* h264: added avc1 -> h264 annexe B stream.
[vlc.git] / 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@netcourrier.com>
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
78 enum
79 {
80     STATE_NOSYNC,
81     STATE_NEXT_SYNC,
82 };
83
84 enum nal_unit_type_e
85 {
86     NAL_UNKNOWN = 0,
87     NAL_SLICE   = 1,
88     NAL_SLICE_DPA   = 2,
89     NAL_SLICE_DPB   = 3,
90     NAL_SLICE_DPC   = 4,
91     NAL_SLICE_IDR   = 5,    /* ref_idc != 0 */
92     NAL_SEI         = 6,    /* ref_idc == 0 */
93     NAL_SPS         = 7,
94     NAL_PPS         = 8
95     /* ref_idc == 0 for 6,9,10,11,12 */
96 };
97
98 enum nal_priority_e
99 {
100     NAL_PRIORITY_DISPOSABLE = 0,
101     NAL_PRIORITY_LOW        = 1,
102     NAL_PRIORITY_HIGH       = 2,
103     NAL_PRIORITY_HIGHEST    = 3,
104 };
105
106 static block_t *ParseNALBlock( decoder_t *, block_t * );
107
108 static block_t *nal_get_encoded( decoder_t *, uint8_t *p, int );
109
110 /*****************************************************************************
111  * Open: probe the packetizer and return score
112  *****************************************************************************/
113 static int Open( vlc_object_t *p_this )
114 {
115     decoder_t     *p_dec = (decoder_t*)p_this;
116     decoder_sys_t *p_sys;
117
118     if( p_dec->fmt_in.i_codec != VLC_FOURCC( 'h', '2', '6', '4') &&
119         p_dec->fmt_in.i_codec != VLC_FOURCC( 'H', '2', '6', '4') &&
120         ( p_dec->fmt_in.i_codec != VLC_FOURCC( 'a', 'v', 'c', '1') || p_dec->fmt_in.i_extra < 7 ) )
121     {
122         return VLC_EGENERIC;
123     }
124
125     /* Allocate the memory needed to store the decoder's structure */
126     if( ( p_dec->p_sys = p_sys = malloc( sizeof(decoder_sys_t) ) ) == NULL )
127     {
128         msg_Err( p_dec, "out of memory" );
129         return VLC_EGENERIC;
130     }
131     p_sys->i_state = STATE_NOSYNC;
132     p_sys->i_offset = 0;
133     p_sys->startcode[0] = 0;
134     p_sys->startcode[1] = 0;
135     p_sys->startcode[2] = 0;
136     p_sys->startcode[3] = 1;
137     p_sys->bytestream = block_BytestreamInit( p_dec );
138     p_sys->b_slice = VLC_FALSE;
139     p_sys->p_frame = NULL;
140     p_sys->i_dts   = 0;
141     p_sys->i_pts   = 0;
142     p_sys->i_flags = 0;
143     p_sys->b_sps   = VLC_FALSE;
144
145     /* Setup properties */
146     es_format_Copy( &p_dec->fmt_out, &p_dec->fmt_in );
147     p_dec->fmt_out.i_codec = VLC_FOURCC( 'h', '2', '6', '4' );
148
149     if( p_dec->fmt_in.i_codec == VLC_FOURCC( 'a', 'v', 'c', '1' ) )
150     {
151         uint8_t *p = &((uint8_t*)p_dec->fmt_in.p_extra)[4];
152         int i_sps, i_pps;
153         int i;
154
155         /* Parse avcC */
156         p_sys->i_avcC_length_size = 1 + ((*p++)&0x03);
157
158         /* Read SPS */
159         i_sps = (*p++)&0x1f;
160
161         for( i = 0; i < i_sps; i++ )
162         {
163             int i_length = GetWBE( p );
164             block_t *p_sps = nal_get_encoded( p_dec, p+2, i_length );
165
166             ParseNALBlock( p_dec, p_sps );
167             p += 2 + i_length;
168         }
169         /* Read PPS */
170         i_pps = *p++;
171         for( i = 0; i < i_pps; i++ )
172         {
173             int i_length = GetWBE( p );
174             block_t *p_pps = nal_get_encoded( p_dec, p+2, i_length );
175
176             ParseNALBlock( p_dec, p_pps );
177             p += 2 + i_length;
178         }
179         msg_Dbg( p_dec, "avcC length size=%d sps=%d pps=%d",
180                  p_sys->i_avcC_length_size, i_sps, i_pps );
181
182         /* Set callback */
183         p_dec->pf_packetize = PacketizeAVC1;
184     }
185     else
186     {
187         /* Set callback */
188         p_dec->pf_packetize = Packetize;
189     }
190
191     return VLC_SUCCESS;
192 }
193
194 /*****************************************************************************
195  * Close: clean up the packetizer
196  *****************************************************************************/
197 static void Close( vlc_object_t *p_this )
198 {
199     decoder_t *p_dec = (decoder_t*)p_this;
200     decoder_sys_t *p_sys = p_dec->p_sys;
201
202     block_BytestreamRelease( &p_sys->bytestream );
203     free( p_sys );
204 }
205
206 /****************************************************************************
207  * Packetize: the whole thing
208  ****************************************************************************/
209 static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
210 {
211     decoder_sys_t *p_sys = p_dec->p_sys;
212     block_t       *p_pic;
213
214     if( !pp_block || !*pp_block ) return NULL;
215
216     block_BytestreamPush( &p_sys->bytestream, *pp_block );
217
218     for( ;; )
219     {
220         switch( p_sys->i_state )
221         {
222             case STATE_NOSYNC:
223                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
224                         &p_sys->i_offset, p_sys->startcode, 4 ) == VLC_SUCCESS )
225                 {
226                     p_sys->i_state = STATE_NEXT_SYNC;
227                 }
228
229                 if( p_sys->i_offset )
230                 {
231                     block_SkipBytes( &p_sys->bytestream, p_sys->i_offset );
232                     p_sys->i_offset = 0;
233                     block_BytestreamFlush( &p_sys->bytestream );
234                 }
235
236                 if( p_sys->i_state != STATE_NEXT_SYNC )
237                 {
238                     /* Need more data */
239                     return NULL;
240                 }
241
242                 p_sys->i_offset = 1; /* To find next startcode */
243
244             case STATE_NEXT_SYNC:
245                 /* Find the next startcode */
246                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
247                         &p_sys->i_offset, p_sys->startcode, 4 ) != VLC_SUCCESS )
248                 {
249                     /* Need more data */
250                     return NULL;
251                 }
252
253                 /* Get the new fragment and set the pts/dts */
254                 p_pic = block_New( p_dec, p_sys->i_offset );
255                 p_pic->i_pts = p_sys->bytestream.p_block->i_pts;
256                 p_pic->i_dts = p_sys->bytestream.p_block->i_dts;
257
258                 block_GetBytes( &p_sys->bytestream, p_pic->p_buffer,
259                                 p_pic->i_buffer );
260
261                 p_sys->i_offset = 0;
262
263                 /* Parse the NAL */
264                 if( !( p_pic = ParseNALBlock( p_dec, p_pic ) ) )
265                 {
266                     p_sys->i_state = STATE_NOSYNC;
267                     break;
268                 }
269
270                 /* So p_block doesn't get re-added several times */
271                 *pp_block = block_BytestreamPop( &p_sys->bytestream );
272
273                 p_sys->i_state = STATE_NOSYNC;
274
275                 return p_pic;
276         }
277     }
278 }
279
280 /****************************************************************************
281  * PacketizeAVC1: the whole thing
282  ****************************************************************************/
283 static block_t *PacketizeAVC1( decoder_t *p_dec, block_t **pp_block )
284 {
285     decoder_sys_t *p_sys = p_dec->p_sys;
286     block_t       *p_block;
287     block_t       *p_ret = NULL;
288     uint8_t       *p;
289
290     if( !pp_block || !*pp_block ) return NULL;
291
292     p_block = *pp_block;
293     *pp_block = NULL;
294
295     for( p = p_block->p_buffer; p < &p_block->p_buffer[p_block->i_buffer]; )
296     {
297         block_t *p_pic;
298         int i_size = 0;
299         int i;
300
301         for( i = 0; i < p_sys->i_avcC_length_size; i++ )
302         {
303             i_size = (i_size << 8) | (*p++);
304         }
305
306         if( i_size > 0 )
307         {
308             block_t *p_part = nal_get_encoded( p_dec, p, i_size );
309
310             p_part->i_dts = p_block->i_dts;
311             p_part->i_pts = p_block->i_pts;
312             /* Parse the NAL */
313             if( ( p_pic = ParseNALBlock( p_dec, p_part ) ) )
314             {
315                 block_ChainAppend( &p_ret, p_pic );
316             }
317         }
318         p += i_size;
319     }
320
321     return p_ret;
322 }
323
324 static block_t *nal_get_encoded( decoder_t *p_dec, uint8_t *p, int i_size )
325 {
326     block_t *p_nal;
327     int     i_nal_size = 5;
328     uint8_t *src = &p[1];
329     uint8_t *end = &p[i_size];
330     uint8_t *dst;
331     int     i_count = 0;
332
333     /* 1: compute real size */
334     while( src < end )
335     {
336         if( i_count == 2 && *src <= 0x03 )
337         {
338             i_nal_size++;
339             i_count = 0;
340         }
341         if( *src == 0 )
342         {
343             i_count++;
344         }
345         else
346         {
347             i_count = 0;
348         }
349         i_nal_size++;
350         src++;
351     }
352
353     /* 2: encode it */
354     p_nal = block_New( p_dec, i_nal_size );
355     i_count = 0;
356     src = p;
357     dst = p_nal->p_buffer;
358
359     /* add start code */
360     *dst++ = 0x00;
361     *dst++ = 0x00;
362     *dst++ = 0x00;
363     *dst++ = 0x01;
364
365     /* nal type */
366     *dst++ = *src++;
367
368     while( src < end )
369     {
370         if( i_count == 2 && *src <= 0x03 )
371         {
372             *dst++ = 0x03;
373             i_count = 0;
374         }
375         if( *src == 0 )
376         {
377             i_count++;
378         }
379         else
380         {
381             i_count = 0;
382         }
383         *dst++ = *src++;
384     }
385
386     return p_nal;
387 }
388
389 static void nal_get_decoded( uint8_t **pp_ret, int *pi_ret, uint8_t *src, int i_src )
390 {
391     uint8_t *end = &src[i_src];
392     uint8_t *dst = malloc( i_src );
393
394     *pp_ret = dst;
395
396     while( src < end )
397     {
398         if( src < end - 3 && src[0] == 0x00 && src[1] == 0x00  && src[2] == 0x03 )
399         {
400             *dst++ = 0x00;
401             *dst++ = 0x00;
402
403             src += 3;
404             continue;
405         }
406         *dst++ = *src++;
407     }
408
409     *pi_ret = dst - *pp_ret;
410 }
411
412 static inline int bs_read_ue( bs_t *s )
413 {
414     int i = 0;
415
416     while( bs_read1( s ) == 0 && s->p < s->p_end && i < 32 )
417     {
418         i++;
419     }
420     return( ( 1 << i) - 1 + bs_read( s, i ) );
421 }
422 static inline int bs_read_se( bs_t *s )
423 {
424     int val = bs_read_ue( s );
425
426     return val&0x01 ? (val+1)/2 : -(val/2);
427 }
428
429
430 static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
431 {
432     decoder_sys_t *p_sys = p_dec->p_sys;
433     block_t *p_pic = NULL;
434
435     const int i_ref_idc = (p_frag->p_buffer[4] >> 5)&0x03;
436     const int i_nal_type= p_frag->p_buffer[4]&0x1f;
437
438     if( p_sys->b_slice &&
439         ( i_nal_type == NAL_SLICE || i_nal_type == NAL_SLICE_IDR ||
440           i_nal_type == NAL_SLICE_DPC || i_nal_type == NAL_SPS || i_nal_type == NAL_PPS ) )
441     {
442         if( p_sys->b_sps )
443         {
444             p_pic = block_ChainGather( p_sys->p_frame );
445             p_pic->i_dts = p_sys->i_dts;
446             p_pic->i_pts = p_sys->i_pts;
447             p_pic->i_length = 0;    /* FIXME */
448             p_pic->i_flags = p_sys->i_flags;
449         }
450         else
451         {
452             block_ChainRelease( p_sys->p_frame );
453             msg_Warn( p_dec, "waiting SPS" );
454         }
455
456         /* reset context */
457         p_sys->p_frame = NULL;
458         p_sys->b_slice = VLC_FALSE;
459         //p_sys->i_dts += 40000;
460     }
461
462     if( i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
463     {
464         uint8_t *dec;
465         int     i_dec;
466         bs_t s;
467
468         p_sys->b_slice = VLC_TRUE;
469         p_sys->i_dts   = p_frag->i_dts;
470         p_sys->i_pts   = p_frag->i_pts;
471
472         /* do not convert the whole frame */
473         nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[5], __MIN( p_frag->i_buffer - 5, 60 ) );
474         bs_init( &s, dec, i_dec );
475
476         /* i_first_mb */
477         bs_read_ue( &s );
478         /* picture type */
479         switch( bs_read_ue( &s ) )
480         {
481             case 0: case 5:
482                 p_sys->i_flags = BLOCK_FLAG_TYPE_P;
483                 break;
484             case 1: case 6:
485                 p_sys->i_flags =BLOCK_FLAG_TYPE_B;
486                 break;
487             case 2: case 7:
488                 p_sys->i_flags = BLOCK_FLAG_TYPE_I;
489                 break;
490             case 3: case 8: /* SP */
491                 p_sys->i_flags = BLOCK_FLAG_TYPE_P;
492                 break;
493             case 4: case 9:
494                 p_sys->i_flags = BLOCK_FLAG_TYPE_I;
495                 break;
496         }
497
498         free( dec );
499     }
500     else if( i_nal_type == NAL_SPS )
501     {
502         uint8_t *dec;
503         int     i_dec;
504         bs_t s;
505         int i_tmp;
506
507         p_sys->b_sps = VLC_TRUE;
508
509         nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[5], p_frag->i_buffer - 5 );
510
511         bs_init( &s, dec, i_dec );
512         /* Skip profile(8), constraint_set012, reserver(5), level(8) */
513         bs_skip( &s, 8 + 1+1+1 + 5 + 8 );
514         /* sps id */
515         bs_read_ue( &s );
516         /* Skip i_log2_max_frame_num */
517         bs_read_ue( &s );
518         /* Read poc_type */
519         i_tmp = bs_read_ue( &s );
520         if( i_tmp == 0 )
521         {
522             /* skip i_log2_max_poc_lsb */
523             bs_read_ue( &s );
524         }
525         else if( i_tmp == 1 )
526         {
527             int i_cycle;
528             /* skip b_delta_pic_order_always_zero */
529             bs_skip( &s, 1 );
530             /* skip i_offset_for_non_ref_pic */
531             bs_read_se( &s );
532             /* skip i_offset_for_top_to_bottom_field */
533             bs_read_se( &s );
534             /* read i_num_ref_frames_in_poc_cycle */
535             i_cycle = bs_read_ue( &s );
536             if( i_cycle > 256 ) i_cycle = 256;
537             while( i_cycle > 0 )
538             {
539                 /* skip i_offset_for_ref_frame */
540                 bs_read_se(&s );
541             }
542         }
543         /* i_num_ref_frames */
544         bs_read_ue( &s );
545         /* b_gaps_in_frame_num_value_allowed */
546         bs_skip( &s, 1 );
547
548         /* Read size */
549         p_dec->fmt_out.video.i_width  = 16 * ( bs_read_ue( &s ) + 1 );
550         p_dec->fmt_out.video.i_height = 16 * ( bs_read_ue( &s ) + 1 );
551
552         /* b_frame_mbs_only */
553         i_tmp = bs_read( &s, 1 );
554         if( i_tmp == 0 )
555         {
556             bs_skip( &s, 1 );
557         }
558         /* b_direct8x8_inference */
559         bs_skip( &s, 1 );
560
561         /* crop ? */
562         i_tmp = bs_read( &s, 1 );
563         if( i_tmp )
564         {
565             /* left */
566             p_dec->fmt_out.video.i_width -= 2 * bs_read_ue( &s );
567             /* right */
568             p_dec->fmt_out.video.i_width -= 2 * bs_read_ue( &s );
569             /* top */
570             p_dec->fmt_out.video.i_height -= 2 * bs_read_ue( &s );
571             /* bottom */
572             p_dec->fmt_out.video.i_height -= 2 * bs_read_ue( &s );
573         }
574
575         /* vui */
576         i_tmp = bs_read( &s, 1 );
577         if( i_tmp )
578         {
579             /* read the aspect ratio part if any FIXME check it */
580             i_tmp = bs_read( &s, 1 );
581             if( i_tmp )
582             {
583                 static const struct { int w, h; } sar[14] =
584                 {
585                     { 0,   0 }, { 1,   1 }, { 12, 11 }, { 10, 11 },
586                     { 16, 11 }, { 40, 33 }, { 24, 11 }, { 20, 11 },
587                     { 32, 11 }, { 80, 33 }, { 18, 11 }, { 15, 11 },
588                     { 64, 33 }, { 160,99 },
589                 };
590                 int i_sar = bs_read( &s, 8 );
591                 int w, h;
592
593                 if( i_sar < 14 )
594                 {
595                     w = sar[i_sar].w;
596                     h = sar[i_sar].h;
597                 }
598                 else
599                 {
600                     w = bs_read( &s, 16 );
601                     h = bs_read( &s, 16 );
602                 }
603                 p_dec->fmt_out.video.i_aspect =
604                     VOUT_ASPECT_FACTOR *
605                     w / h *
606                     p_dec->fmt_out.video.i_width / p_dec->fmt_out.video.i_height;
607             }
608         }
609
610         free( dec );
611     }
612     else if( i_nal_type == NAL_PPS )
613     {
614         bs_t s;
615         bs_init( &s, &p_frag->p_buffer[5], p_frag->i_buffer - 5 );
616
617         /* TODO */
618     }
619
620
621     /* Append the block */
622     block_ChainAppend( &p_sys->p_frame, p_frag );
623
624     return p_pic;
625 }
626