2,
  99     NAL_SLICE_DPB   = 3,
 100     NAL_SLICE_DPC   = 4,
 101     NAL_SLICE_IDR   = 5,    /* ref_idc != 0 */
 102     NAL_SEI         = 6,    /* ref_idc == 0 */
 103     NAL_SPS         = 7,
 104     NAL_PPS         = 8
 105     /* ref_idc == 0 for 6,9,10,11,12 */
 106 };
 107
 108 enum nal_priority_e
 109 {
 110     NAL_PRIORITY_DISPOSABLE = 0,
 111     NAL_PRIORITY_LOW        = 1,
 112     NAL_PRIORITY_HIGH       = 2,
 113     NAL_PRIORITY_HIGHEST    = 3,
 114 };
 115
 116 static block_t *ParseNALBlock( decoder_t *, block_t * );
 117
 118 static block_t *nal_get_annexeb( decoder_t *, uint8_t *p, int );
 119
 120 /*****************************************************************************
 121  * Open: probe the packetizer and return score
 122  *****************************************************************************/
 123 static int Open( vlc_object_t *p_this )
 124 {
 125     decoder_t     *p_dec = (decoder_t*)p_this;
 126     decoder_sys_t *p_sys;
 127
 128     if( p_dec->fmt_in.i_codec != VLC_FOURCC( 'h', '2', '6', '4') &&
 129         p_dec->fmt_in.i_codec != VLC_FOURCC( 'H', '2', '6', '4') &&
 130         p_dec->fmt_in.i_codec != VLC_FOURCC( 'V', 'S', 'S', 'H') &&
 131         p_dec->fmt_in.i_codec != VLC_FOURCC( 'v', 's', 's', 'h') &&
 132         ( p_dec->fmt_in.i_codec != VLC_FOURCC( 'a', 'v', 'c', '1') ||
 133           p_dec->fmt_in.i_extra < 7 ) )
 134     {
 135         return VLC_EGENERIC;
 136     }
 137
 138     /* Allocate the memory needed to store the decoder's structure */
 139     if( ( p_dec->p_sys = p_sys = malloc( sizeof(decoder_sys_t) ) ) == NULL )
 140     {
 141         msg_Err( p_dec, "out of memory" );
 142         return VLC_EGENERIC;
 143     }
 144     p_sys->i_state = STATE_NOSYNC;
 145     p_sys->i_offset = 0;
 146     p_sys->startcode[0] = 0;
 147     p_sys->startcode[1] = 0;
 148     p_sys->startcode[2] = 0;
 149     p_sys->startcode[3] = 1;
 150     p_sys->bytestream = block_BytestreamInit( p_dec );
 151     p_sys->b_slice = VLC_FALSE;
 152     p_sys->p_frame = NULL;
 153     p_sys->i_dts   = 0;
 154     p_sys->i_pts   = 0;
 155     p_sys->i_flags = 0;
 156     p_sys->b_sps   = VLC_FALSE;
 157
 158     p_sys->i_nal_type = -1;
 159     p_sys->i_nal_ref_idc = -1;
 160     p_sys->i_idr_pic_id = -1;
 161     p_sys->i_frame_num = -1;
 162
 163     /* Setup properties */
 164     es_format_Copy( &p_dec->fmt_out, &p_dec->fmt_in );
 165     p_dec->fmt_out.i_codec = VLC_FOURCC( 'h', '2', '6', '4' );
 166     /* FIXME: FFMPEG isn't happy at all if you leave this */
 167     if( p_dec->fmt_out.i_extra ) free( p_dec->fmt_out.p_extra );
 168     p_dec->fmt_out.i_extra = 0; p_dec->fmt_out.p_extra = 0;
 169
 170     if( p_dec->fmt_in.i_codec == VLC_FOURCC( 'a', 'v', 'c', '1' ) )
 171     {
 172         uint8_t *p = &((uint8_t*)p_dec->fmt_in.p_extra)[4];
 173         int i_sps, i_pps;
 174         int i;
 175
 176         /* Parse avcC */
 177         p_sys->i_avcC_length_size = 1 + ((*p++)&0x03);
 178
 179         /* Read SPS */
 180         i_sps = (*p++)&0x1f;
 181
 182         for( i = 0; i < i_sps; i++ )
 183         {
 184             int i_length = GetWBE( p );
 185             block_t *p_sps = nal_get_annexeb( p_dec, p + 2, i_length );
 186
 187             ParseNALBlock( p_dec, p_sps );
 188             p += 2 + i_length;
 189         }
 190         /* Read PPS */
 191         i_pps = *p++;
 192         for( i = 0; i < i_pps; i++ )
 193         {
 194             int i_length = GetWBE( p );
 195             block_t *p_pps = nal_get_annexeb( p_dec, p + 2, i_length );
 196
 197             ParseNALBlock( p_dec, p_pps );
 198             p += 2 + i_length;
 199         }
 200         msg_Dbg( p_dec, "avcC length size=%d sps=%d pps=%d",
 201                  p_sys->i_avcC_length_size, i_sps, i_pps );
 202
 203         /* Set callback */
 204         p_dec->pf_packetize = PacketizeAVC1;
 205     }
 206     else
 207     {
 208         /* Set callback */
 209         p_dec->pf_packetize = Packetize;
 210     }
 211
 212     return VLC_SUCCESS;
 213 }
 214
 215 /*****************************************************************************
 216  * Close: clean up the packetizer
 217  *****************************************************************************/
 218 static void Close( vlc_object_t *p_this )
 219 {
 220     decoder_t *p_dec = (decoder_t*)p_this;
 221     decoder_sys_t *p_sys = p_dec->p_sys;
 222
 223     block_BytestreamRelease( &p_sys->bytestream );
 224     free( p_sys );
 225 }
 226
 227 /****************************************************************************
 228  * Packetize: the whole thing
 229  ****************************************************************************/
 230 static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
 231 {
 232     decoder_sys_t *p_sys = p_dec->p_sys;
 233     block_t       *p_pic;
 234
 235     if( !pp_block || !*pp_block ) return NULL;
 236
 237     block_BytestreamPush( &p_sys->bytestream, *pp_block );
 238
 239     for( ;; )
 240     {
 241         switch( p_sys->i_state )
 242         {
 243             case STATE_NOSYNC:
 244                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
 245                       &p_sys->i_offset, p_sys->startcode+1, 3 ) == VLC_SUCCESS)
 246                 {
 247                     p_sys->i_state = STATE_NEXT_SYNC;
 248                 }
 249
 250                 if( p_sys->i_offset )
 251                 {
 252                     block_SkipBytes( &p_sys->bytestream, p_sys->i_offset );
 253                     p_sys->i_offset = 0;
 254                     block_BytestreamFlush( &p_sys->bytestream );
 255                 }
 256
 257                 if( p_sys->i_state != STATE_NEXT_SYNC )
 258                 {
 259                     /* Need more data */
 260                     return NULL;
 261                 }
 262
 263                 p_sys->i_offset = 1; /* To find next startcode */
 264
 265             case STATE_NEXT_SYNC:
 266                 /* Find the next startcode */
 267                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
 268                       &p_sys->i_offset, p_sys->startcode, 3 ) != VLC_SUCCESS)
 269                 {
 270                     if( block_FindStartcodeFromOffset( &p_sys->bytestream,
 271                           &p_sys->i_offset, p_sys->startcode+1, 3 ) !=
 272                         VLC_SUCCESS )
 273                     {
 274                         /* Need more data */
 275                         return NULL;
 276                     }
 277                 }
 278
 279                 /* Get the new fragment and set the pts/dts */
 280                 p_pic = block_New( p_dec, p_sys->i_offset );
 281                 p_pic->i_pts = p_sys->bytestream.p_block->i_pts;
 282                 p_pic->i_dts = p_sys->bytestream.p_block->i_dts;
 283
 284                 block_GetBytes( &p_sys->bytestream, p_pic->p_buffer,
 285                                 p_pic->i_buffer );
 286
 287                 p_sys->i_offset = 0;
 288
 289                 /* Parse the NAL */
 290                 if( !( p_pic = ParseNALBlock( p_dec, p_pic ) ) )
 291                 {
 292                     p_sys->i_state = STATE_NOSYNC;
 293                     break;
 294                 }
 295
 296                 /* So p_block doesn't get re-added several times */
 297                 *pp_block = block_BytestreamPop( &p_sys->bytestream );
 298
 299                 p_sys->i_state = STATE_NOSYNC;
 300
 301                 return p_pic;
 302         }
 303     }
 304 }
 305
 306 /****************************************************************************
 307  * PacketizeAVC1: the whole thing
 308  ****************************************************************************/
 309 static block_t *PacketizeAVC1( decoder_t *p_dec, block_t **pp_block )
 310 {
 311     decoder_sys_t *p_sys = p_dec->p_sys;
 312     block_t       *p_block;
 313     block_t       *p_ret = NULL;
 314     uint8_t       *p;
 315
 316     if( !pp_block || !*pp_block ) return NULL;
 317
 318     p_block = *pp_block;
 319     *pp_block = NULL;
 320
 321     for( p = p_block->p_buffer; p < &p_block->p_buffer[p_block->i_buffer]; )
 322     {
 323         block_t *p_pic;
 324         int i_size = 0;
 325         int i;
 326
 327         for( i = 0; i < p_sys->i_avcC_length_size; i++ )
 328         {
 329             i_size = (i_size << 8) | (*p++);
 330         }
 331
 332         if( i_size > 0 )
 333         {
 334             block_t *p_part = nal_get_annexeb( p_dec, p, i_size );
 335
 336             p_part->i_dts = p_block->i_dts;
 337             p_part->i_pts = p_block->i_pts;
 338
 339             /* Parse the NAL */
 340             if( ( p_pic = ParseNALBlock( p_dec, p_part ) ) )
 341             {
 342                 block_ChainAppend( &p_ret, p_pic );
 343             }
 344         }
 345         p += i_size;
 346     }
 347
 348     return p_ret;
 349 }
 350
 351 static block_t *nal_get_annexeb( decoder_t *p_dec, uint8_t *p, int i_size )
 352 {
 353     block_t *p_nal;
 354
 355     p_nal = block_New( p_dec, 3 + i_size );
 356
 357     /* Add start code */
 358     p_nal->p_buffer[0] = 0x00;
 359     p_nal->p_buffer[1] = 0x00;
 360     p_nal->p_buffer[2] = 0x01;
 361
 362     /* Copy nalu */
 363     memcpy( &p_nal->p_buffer[3], p, i_size );
 364
 365     return p_nal;
 366 }
 367
 368 static void nal_get_decoded( uint8_t **pp_ret, int *pi_ret,
 369                              uint8_t *src, int i_src )
 370 {
 371     uint8_t *end = &src[i_src];
 372     uint8_t *dst = malloc( i_src );
 373
 374     *pp_ret = dst;
 375
 376     while( src < end )
 377     {
 378         if( src < end - 3 && src[0] == 0x00 && src[1] == 0x00 &&
 379             src[2] == 0x03 )
 380         {
 381             *dst++ = 0x00;
 382             *dst++ = 0x00;
 383
 384             src += 3;
 385             continue;
 386         }
 387         *dst++ = *src++;
 388     }
 389
 390     *pi_ret = dst - *pp_ret;
 391 }
 392
 393 static inline int bs_read_ue( bs_t *s )
 394 {
 395     int i = 0;
 396
 397     while( bs_read1( s ) == 0 && s->p < s->p_end && i < 32 )
 398     {
 399         i++;
 400     }
 401     return( ( 1 << i) - 1 + bs_read( s, i ) );
 402 }
 403
 404 static inline int bs_read_se( bs_t *s )
 405 {
 406     int val = bs_read_ue( s );
 407
 408     return val&0x01 ? (val+1)/2 : -(val/2);
 409 }
 410
 411
 412 static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
 413 {
 414     decoder_sys_t *p_sys = p_dec->p_sys;
 415     block_t *p_pic = NULL;
 416
 417     const int i_nal_ref_idc = (p_frag->p_buffer[3] >> 5)&0x03;
 418     const int i_nal_type = p_frag->p_buffer[3]&0x1f;
 419
 420     if( p_sys->b_slice && !p_sys->b_sps )
 421     {
 422         block_ChainRelease( p_sys->p_frame );
 423         msg_Warn( p_dec, "waiting for SPS" );
 424
 425         /* Reset context */
 426         p_sys->p_frame = NULL;
 427         p_sys->b_slice = VLC_FALSE;
 428     }
 429
 430     if( !p_sys->b_sps &&
 431         i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
 432     {
 433         p_sys->b_slice = VLC_TRUE;
 434         /* Fragment will be discarded later on */
 435     }
 436     else if( i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
 437     {
 438         uint8_t *dec;
 439         int i_dec, i_first_mb, i_slice_type, i_frame_num, i_pic_flags = 0;
 440         vlc_bool_t b_pic = VLC_FALSE;
 441         bs_t s;
 442
 443         /* do not convert the whole frame */
 444         nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[4],
 445                          __MIN( p_frag->i_buffer - 4, 60 ) );
 446         bs_init( &s, dec, i_dec );
 447
 448         /* first_mb_in_slice */
 449         i_first_mb = bs_read_ue( &s );
 450
 451         /* slice_type */
 452         switch( (i_slice_type = bs_read_ue( &s )) )
 453         {
 454             case 0: case 5:
 455                 i_pic_flags = BLOCK_FLAG_TYPE_P;
 456                 break;
 457             case 1: case 6:
 458                 i_pic_flags = BLOCK_FLAG_TYPE_B;
 459                 break;
 460             case 2: case 7:
 461                 i_pic_flags = BLOCK_FLAG_TYPE_I;
 462                 break;
 463             case 3: case 8: /* SP */
 464                 i_pic_flags = BLOCK_FLAG_TYPE_P;
 465                 break;
 466             case 4: case 9:
 467                 i_pic_flags = BLOCK_FLAG_TYPE_I;
 468                 break;
 469         }
 470
 471         /* pic_parameter_set_id */
 472         bs_read_ue( &s );
 473         /* frame_num */
 474         i_frame_num = bs_read( &s, p_sys->i_log2_max_frame_num + 4 );
 475
 476         /* Detection of the first VCL NAL unit of a primary coded picture
 477          * (cf. 7.4.1.2.4) */
 478         if( i_frame_num != p_sys->i_frame_num ||
 479             ( (i_nal_ref_idc != p_sys->i_nal_ref_idc) &&
 480               (!i_nal_ref_idc || !p_sys->i_nal_ref_idc) ) )
 481         {
 482             b_pic = VLC_TRUE;
 483         }
 484         p_sys->i_frame_num = i_frame_num;
 485         p_sys->i_nal_ref_idc = i_nal_ref_idc;
 486
 487         if( !p_sys->b_frame_mbs_only )
 488         {
 489             /* field_pic_flag */
 490             if( bs_read( &s, 1 ) )
 491             {
 492                 /* bottom_field_flag */
 493                 bs_read( &s, 1 );
 494             }
 495         }
 496
 497         if( i_nal_type == NAL_SLICE_IDR )
 498         {
 499             /* id_pic_id */
 500             int i_idr_pic_id = bs_read_ue( &s );
 501             if( p_sys->i_nal_type != i_nal_type ) b_pic = VLC_TRUE;
 502             if( p_sys->i_idr_pic_id != i_idr_pic_id ) b_pic = VLC_TRUE;
 503             p_sys->i_idr_pic_id = i_idr_pic_id;
 504         }
 505         p_sys->i_nal_type = i_nal_type;
 506
 507         if( b_pic && p_sys->b_slice )
 508         {
 509             p_pic = block_ChainGather( p_sys->p_frame );
 510             p_pic->i_dts = p_sys->i_dts;
 511             p_pic->i_pts = p_sys->i_pts;
 512             p_pic->i_length = 0;    /* FIXME */
 513             p_pic->i_flags = p_sys->i_flags;
 514
 515             /* Reset context */
 516             p_sys->p_frame = NULL;
 517             p_sys->b_slice = VLC_FALSE;
 518         }
 519
 520         p_sys->b_slice = VLC_TRUE;
 521         p_sys->i_flags = i_pic_flags;
 522         p_sys->i_dts   = p_frag->i_dts;
 523         p_sys->i_pts   = p_frag->i_pts;
 524
 525         free( dec );
 526     }
 527     else if( i_nal_type == NAL_SPS )
 528     {
 529         uint8_t *dec;
 530         int     i_dec;
 531         bs_t s;
 532         int i_tmp;
 533
 534         msg_Dbg( p_dec, "found NAL_SPS" );
 535
 536         p_sys->b_sps = VLC_TRUE;
 537
 538         nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[4],
 539                          p_frag->i_buffer - 4 );
 540
 541         bs_init( &s, dec, i_dec );
 542         /* Skip profile(8), constraint_set012, reserver(5), level(8) */
 543         bs_skip( &s, 8 + 1+1+1 + 5 + 8 );
 544         /* sps id */
 545         bs_read_ue( &s );
 546         /* Skip i_log2_max_frame_num */
 547         p_sys->i_log2_max_frame_num = bs_read_ue( &s );
 548         /* Read poc_type */
 549         i_tmp = bs_read_ue( &s );
 550         if( i_tmp == 0 )
 551         {
 552             /* skip i_log2_max_poc_lsb */
 553             bs_read_ue( &s );
 554         }
 555         else if( i_tmp == 1 )
 556         {
 557             int i_cycle;
 558             /* skip b_delta_pic_order_always_zero */
 559             bs_skip( &s, 1 );
 560             /* skip i_offset_for_non_ref_pic */
 561             bs_read_se( &s );
 562             /* skip i_offset_for_top_to_bottom_field */
 563             bs_read_se( &s );
 564             /* read i_num_ref_frames_in_poc_cycle */
 565             i_cycle = bs_read_ue( &s );
 566             if( i_cycle > 256 ) i_cycle = 256;
 567             while( i_cycle > 0 )
 568             {
 569                 /* skip i_offset_for_ref_frame */
 570                 bs_read_se(&s );
 571             }
 572         }
 573         /* i_num_ref_frames */
 574         bs_read_ue( &s );
 575         /* b_gaps_in_frame_num_value_allowed */
 576         bs_skip( &s, 1 );
 577
 578         /* Read size */
 579         p_dec->fmt_out.video.i_width  = 16 * ( bs_read_ue( &s ) + 1 );
 580         p_dec->fmt_out.video.i_height = 16 * ( bs_read_ue( &s ) + 1 );
 581
 582         /* b_frame_mbs_only */
 583         p_sys->b_frame_mbs_only = bs_read( &s, 1 );
 584         if( p_sys->b_frame_mbs_only == 0 )
 585         {
 586             bs_skip( &s, 1 );
 587         }
 588         /* b_direct8x8_inference */
 589         bs_skip( &s, 1 );
 590
 591         /* crop */
 592         i_tmp = bs_read( &s, 1 );
 593         if( i_tmp )
 594         {
 595             /* left */
 596             bs_read_ue( &s );
 597             /* right */
 598             bs_read_ue( &s );
 599             /* top */
 600             bs_read_ue( &s );
 601             /* bottom */
 602             bs_read_ue( &s );
 603         }
 604
 605         /* vui */
 606         i_tmp = bs_read( &s, 1 );
 607         if( i_tmp )
 608         {
 609             /* read the aspect ratio part if any FIXME check it */
 610             i_tmp = bs_read( &s, 1 );
 611             if( i_tmp )
 612             {
 613                 static const struct { int w, h; } sar[14] =
 614                 {
 615                     { 0,   0 }, { 1,   1 }, { 12, 11 }, { 10, 11 },
 616                     { 16, 11 }, { 40, 33 }, { 24, 11 }, { 20, 11 },
 617                     { 32, 11 }, { 80, 33 }, { 18, 11 }, { 15, 11 },
 618                     { 64, 33 }, { 160,99 },
 619                 };
 620                 int i_sar = bs_read( &s, 8 );
 621                 int w, h;
 622
 623                 if( i_sar < 14 )
 624                 {
 625                     w = sar[i_sar].w;
 626                     h = sar[i_sar].h;
 627                 }
 628                 else
 629                 {
 630                     w = bs_read( &s, 16 );
 631                     h = bs_read( &s, 16 );
 632                 }
 633                 p_dec->fmt_out.video.i_aspect =
 634                     VOUT_ASPECT_FACTOR * w / h * p_dec->fmt_out.video.i_width /
 635                     p_dec->fmt_out.video.i_height;
 636             }
 637         }
 638
 639         free( dec );
 640     }
 641     else if( i_nal_type == NAL_PPS )
 642     {
 643         bs_t s;
 644         bs_init( &s, &p_frag->p_buffer[4], p_frag->i_buffer - 4 );
 645
 646         /* TODO */
 647         msg_Dbg( p_dec, "found NAL_PPS" );
 648     }
 649
 650     /* Append the block */
 651     block_ChainAppend( &p_sys->p_frame, p_frag );
 652
 653     return p_pic;
 654 }