1 /*****************************************************************************
2 * h264.c: h264/avc video packetizer
3 *****************************************************************************
4 * Copyright (C) 2001, 2002, 2006 the VideoLAN team
7 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
8 * Eric Petit <titer@videolan.org>
9 * Gildas Bazin <gbazin@videolan.org>
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.
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.
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., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
24 *****************************************************************************/
26 /*****************************************************************************
28 *****************************************************************************/
29 #include <stdlib.h> /* malloc(), free() */
32 #include <vlc/decoder.h>
35 #include "vlc_block_helper.h"
38 /*****************************************************************************
40 *****************************************************************************/
41 static int Open ( vlc_object_t * );
42 static void Close( vlc_object_t * );
45 set_category( CAT_SOUT );
46 set_subcategory( SUBCAT_SOUT_PACKETIZER );
47 set_description( _("H.264 video packetizer") );
48 set_capability( "packetizer", 50 );
49 set_callbacks( Open, Close );
53 /****************************************************************************
55 ****************************************************************************/
56 static block_t *Packetize( decoder_t *, block_t ** );
57 static block_t *PacketizeAVC1( decoder_t *, block_t ** );
61 block_bytestream_t bytestream;
74 int i_avcC_length_size;
78 /* Useful values of the Sequence Parameter Set */
79 int i_log2_max_frame_num;
82 /* Useful values of the Slice Header */
102 NAL_SLICE_IDR = 5, /* ref_idc != 0 */
103 NAL_SEI = 6, /* ref_idc == 0 */
107 /* ref_idc == 0 for 6,9,10,11,12 */
112 NAL_PRIORITY_DISPOSABLE = 0,
113 NAL_PRIORITY_LOW = 1,
114 NAL_PRIORITY_HIGH = 2,
115 NAL_PRIORITY_HIGHEST = 3,
118 static block_t *ParseNALBlock( decoder_t *, block_t * );
120 static block_t *nal_get_annexeb( decoder_t *, uint8_t *p, int );
122 /*****************************************************************************
123 * Open: probe the packetizer and return score
124 * When opening after demux, the packetizer is only loaded AFTER the decoder
125 * That means that what you set in fmt_out is ignored by the decoder in this special case
126 *****************************************************************************/
127 static int Open( vlc_object_t *p_this )
129 decoder_t *p_dec = (decoder_t*)p_this;
130 decoder_sys_t *p_sys;
132 if( p_dec->fmt_in.i_codec != VLC_FOURCC( 'h', '2', '6', '4') &&
133 p_dec->fmt_in.i_codec != VLC_FOURCC( 'H', '2', '6', '4') &&
134 p_dec->fmt_in.i_codec != VLC_FOURCC( 'V', 'S', 'S', 'H') &&
135 p_dec->fmt_in.i_codec != VLC_FOURCC( 'v', 's', 's', 'h') &&
136 ( p_dec->fmt_in.i_codec != VLC_FOURCC( 'a', 'v', 'c', '1') ||
137 p_dec->fmt_in.i_extra < 7 ) )
142 /* Allocate the memory needed to store the decoder's structure */
143 if( ( p_dec->p_sys = p_sys = malloc( sizeof(decoder_sys_t) ) ) == NULL )
145 msg_Err( p_dec, "out of memory" );
148 p_sys->i_state = STATE_NOSYNC;
150 p_sys->startcode[0] = 0;
151 p_sys->startcode[1] = 0;
152 p_sys->startcode[2] = 0;
153 p_sys->startcode[3] = 1;
154 p_sys->bytestream = block_BytestreamInit( p_dec );
155 p_sys->b_slice = VLC_FALSE;
156 p_sys->p_frame = NULL;
157 p_sys->b_sps = VLC_FALSE;
158 p_sys->b_pps = VLC_FALSE;
162 p_sys->i_nal_type = -1;
163 p_sys->i_nal_ref_idc = -1;
164 p_sys->i_idr_pic_id = -1;
165 p_sys->i_frame_num = -1;
167 /* Setup properties */
168 es_format_Copy( &p_dec->fmt_out, &p_dec->fmt_in );
169 p_dec->fmt_out.i_codec = VLC_FOURCC( 'h', '2', '6', '4' );
171 if( p_dec->fmt_in.i_codec == VLC_FOURCC( 'a', 'v', 'c', '1' ) )
173 /* This type of stream is produced by mp4 and matroska
174 * when we want to store it in another streamformat, you need to convert
175 * The fmt_in.p_extra should ALWAYS contain the avcC
176 * The fmt_out.p_extra should contain all the SPS and PPS with 4 byte startcodes */
177 uint8_t *p = &((uint8_t*)p_dec->fmt_in.p_extra)[4];
182 p_sys->i_avcC_length_size = 1 + ((*p++)&0x03);
186 for( i = 0; i < i_sps; i++ )
188 int i_length = GetWBE( p );
189 block_t *p_sps = nal_get_annexeb( p_dec, p + 2, i_length );
191 p_sys->p_sps = block_Duplicate( p_sps );
192 p_sps->i_pts = p_sps->i_dts = mdate();
193 ParseNALBlock( p_dec, p_sps );
198 for( i = 0; i < i_pps; i++ )
200 int i_length = GetWBE( p );
201 block_t *p_pps = nal_get_annexeb( p_dec, p + 2, i_length );
203 p_sys->p_pps = block_Duplicate( p_pps );
204 p_pps->i_pts = p_pps->i_dts = mdate();
205 ParseNALBlock( p_dec, p_pps );
208 msg_Dbg( p_dec, "avcC length size=%d, sps=%d, pps=%d",
209 p_sys->i_avcC_length_size, i_sps, i_pps );
211 /* FIXME: FFMPEG isn't happy at all if you leave this */
212 if( p_dec->fmt_out.i_extra ) free( p_dec->fmt_out.p_extra );
213 p_dec->fmt_out.i_extra = 0; p_dec->fmt_out.p_extra = NULL;
215 /* Set the new extradata */
216 p_dec->fmt_out.i_extra = p_sys->p_pps->i_buffer + p_sys->p_sps->i_buffer;
217 p_dec->fmt_out.p_extra = (uint8_t*)malloc( p_dec->fmt_out.i_extra );
218 memcpy( p_dec->fmt_out.p_extra, p_sys->p_pps->p_buffer, p_sys->p_pps->i_buffer);
219 memcpy( p_dec->fmt_out.p_extra+p_sys->p_pps->i_buffer, p_sys->p_sps->p_buffer, p_sys->p_sps->i_buffer);
222 p_dec->pf_packetize = PacketizeAVC1;
226 /* This type of stream contains data with 3 of 4 byte startcodes
227 * The fmt_in.p_extra MAY contain SPS/PPS with 4 byte startcodes
228 * The fmt_out.p_extra should be the same */
231 p_dec->pf_packetize = Packetize;
234 if( p_dec->fmt_in.i_extra > 0 )
236 block_t *p_init = block_New( p_dec, p_dec->fmt_in.i_extra );
239 memcpy( p_init->p_buffer, p_dec->fmt_in.p_extra,
240 p_dec->fmt_in.i_extra );
242 while( ( p_pic = Packetize( p_dec, &p_init ) ) )
244 /* Should not occur because we should only receive SPS/PPS */
245 block_Release( p_pic );
253 /*****************************************************************************
254 * Close: clean up the packetizer
255 *****************************************************************************/
256 static void Close( vlc_object_t *p_this )
258 decoder_t *p_dec = (decoder_t*)p_this;
259 decoder_sys_t *p_sys = p_dec->p_sys;
261 if( p_sys->p_frame ) block_ChainRelease( p_sys->p_frame );
262 if( p_sys->p_sps ) block_Release( p_sys->p_sps );
263 if( p_sys->p_pps ) block_Release( p_sys->p_pps );
264 block_BytestreamRelease( &p_sys->bytestream );
268 /****************************************************************************
269 * Packetize: the whole thing
270 * Search for the startcodes 3 or more bytes
271 * Feed ParseNALBlock ALWAYS with 4 byte startcode prepended NALs
272 ****************************************************************************/
273 static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
275 decoder_sys_t *p_sys = p_dec->p_sys;
278 if( !pp_block || !*pp_block ) return NULL;
280 block_BytestreamPush( &p_sys->bytestream, *pp_block );
284 switch( p_sys->i_state )
287 /* Skip untill 3 byte startcode 0 0 1 */
288 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
289 &p_sys->i_offset, p_sys->startcode+1, 3 ) == VLC_SUCCESS)
291 p_sys->i_state = STATE_NEXT_SYNC;
294 if( p_sys->i_offset )
297 block_SkipBytes( &p_sys->bytestream, p_sys->i_offset );
299 block_BytestreamFlush( &p_sys->bytestream );
302 if( p_sys->i_state != STATE_NEXT_SYNC )
308 p_sys->i_offset = 1; /* To find next startcode */
310 case STATE_NEXT_SYNC:
311 /* Find the next 3 byte startcode 0 0 1*/
312 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
313 &p_sys->i_offset, p_sys->startcode+1, 3 ) != VLC_SUCCESS)
319 /* Get the new fragment and set the pts/dts */
320 p_pic = block_New( p_dec, p_sys->i_offset +1 );
321 p_pic->i_pts = p_sys->bytestream.p_block->i_pts;
322 p_pic->i_dts = p_sys->bytestream.p_block->i_dts;
323 /* Force 4 byte startcode 0 0 0 1 */
324 p_pic->p_buffer[0] = 0;
326 block_GetBytes( &p_sys->bytestream, &p_pic->p_buffer[1],
329 /* Remove trailing 0 bytes */
330 while( p_pic->i_buffer && (!p_pic->p_buffer[p_pic->i_buffer-1] ) ) p_pic->i_buffer--;
334 if( !( p_pic = ParseNALBlock( p_dec, p_pic ) ) )
336 p_sys->i_state = STATE_NOSYNC;
340 msg_Dbg( p_dec, "pts="I64Fd" dts="I64Fd,
341 p_pic->i_pts, p_pic->i_dts );
344 /* So p_block doesn't get re-added several times */
345 *pp_block = block_BytestreamPop( &p_sys->bytestream );
347 p_sys->i_state = STATE_NOSYNC;
354 /****************************************************************************
355 * PacketizeAVC1: Takes VCL blocks of data and creates annexe B type NAL stream
356 * Will always use 4 byte 0 0 0 1 startcodes
357 * Should prepend the SPS and PPS to the front of the stream
358 ****************************************************************************/
359 static block_t *PacketizeAVC1( decoder_t *p_dec, block_t **pp_block )
361 decoder_sys_t *p_sys = p_dec->p_sys;
363 block_t *p_ret = NULL;
366 if( !pp_block || !*pp_block ) return NULL;
372 if( //(p_block->i_flags & BLOCK_FLAG_TYPE_I) &&
373 p_sys->p_sps && p_sys->p_pps )
376 block_t *p_sps = block_Duplicate( p_sys->p_sps );
377 block_t *p_pps = block_Duplicate( p_sys->p_pps );
378 p_sps->i_dts = p_pps->i_dts = p_block->i_dts;
379 p_sps->i_pts = p_pps->i_pts = p_block->i_pts;
380 p_pic = ParseNALBlock( p_dec, p_sps );
381 if( p_pic ) block_ChainAppend( &p_ret, p_pic );
382 p_pic = ParseNALBlock( p_dec, p_pps );
383 if( p_pic ) block_ChainAppend( &p_ret, p_pic );
387 for( p = p_block->p_buffer; p < &p_block->p_buffer[p_block->i_buffer]; )
393 for( i = 0; i < p_sys->i_avcC_length_size; i++ )
395 i_size = (i_size << 8) | (*p++);
400 block_t *p_part = nal_get_annexeb( p_dec, p, i_size );
402 p_part->i_dts = p_block->i_dts;
403 p_part->i_pts = p_block->i_pts;
406 if( ( p_pic = ParseNALBlock( p_dec, p_part ) ) )
408 block_ChainAppend( &p_ret, p_pic );
413 block_Release( p_block );
418 static block_t *nal_get_annexeb( decoder_t *p_dec, uint8_t *p, int i_size )
422 p_nal = block_New( p_dec, 4 + i_size );
425 p_nal->p_buffer[0] = 0x00;
426 p_nal->p_buffer[1] = 0x00;
427 p_nal->p_buffer[2] = 0x00;
428 p_nal->p_buffer[3] = 0x01;
431 memcpy( &p_nal->p_buffer[4], p, i_size );
436 static void nal_get_decoded( uint8_t **pp_ret, int *pi_ret,
437 uint8_t *src, int i_src )
439 uint8_t *end = &src[i_src];
440 uint8_t *dst = malloc( i_src );
446 if( src < end - 3 && src[0] == 0x00 && src[1] == 0x00 &&
458 *pi_ret = dst - *pp_ret;
461 static inline int bs_read_ue( bs_t *s )
465 while( bs_read1( s ) == 0 && s->p < s->p_end && i < 32 )
469 return( ( 1 << i) - 1 + bs_read( s, i ) );
472 static inline int bs_read_se( bs_t *s )
474 int val = bs_read_ue( s );
476 return val&0x01 ? (val+1)/2 : -(val/2);
480 /*****************************************************************************
481 * ParseNALBlock: parses annexB type NALs
482 * All p_frag blocks are required to start with 0 0 0 1 4-byte startcode
483 *****************************************************************************/
484 static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
486 decoder_sys_t *p_sys = p_dec->p_sys;
487 block_t *p_pic = NULL;
489 const int i_nal_ref_idc = (p_frag->p_buffer[4] >> 5)&0x03;
490 const int i_nal_type = p_frag->p_buffer[4]&0x1f;
494 p_pic = block_ChainGather( p_sys->p_frame ); \
495 p_pic->i_length = 0; /* FIXME */ \
497 p_sys->p_frame = NULL; \
498 p_sys->b_slice = VLC_FALSE; \
502 if( p_sys->b_slice && !p_sys->b_sps )
504 block_ChainRelease( p_sys->p_frame );
505 msg_Warn( p_dec, "waiting for SPS" );
508 p_sys->p_frame = NULL;
509 p_sys->b_slice = VLC_FALSE;
513 i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
515 p_sys->b_slice = VLC_TRUE;
516 /* Fragment will be discarded later on */
518 else if( i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
521 int i_dec, i_first_mb, i_slice_type, i_frame_num, i_pic_flags = 0;
522 vlc_bool_t b_pic = VLC_FALSE;
525 /* do not convert the whole frame */
526 nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[5],
527 __MIN( p_frag->i_buffer - 5, 60 ) );
528 bs_init( &s, dec, i_dec );
530 /* first_mb_in_slice */
531 i_first_mb = bs_read_ue( &s );
534 switch( (i_slice_type = bs_read_ue( &s )) )
537 i_pic_flags = BLOCK_FLAG_TYPE_P;
540 i_pic_flags = BLOCK_FLAG_TYPE_B;
543 i_pic_flags = BLOCK_FLAG_TYPE_I;
545 case 3: case 8: /* SP */
546 i_pic_flags = BLOCK_FLAG_TYPE_P;
549 i_pic_flags = BLOCK_FLAG_TYPE_I;
553 /* pic_parameter_set_id */
556 i_frame_num = bs_read( &s, p_sys->i_log2_max_frame_num + 4 );
558 /* Detection of the first VCL NAL unit of a primary coded picture
560 if( i_frame_num != p_sys->i_frame_num ||
561 ( (i_nal_ref_idc != p_sys->i_nal_ref_idc) &&
562 (!i_nal_ref_idc || !p_sys->i_nal_ref_idc) ) )
566 p_sys->i_frame_num = i_frame_num;
567 p_sys->i_nal_ref_idc = i_nal_ref_idc;
569 if( !p_sys->b_frame_mbs_only )
572 if( bs_read( &s, 1 ) )
574 /* bottom_field_flag */
579 if( i_nal_type == NAL_SLICE_IDR )
582 int i_idr_pic_id = bs_read_ue( &s );
583 if( p_sys->i_nal_type != i_nal_type ) b_pic = VLC_TRUE;
584 if( p_sys->i_idr_pic_id != i_idr_pic_id ) b_pic = VLC_TRUE;
585 p_sys->i_idr_pic_id = i_idr_pic_id;
587 p_sys->i_nal_type = i_nal_type;
589 if( b_pic && p_sys->b_slice ) OUTPUT;
591 p_sys->b_slice = VLC_TRUE;
595 else if( i_nal_type == NAL_SPS )
602 if( !p_sys->b_sps ) msg_Dbg( p_dec, "found NAL_SPS" );
604 p_sys->b_sps = VLC_TRUE;
606 nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[5],
607 p_frag->i_buffer - 5 );
609 bs_init( &s, dec, i_dec );
610 /* Skip profile(8), constraint_set012, reserver(5), level(8) */
611 bs_skip( &s, 8 + 1+1+1 + 5 + 8 );
614 /* Skip i_log2_max_frame_num */
615 p_sys->i_log2_max_frame_num = bs_read_ue( &s );
617 i_tmp = bs_read_ue( &s );
620 /* skip i_log2_max_poc_lsb */
623 else if( i_tmp == 1 )
626 /* skip b_delta_pic_order_always_zero */
628 /* skip i_offset_for_non_ref_pic */
630 /* skip i_offset_for_top_to_bottom_field */
632 /* read i_num_ref_frames_in_poc_cycle */
633 i_cycle = bs_read_ue( &s );
634 if( i_cycle > 256 ) i_cycle = 256;
637 /* skip i_offset_for_ref_frame */
641 /* i_num_ref_frames */
643 /* b_gaps_in_frame_num_value_allowed */
647 p_dec->fmt_out.video.i_width = 16 * ( bs_read_ue( &s ) + 1 );
648 p_dec->fmt_out.video.i_height = 16 * ( bs_read_ue( &s ) + 1 );
650 /* b_frame_mbs_only */
651 p_sys->b_frame_mbs_only = bs_read( &s, 1 );
652 if( p_sys->b_frame_mbs_only == 0 )
656 /* b_direct8x8_inference */
660 i_tmp = bs_read( &s, 1 );
674 i_tmp = bs_read( &s, 1 );
677 /* read the aspect ratio part if any FIXME check it */
678 i_tmp = bs_read( &s, 1 );
681 static const struct { int w, h; } sar[14] =
683 { 0, 0 }, { 1, 1 }, { 12, 11 }, { 10, 11 },
684 { 16, 11 }, { 40, 33 }, { 24, 11 }, { 20, 11 },
685 { 32, 11 }, { 80, 33 }, { 18, 11 }, { 15, 11 },
686 { 64, 33 }, { 160,99 },
688 int i_sar = bs_read( &s, 8 );
698 w = bs_read( &s, 16 );
699 h = bs_read( &s, 16 );
702 p_dec->fmt_out.video.i_aspect = VOUT_ASPECT_FACTOR * w /
703 h * p_dec->fmt_out.video.i_width /
704 p_dec->fmt_out.video.i_height;
706 p_dec->fmt_out.video.i_aspect = VOUT_ASPECT_FACTOR;
713 if( p_sys->b_slice ) OUTPUT;
715 else if( i_nal_type == NAL_PPS )
718 bs_init( &s, &p_frag->p_buffer[5], p_frag->i_buffer - 5 );
720 if( !p_sys->b_pps ) msg_Dbg( p_dec, "found NAL_PPS" );
721 p_sys->b_pps = VLC_TRUE;
725 if( p_sys->b_slice ) OUTPUT;
727 else if( i_nal_type == NAL_AU_DELIMITER ||
728 i_nal_type == NAL_SEI ||
729 ( i_nal_type >= 13 && i_nal_type <= 18 ) )
731 if( p_sys->b_slice ) OUTPUT;
736 /* Append the block */
737 block_ChainAppend( &p_sys->p_frame, p_frag );