* all: added VSSH fourcc.
[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( 'V', 'S', 'S', 'H') &&
121         p_dec->fmt_in.i_codec != VLC_FOURCC( 'v', 's', 's', 'h') &&
122         ( p_dec->fmt_in.i_codec != VLC_FOURCC( 'a', 'v', 'c', '1') || p_dec->fmt_in.i_extra < 7 ) )
123     {
124         return VLC_EGENERIC;
125     }
126
127     /* Allocate the memory needed to store the decoder's structure */
128     if( ( p_dec->p_sys = p_sys = malloc( sizeof(decoder_sys_t) ) ) == NULL )
129     {
130         msg_Err( p_dec, "out of memory" );
131         return VLC_EGENERIC;
132     }
133     p_sys->i_state = STATE_NOSYNC;
134     p_sys->i_offset = 0;
135     p_sys->startcode[0] = 0;
136     p_sys->startcode[1] = 0;
137     p_sys->startcode[2] = 0;
138     p_sys->startcode[3] = 1;
139     p_sys->bytestream = block_BytestreamInit( p_dec );
140     p_sys->b_slice = VLC_FALSE;
141     p_sys->p_frame = NULL;
142     p_sys->i_dts   = 0;
143     p_sys->i_pts   = 0;
144     p_sys->i_flags = 0;
145     p_sys->b_sps   = VLC_FALSE;
146
147     /* Setup properties */
148     es_format_Copy( &p_dec->fmt_out, &p_dec->fmt_in );
149     p_dec->fmt_out.i_codec = VLC_FOURCC( 'h', '2', '6', '4' );
150
151     if( p_dec->fmt_in.i_codec == VLC_FOURCC( 'a', 'v', 'c', '1' ) )
152     {
153         uint8_t *p = &((uint8_t*)p_dec->fmt_in.p_extra)[4];
154         int i_sps, i_pps;
155         int i;
156
157         /* Parse avcC */
158         p_sys->i_avcC_length_size = 1 + ((*p++)&0x03);
159
160         /* Read SPS */
161         i_sps = (*p++)&0x1f;
162
163         for( i = 0; i < i_sps; i++ )
164         {
165             int i_length = GetWBE( p );
166             block_t *p_sps = nal_get_encoded( p_dec, p+2, i_length );
167
168             ParseNALBlock( p_dec, p_sps );
169             p += 2 + i_length;
170         }
171         /* Read PPS */
172         i_pps = *p++;
173         for( i = 0; i < i_pps; i++ )
174         {
175             int i_length = GetWBE( p );
176             block_t *p_pps = nal_get_encoded( p_dec, p+2, i_length );
177
178             ParseNALBlock( p_dec, p_pps );
179             p += 2 + i_length;
180         }
181         msg_Dbg( p_dec, "avcC length size=%d sps=%d pps=%d",
182                  p_sys->i_avcC_length_size, i_sps, i_pps );
183
184         /* Set callback */
185         p_dec->pf_packetize = PacketizeAVC1;
186     }
187     else
188     {
189         /* Set callback */
190         p_dec->pf_packetize = Packetize;
191     }
192
193     return VLC_SUCCESS;
194 }
195
196 /*****************************************************************************
197  * Close: clean up the packetizer
198  *****************************************************************************/
199 static void Close( vlc_object_t *p_this )
200 {
201     decoder_t *p_dec = (decoder_t*)p_this;
202     decoder_sys_t *p_sys = p_dec->p_sys;
203
204     block_BytestreamRelease( &p_sys->bytestream );
205     free( p_sys );
206 }
207
208 /****************************************************************************
209  * Packetize: the whole thing
210  ****************************************************************************/
211 static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
212 {
213     decoder_sys_t *p_sys = p_dec->p_sys;
214     block_t       *p_pic;
215
216     if( !pp_block || !*pp_block ) return NULL;
217
218     block_BytestreamPush( &p_sys->bytestream, *pp_block );
219
220     for( ;; )
221     {
222         switch( p_sys->i_state )
223         {
224             case STATE_NOSYNC:
225                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
226                         &p_sys->i_offset, p_sys->startcode, 4 ) == VLC_SUCCESS )
227                 {
228                     p_sys->i_state = STATE_NEXT_SYNC;
229                 }
230
231                 if( p_sys->i_offset )
232                 {
233                     block_SkipBytes( &p_sys->bytestream, p_sys->i_offset );
234                     p_sys->i_offset = 0;
235                     block_BytestreamFlush( &p_sys->bytestream );
236                 }
237
238                 if( p_sys->i_state != STATE_NEXT_SYNC )
239                 {
240                     /* Need more data */
241                     return NULL;
242                 }
243
244                 p_sys->i_offset = 1; /* To find next startcode */
245
246             case STATE_NEXT_SYNC:
247                 /* Find the next startcode */
248                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
249                         &p_sys->i_offset, p_sys->startcode, 4 ) != VLC_SUCCESS )
250                 {
251                     /* Need more data */
252                     return NULL;
253                 }
254
255                 /* Get the new fragment and set the pts/dts */
256                 p_pic = block_New( p_dec, p_sys->i_offset );
257                 p_pic->i_pts = p_sys->bytestream.p_block->i_pts;
258                 p_pic->i_dts = p_sys->bytestream.p_block->i_dts;
259
260                 block_GetBytes( &p_sys->bytestream, p_pic->p_buffer,
261                                 p_pic->i_buffer );
262
263                 p_sys->i_offset = 0;
264
265                 /* Parse the NAL */
266                 if( !( p_pic = ParseNALBlock( p_dec, p_pic ) ) )
267                 {
268                     p_sys->i_state = STATE_NOSYNC;
269                     break;
270                 }
271
272                 /* So p_block doesn't get re-added several times */
273                 *pp_block = block_BytestreamPop( &p_sys->bytestream );
274
275                 p_sys->i_state = STATE_NOSYNC;
276
277                 return p_pic;
278         }
279     }
280 }
281
282 /****************************************************************************
283  * PacketizeAVC1: the whole thing
284  ****************************************************************************/
285 static block_t *PacketizeAVC1( decoder_t *p_dec, block_t **pp_block )
286 {
287     decoder_sys_t *p_sys = p_dec->p_sys;
288     block_t       *p_block;
289     block_t       *p_ret = NULL;
290     uint8_t       *p;
291
292     if( !pp_block || !*pp_block ) return NULL;
293
294     p_block = *pp_block;
295     *pp_block = NULL;
296
297     for( p = p_block->p_buffer; p < &p_block->p_buffer[p_block->i_buffer]; )
298     {
299         block_t *p_pic;
300         int i_size = 0;
301         int i;
302
303         for( i = 0; i < p_sys->i_avcC_length_size; i++ )
304         {
305             i_size = (i_size << 8) | (*p++);
306         }
307
308         if( i_size > 0 )
309         {
310             block_t *p_part = nal_get_encoded( p_dec, p, i_size );
311
312             p_part->i_dts = p_block->i_dts;
313             p_part->i_pts = p_block->i_pts;
314             /* Parse the NAL */
315             if( ( p_pic = ParseNALBlock( p_dec, p_part ) ) )
316             {
317                 block_ChainAppend( &p_ret, p_pic );
318             }
319         }
320         p += i_size;
321     }
322
323     return p_ret;
324 }
325
326 static block_t *nal_get_encoded( decoder_t *p_dec, uint8_t *p, int i_size )
327 {
328     block_t *p_nal;
329     int     i_nal_size = 5;
330     uint8_t *src = &p[1];
331     uint8_t *end = &p[i_size];
332     uint8_t *dst;
333     int     i_count = 0;
334
335     /* 1: compute real size */
336     while( src < end )
337     {
338         if( i_count == 2 && *src <= 0x03 )
339         {
340             i_nal_size++;
341             i_count = 0;
342         }
343         if( *src == 0 )
344         {
345             i_count++;
346         }
347         else
348         {
349             i_count = 0;
350         }
351         i_nal_size++;
352         src++;
353     }
354
355     /* 2: encode it */
356     p_nal = block_New( p_dec, i_nal_size );
357     i_count = 0;
358     src = p;
359     dst = p_nal->p_buffer;
360
361     /* add start code */
362     *dst++ = 0x00;
363     *dst++ = 0x00;
364     *dst++ = 0x00;
365     *dst++ = 0x01;
366
367     /* nal type */
368     *dst++ = *src++;
369
370     while( src < end )
371     {
372         if( i_count == 2 && *src <= 0x03 )
373         {
374             *dst++ = 0x03;
375             i_count = 0;
376         }
377         if( *src == 0 )
378         {
379             i_count++;
380         }
381         else
382         {
383             i_count = 0;
384         }
385         *dst++ = *src++;
386     }
387
388     return p_nal;
389 }
390
391 static void nal_get_decoded( uint8_t **pp_ret, int *pi_ret, uint8_t *src, int i_src )
392 {
393     uint8_t *end = &src[i_src];
394     uint8_t *dst = malloc( i_src );
395
396     *pp_ret = dst;
397
398     while( src < end )
399     {
400         if( src < end - 3 && src[0] == 0x00 && src[1] == 0x00  && src[2] == 0x03 )
401         {
402             *dst++ = 0x00;
403             *dst++ = 0x00;
404
405             src += 3;
406             continue;
407         }
408         *dst++ = *src++;
409     }
410
411     *pi_ret = dst - *pp_ret;
412 }
413
414 static inline int bs_read_ue( bs_t *s )
415 {
416     int i = 0;
417
418     while( bs_read1( s ) == 0 && s->p < s->p_end && i < 32 )
419     {
420         i++;
421     }
422     return( ( 1 << i) - 1 + bs_read( s, i ) );
423 }
424 static inline int bs_read_se( bs_t *s )
425 {
426     int val = bs_read_ue( s );
427
428     return val&0x01 ? (val+1)/2 : -(val/2);
429 }
430
431
432 static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
433 {
434     decoder_sys_t *p_sys = p_dec->p_sys;
435     block_t *p_pic = NULL;
436
437     const int i_ref_idc = (p_frag->p_buffer[4] >> 5)&0x03;
438     const int i_nal_type= p_frag->p_buffer[4]&0x1f;
439
440     if( p_sys->b_slice &&
441         ( i_nal_type == NAL_SLICE || i_nal_type == NAL_SLICE_IDR ||
442           i_nal_type == NAL_SLICE_DPC || i_nal_type == NAL_SPS || i_nal_type == NAL_PPS ) )
443     {
444         if( p_sys->b_sps )
445         {
446             p_pic = block_ChainGather( p_sys->p_frame );
447             p_pic->i_dts = p_sys->i_dts;
448             p_pic->i_pts = p_sys->i_pts;
449             p_pic->i_length = 0;    /* FIXME */
450             p_pic->i_flags = p_sys->i_flags;
451         }
452         else
453         {
454             block_ChainRelease( p_sys->p_frame );
455             msg_Warn( p_dec, "waiting SPS" );
456         }
457
458         /* reset context */
459         p_sys->p_frame = NULL;
460         p_sys->b_slice = VLC_FALSE;
461         //p_sys->i_dts += 40000;
462     }
463
464     if( i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
465     {
466         uint8_t *dec;
467         int     i_dec;
468         bs_t s;
469
470         p_sys->b_slice = VLC_TRUE;
471         p_sys->i_dts   = p_frag->i_dts;
472         p_sys->i_pts   = p_frag->i_pts;
473
474         /* do not convert the whole frame */
475         nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[5], __MIN( p_frag->i_buffer - 5, 60 ) );
476         bs_init( &s, dec, i_dec );
477
478         /* i_first_mb */
479         bs_read_ue( &s );
480         /* picture type */
481         switch( bs_read_ue( &s ) )
482         {
483             case 0: case 5:
484                 p_sys->i_flags = BLOCK_FLAG_TYPE_P;
485                 break;
486             case 1: case 6:
487                 p_sys->i_flags =BLOCK_FLAG_TYPE_B;
488                 break;
489             case 2: case 7:
490                 p_sys->i_flags = BLOCK_FLAG_TYPE_I;
491                 break;
492             case 3: case 8: /* SP */
493                 p_sys->i_flags = BLOCK_FLAG_TYPE_P;
494                 break;
495             case 4: case 9:
496                 p_sys->i_flags = BLOCK_FLAG_TYPE_I;
497                 break;
498         }
499
500         free( dec );
501     }
502     else if( i_nal_type == NAL_SPS )
503     {
504         uint8_t *dec;
505         int     i_dec;
506         bs_t s;
507         int i_tmp;
508
509         p_sys->b_sps = VLC_TRUE;
510
511         nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[5], p_frag->i_buffer - 5 );
512
513         bs_init( &s, dec, i_dec );
514         /* Skip profile(8), constraint_set012, reserver(5), level(8) */
515         bs_skip( &s, 8 + 1+1+1 + 5 + 8 );
516         /* sps id */
517         bs_read_ue( &s );
518         /* Skip i_log2_max_frame_num */
519         bs_read_ue( &s );
520         /* Read poc_type */
521         i_tmp = bs_read_ue( &s );
522         if( i_tmp == 0 )
523         {
524             /* skip i_log2_max_poc_lsb */
525             bs_read_ue( &s );
526         }
527         else if( i_tmp == 1 )
528         {
529             int i_cycle;
530             /* skip b_delta_pic_order_always_zero */
531             bs_skip( &s, 1 );
532             /* skip i_offset_for_non_ref_pic */
533             bs_read_se( &s );
534             /* skip i_offset_for_top_to_bottom_field */
535             bs_read_se( &s );
536             /* read i_num_ref_frames_in_poc_cycle */
537             i_cycle = bs_read_ue( &s );
538             if( i_cycle > 256 ) i_cycle = 256;
539             while( i_cycle > 0 )
540             {
541                 /* skip i_offset_for_ref_frame */
542                 bs_read_se(&s );
543             }
544         }
545         /* i_num_ref_frames */
546         bs_read_ue( &s );
547         /* b_gaps_in_frame_num_value_allowed */
548         bs_skip( &s, 1 );
549
550         /* Read size */
551         p_dec->fmt_out.video.i_width  = 16 * ( bs_read_ue( &s ) + 1 );
552         p_dec->fmt_out.video.i_height = 16 * ( bs_read_ue( &s ) + 1 );
553
554         /* b_frame_mbs_only */
555         i_tmp = bs_read( &s, 1 );
556         if( i_tmp == 0 )
557         {
558             bs_skip( &s, 1 );
559         }
560         /* b_direct8x8_inference */
561         bs_skip( &s, 1 );
562
563         /* crop ? */
564         i_tmp = bs_read( &s, 1 );
565         if( i_tmp )
566         {
567             /* left */
568             p_dec->fmt_out.video.i_width -= 2 * bs_read_ue( &s );
569             /* right */
570             p_dec->fmt_out.video.i_width -= 2 * bs_read_ue( &s );
571             /* top */
572             p_dec->fmt_out.video.i_height -= 2 * bs_read_ue( &s );
573             /* bottom */
574             p_dec->fmt_out.video.i_height -= 2 * bs_read_ue( &s );
575         }
576
577         /* vui */
578         i_tmp = bs_read( &s, 1 );
579         if( i_tmp )
580         {
581             /* read the aspect ratio part if any FIXME check it */
582             i_tmp = bs_read( &s, 1 );
583             if( i_tmp )
584             {
585                 static const struct { int w, h; } sar[14] =
586                 {
587                     { 0,   0 }, { 1,   1 }, { 12, 11 }, { 10, 11 },
588                     { 16, 11 }, { 40, 33 }, { 24, 11 }, { 20, 11 },
589                     { 32, 11 }, { 80, 33 }, { 18, 11 }, { 15, 11 },
590                     { 64, 33 }, { 160,99 },
591                 };
592                 int i_sar = bs_read( &s, 8 );
593                 int w, h;
594
595                 if( i_sar < 14 )
596                 {
597                     w = sar[i_sar].w;
598                     h = sar[i_sar].h;
599                 }
600                 else
601                 {
602                     w = bs_read( &s, 16 );
603                     h = bs_read( &s, 16 );
604                 }
605                 p_dec->fmt_out.video.i_aspect =
606                     VOUT_ASPECT_FACTOR *
607                     w / h *
608                     p_dec->fmt_out.video.i_width / p_dec->fmt_out.video.i_height;
609             }
610         }
611
612         free( dec );
613     }
614     else if( i_nal_type == NAL_PPS )
615     {
616         bs_t s;
617         bs_init( &s, &p_frag->p_buffer[5], p_frag->i_buffer - 5 );
618
619         /* TODO */
620     }
621
622
623     /* Append the block */
624     block_ChainAppend( &p_sys->p_frame, p_frag );
625
626     return p_pic;
627 }
628