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