]> git.sesse.net Git - vlc/blob - modules/packetizer/mpeg4video.c
* modules/packetizer/mpeg4video.c: rewrite using the block helper facility. Fixes...
[vlc] / modules / packetizer / mpeg4video.c
1 /*****************************************************************************
2  * mpeg4video.c: mpeg 4 video packetizer
3  *****************************************************************************
4  * Copyright (C) 2001-2006 the VideoLAN team
5  * $Id$
6  *
7  * Authors: Gildas Bazin <gbazin@videolan.org>
8  *          Laurent Aimar <fenrir@via.ecp.fr>
9  *          Eric Petit <titer@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., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, 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 #include <vlc/input.h>                  /* hmmm, just for INPUT_RATE_DEFAULT */
35
36 #include "vlc_bits.h"
37 #include "vlc_block_helper.h"
38
39 /*****************************************************************************
40  * Module descriptor
41  *****************************************************************************/
42 static int  Open ( vlc_object_t * );
43 static void Close( vlc_object_t * );
44
45 vlc_module_begin();
46     set_category( CAT_SOUT );
47     set_subcategory( SUBCAT_SOUT_PACKETIZER );
48     set_description( _("MPEG4 video packetizer") );
49     set_capability( "packetizer", 50 );
50     set_callbacks( Open, Close );
51 vlc_module_end();
52
53 /****************************************************************************
54  * Local prototypes
55  ****************************************************************************/
56 static block_t *Packetize( decoder_t *, block_t ** );
57
58 struct decoder_sys_t
59 {
60     /*
61      * Input properties
62      */
63     block_bytestream_t bytestream;
64     int i_state;
65     int i_offset;
66     uint8_t p_startcode[3];
67
68     /*
69      * Common properties
70      */
71     mtime_t i_interpolated_pts;
72     mtime_t i_interpolated_dts;
73     mtime_t i_last_ref_pts;
74     mtime_t i_last_time_ref;
75     mtime_t i_time_ref;
76     mtime_t i_last_time;
77     mtime_t i_last_timeincr;
78
79     unsigned int i_flags;
80
81     int         i_fps_num;
82     int         i_fps_den;
83     int         i_last_incr;
84     int         i_last_incr_diff;
85
86     vlc_bool_t  b_frame;
87
88     /* Current frame being built */
89     block_t    *p_frame;
90     block_t    **pp_last;
91 };
92
93 enum {
94     STATE_NOSYNC,
95     STATE_NEXT_SYNC
96 };
97
98 static block_t *ParseMPEGBlock( decoder_t *, block_t * );
99 static int ParseVOL( decoder_t *, es_format_t *, uint8_t *, int );
100 static int ParseVOP( decoder_t *, block_t * );
101 static int vlc_log2( unsigned int );
102
103 #define VIDEO_OBJECT_MASK                       0x01f
104 #define VIDEO_OBJECT_LAYER_MASK                 0x00f
105
106 #define VIDEO_OBJECT_START_CODE                 0x100
107 #define VIDEO_OBJECT_LAYER_START_CODE           0x120
108 #define VISUAL_OBJECT_SEQUENCE_START_CODE       0x1b0
109 #define VISUAL_OBJECT_SEQUENCE_END_CODE         0x1b1
110 #define USER_DATA_START_CODE                    0x1b2
111 #define GROUP_OF_VOP_START_CODE                 0x1b3
112 #define VIDEO_SESSION_ERROR_CODE                0x1b4
113 #define VISUAL_OBJECT_START_CODE                0x1b5
114 #define VOP_START_CODE                          0x1b6
115 #define FACE_OBJECT_START_CODE                  0x1ba
116 #define FACE_OBJECT_PLANE_START_CODE            0x1bb
117 #define MESH_OBJECT_START_CODE                  0x1bc
118 #define MESH_OBJECT_PLANE_START_CODE            0x1bd
119 #define STILL_TEXTURE_OBJECT_START_CODE         0x1be
120 #define TEXTURE_SPATIAL_LAYER_START_CODE        0x1bf
121 #define TEXTURE_SNR_LAYER_START_CODE            0x1c0
122
123 /*****************************************************************************
124  * Open: probe the packetizer and return score
125  *****************************************************************************/
126 static int Open( vlc_object_t *p_this )
127 {
128     decoder_t     *p_dec = (decoder_t*)p_this;
129     decoder_sys_t *p_sys;
130
131     switch( p_dec->fmt_in.i_codec )
132     {
133         case VLC_FOURCC( 'm', '4', 's', '2'):
134         case VLC_FOURCC( 'M', '4', 'S', '2'):
135         case VLC_FOURCC( 'm', 'p', '4', 's'):
136         case VLC_FOURCC( 'M', 'P', '4', 'S'):
137         case VLC_FOURCC( 'm', 'p', '4', 'v'):
138         case VLC_FOURCC( 'M', 'P', '4', 'V'):
139         case VLC_FOURCC( 'D', 'I', 'V', 'X'):
140         case VLC_FOURCC( 'd', 'i', 'v', 'x'):
141         case VLC_FOURCC( 'X', 'V', 'I', 'D'):
142         case VLC_FOURCC( 'X', 'v', 'i', 'D'):
143         case VLC_FOURCC( 'x', 'v', 'i', 'd'):
144         case VLC_FOURCC( 'D', 'X', '5', '0'):
145         case VLC_FOURCC( 'd', 'x', '5', '0'):
146         case VLC_FOURCC( 0x04, 0,   0,   0):
147         case VLC_FOURCC( '3', 'I', 'V', '2'):
148         case VLC_FOURCC( 'm', '4', 'c', 'c'):
149         case VLC_FOURCC( 'M', '4', 'C', 'C'):
150             break;
151
152         default:
153             return VLC_EGENERIC;
154     }
155
156     /* Allocate the memory needed to store the decoder's structure */
157     if( ( p_dec->p_sys = p_sys = malloc( sizeof(decoder_sys_t) ) ) == NULL )
158     {
159         msg_Err( p_dec, "out of memory" );
160         return VLC_EGENERIC;
161     }
162     memset( p_sys, 0, sizeof(decoder_sys_t) );
163
164     /* Misc init */
165     p_sys->i_state = STATE_NOSYNC;
166     p_sys->bytestream = block_BytestreamInit( p_dec );
167     p_sys->p_startcode[0] = 0;
168     p_sys->p_startcode[1] = 0;
169     p_sys->p_startcode[2] = 1;
170     p_sys->i_offset = 0;
171     p_sys->p_frame = NULL;
172     p_sys->pp_last = &p_sys->p_frame;
173
174     /* Setup properties */
175     es_format_Copy( &p_dec->fmt_out, &p_dec->fmt_in );
176     p_dec->fmt_out.i_codec = VLC_FOURCC( 'm', 'p', '4', 'v' );
177
178     if( p_dec->fmt_in.i_extra )
179     {
180         /* We have a vol */
181         p_dec->fmt_out.i_extra = p_dec->fmt_in.i_extra;
182         p_dec->fmt_out.p_extra = malloc( p_dec->fmt_in.i_extra );
183         memcpy( p_dec->fmt_out.p_extra, p_dec->fmt_in.p_extra,
184                 p_dec->fmt_in.i_extra );
185
186         msg_Dbg( p_dec, "opening with vol size: %d", p_dec->fmt_in.i_extra );
187         ParseVOL( p_dec, &p_dec->fmt_out,
188                   p_dec->fmt_out.p_extra, p_dec->fmt_out.i_extra );
189     }
190     else
191     {
192         /* No vol, we'll have to look for one later on */
193         p_dec->fmt_out.i_extra = 0;
194         p_dec->fmt_out.p_extra = 0;
195     }
196
197     /* Set callback */
198     p_dec->pf_packetize = Packetize;
199
200     return VLC_SUCCESS;
201 }
202
203 /*****************************************************************************
204  * Close: clean up the packetizer
205  *****************************************************************************/
206 static void Close( vlc_object_t *p_this )
207 {
208     decoder_t *p_dec = (decoder_t*)p_this;
209
210     block_BytestreamRelease( &p_dec->p_sys->bytestream );
211     if( p_dec->p_sys->p_frame ) block_ChainRelease( p_dec->p_sys->p_frame );
212     free( p_dec->p_sys );
213 }
214
215 /****************************************************************************
216  * Packetize: the whole thing
217  ****************************************************************************/
218 static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
219 {
220     decoder_sys_t *p_sys = p_dec->p_sys;
221     block_t       *p_pic;
222     mtime_t       i_pts, i_dts;
223
224     if( pp_block == NULL || *pp_block == NULL ) return NULL;
225
226     if( (*pp_block)->i_flags & BLOCK_FLAG_DISCONTINUITY )
227     {
228         p_sys->i_state = STATE_NOSYNC;
229         if( p_sys->p_frame ) block_ChainRelease( p_sys->p_frame );
230         p_sys->p_frame = NULL;
231         p_sys->pp_last = &p_sys->p_frame;
232         block_Release( *pp_block );
233         return NULL;
234     }
235
236     block_BytestreamPush( &p_sys->bytestream, *pp_block );
237
238     while( 1 )
239     {
240         switch( p_sys->i_state )
241         {
242
243         case STATE_NOSYNC:
244             if( block_FindStartcodeFromOffset( &p_sys->bytestream,
245                     &p_sys->i_offset, p_sys->p_startcode, 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             /* TODO: If p_block == NULL, flush the buffer without checking the
267              * next sync word */
268
269             /* Find the next startcode */
270             if( block_FindStartcodeFromOffset( &p_sys->bytestream,
271                     &p_sys->i_offset, p_sys->p_startcode, 3 ) != VLC_SUCCESS )
272             {
273                 /* Need more data */
274                 return NULL;
275             }
276
277             /* Get the new fragment and set the pts/dts */
278             p_pic = block_New( p_dec, p_sys->i_offset );
279             block_BytestreamFlush( &p_sys->bytestream );
280             p_pic->i_pts = i_pts = p_sys->bytestream.p_block->i_pts;
281             p_pic->i_dts = i_dts = p_sys->bytestream.p_block->i_dts;
282             p_pic->i_rate = p_sys->bytestream.p_block->i_rate;
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             /* Get picture if any */
290             if( !( p_pic = ParseMPEGBlock( p_dec, p_pic ) ) )
291             {
292                 p_sys->i_state = STATE_NOSYNC;
293                 break;
294             }
295
296             /* don't reuse the same timestamps several times */
297             if( i_pts == p_sys->bytestream.p_block->i_pts &&
298                 i_dts == p_sys->bytestream.p_block->i_dts )
299             {
300                 p_sys->bytestream.p_block->i_pts = 0;
301                 p_sys->bytestream.p_block->i_dts = 0;
302             }
303
304             /* We've just started the stream, wait for the first PTS.
305              * We discard here so we can still get the sequence header. */
306             if( p_sys->i_interpolated_pts <= 0 &&
307                 p_sys->i_interpolated_dts <= 0 )
308             {
309                 msg_Dbg( p_dec, "need a starting pts/dts" );
310                 p_sys->i_state = STATE_NOSYNC;
311                 block_Release( p_pic );
312                 break;
313             }
314
315             /* When starting the stream we can have the first frame with
316              * a null DTS (i_interpolated_pts is initialized to 0) */
317             if( !p_pic->i_dts ) p_pic->i_dts = p_pic->i_pts;
318
319             /* So p_block doesn't get re-added several times */
320             *pp_block = block_BytestreamPop( &p_sys->bytestream );
321
322             p_sys->i_state = STATE_NOSYNC;
323
324             return p_pic;
325         }
326     }
327 }
328
329 /*****************************************************************************
330  * ParseMPEGBlock: Re-assemble fragments into a block containing a picture
331  *****************************************************************************/
332 static block_t *ParseMPEGBlock( decoder_t *p_dec, block_t *p_frag )
333 {
334     decoder_sys_t *p_sys = p_dec->p_sys;
335     block_t *p_pic = NULL;
336
337     if( p_frag->p_buffer[3] == 0xB0 || p_frag->p_buffer[3] == 0xB1 )
338     {
339         /* Remove VOS start/end code from the original stream */
340         block_Release( p_frag );
341         return NULL;
342     }
343     if( p_frag->p_buffer[3] >= 0x20 && p_frag->p_buffer[3] <= 0x2f )
344     {
345         /* Copy the complete VOL */
346         p_dec->fmt_out.i_extra = p_frag->i_buffer;
347         p_dec->fmt_out.p_extra =
348             realloc( p_dec->fmt_out.p_extra, p_dec->fmt_out.i_extra );
349         memcpy( p_dec->fmt_out.p_extra, p_frag->p_buffer, p_frag->i_buffer );
350         ParseVOL( p_dec, &p_dec->fmt_out,
351                   p_dec->fmt_out.p_extra, p_dec->fmt_out.i_extra );
352
353         /* Remove from the original stream */
354         block_Release( p_frag );
355         return NULL;
356     }
357     else
358     {
359         if( !p_dec->fmt_out.i_extra )
360         {
361             msg_Warn( p_dec, "waiting for VOL" );
362             block_Release( p_frag );
363             return NULL;
364         }
365
366         /* Append the block */
367         block_ChainLastAppend( &p_sys->pp_last, p_frag );
368     }
369
370     if( p_frag->p_buffer[3] == 0xb6 &&
371         ParseVOP( p_dec, p_frag ) == VLC_SUCCESS )
372     {
373         /* We are dealing with a VOP */
374         p_pic = block_ChainGather( p_sys->p_frame );
375         p_pic->i_pts = p_sys->i_interpolated_pts;
376         p_pic->i_dts = p_sys->i_interpolated_dts;
377
378         /* Reset context */
379         p_sys->p_frame = NULL;
380         p_sys->pp_last = &p_sys->p_frame;
381     }
382
383     return p_pic;
384 }
385
386 /* ParseVOL:
387  *  TODO:
388  *      - support aspect ratio
389  */
390 static int ParseVOL( decoder_t *p_dec, es_format_t *fmt,
391                      uint8_t *p_vol, int i_vol )
392 {
393     decoder_sys_t *p_sys = p_dec->p_sys;
394     int i_vo_type, i_vo_ver_id, i_ar, i_shape;
395     bs_t s;
396
397     for( ;; )
398     {
399         if( p_vol[0] == 0x00 && p_vol[1] == 0x00 && p_vol[2] == 0x01 &&
400             p_vol[3] >= 0x20 && p_vol[3] <= 0x2f ) break;
401
402         p_vol++; i_vol--;
403         if( i_vol <= 4 ) return VLC_EGENERIC;
404     }
405
406     bs_init( &s, &p_vol[4], i_vol - 4 );
407
408     bs_skip( &s, 1 );   /* random access */
409     i_vo_type = bs_read( &s, 8 );
410     if( bs_read1( &s ) )
411     {
412         i_vo_ver_id = bs_read( &s, 4 );
413         bs_skip( &s, 3 );
414     }
415     else
416     {
417         i_vo_ver_id = 1;
418     }
419     i_ar = bs_read( &s, 4 );
420     if( i_ar == 0xf )
421     {
422         int i_ar_width, i_ar_height;
423
424         i_ar_width = bs_read( &s, 8 );
425         i_ar_height= bs_read( &s, 8 );
426     }
427     if( bs_read1( &s ) )
428     {
429         int i_chroma_format;
430         int i_low_delay;
431
432         /* vol control parameter */
433         i_chroma_format = bs_read( &s, 2 );
434         i_low_delay = bs_read1( &s );
435
436         if( bs_read1( &s ) )
437         {
438             bs_skip( &s, 16 );
439             bs_skip( &s, 16 );
440             bs_skip( &s, 16 );
441             bs_skip( &s, 3 );
442             bs_skip( &s, 11 );
443             bs_skip( &s, 1 );
444             bs_skip( &s, 16 );
445         }
446     }
447     /* shape 0->RECT, 1->BIN, 2->BIN_ONLY, 3->GRAY */
448     i_shape = bs_read( &s, 2 );
449     if( i_shape == 3 && i_vo_ver_id != 1 )
450     {
451         bs_skip( &s, 4 );
452     }
453
454     if( !bs_read1( &s ) ) return VLC_EGENERIC; /* Marker */
455
456     p_sys->i_fps_num = bs_read( &s, 16 ); /* Time increment resolution*/
457     if( !p_sys->i_fps_num ) p_sys->i_fps_num = 1;
458
459     if( !bs_read1( &s ) ) return VLC_EGENERIC; /* Marker */
460
461     if( bs_read1( &s ) )
462     {
463         int i_time_increment_bits = vlc_log2( p_sys->i_fps_num - 1 ) + 1;
464
465         if( i_time_increment_bits < 1 ) i_time_increment_bits = 1;
466
467         p_sys->i_fps_den = bs_read( &s, i_time_increment_bits );
468     }
469     if( i_shape == 0 )
470     {
471         bs_skip( &s, 1 );
472         fmt->video.i_width = bs_read( &s, 13 );
473         bs_skip( &s, 1 );
474         fmt->video.i_height= bs_read( &s, 13 );
475         bs_skip( &s, 1 );
476     }
477
478     return VLC_SUCCESS;
479 }
480
481 static int ParseVOP( decoder_t *p_dec, block_t *p_vop )
482 {
483     decoder_sys_t *p_sys = p_dec->p_sys;
484     int64_t i_time_increment, i_time_ref;
485     int i_modulo_time_base = 0, i_time_increment_bits;
486     bs_t s;
487
488     bs_init( &s, &p_vop->p_buffer[4], p_vop->i_buffer - 4 );
489
490     switch( bs_read( &s, 2 ) )
491     {
492     case 0:
493         p_sys->i_flags = BLOCK_FLAG_TYPE_I;
494         break;
495     case 1:
496         p_sys->i_flags = BLOCK_FLAG_TYPE_P;
497         break;
498     case 2:
499         p_sys->i_flags = BLOCK_FLAG_TYPE_B;
500         p_sys->b_frame = VLC_TRUE;
501         break;
502     case 3: /* gni ? */
503         p_sys->i_flags = BLOCK_FLAG_TYPE_PB;
504         break;
505     }
506
507     while( bs_read( &s, 1 ) ) i_modulo_time_base++;
508     if( !bs_read1( &s ) ) return VLC_EGENERIC; /* Marker */
509
510     /* VOP time increment */
511     i_time_increment_bits = vlc_log2(p_dec->p_sys->i_fps_num - 1) + 1;
512     if( i_time_increment_bits < 1 ) i_time_increment_bits = 1;
513     i_time_increment = bs_read( &s, i_time_increment_bits );
514
515     /* Interpolate PTS/DTS */
516     if( !(p_sys->i_flags & BLOCK_FLAG_TYPE_B) )
517     {
518         p_sys->i_last_time_ref = p_sys->i_time_ref;
519         p_sys->i_time_ref +=
520             (i_modulo_time_base * p_dec->p_sys->i_fps_num);
521         i_time_ref = p_sys->i_time_ref;
522     }
523     else
524     {
525         i_time_ref = p_sys->i_last_time_ref +
526             (i_modulo_time_base * p_dec->p_sys->i_fps_num);
527     }
528
529 #if 0
530     msg_Err( p_dec, "interp pts/dts (%lli,%lli), pts/dts (%lli,%lli)",
531              p_sys->i_interpolated_pts, p_sys->i_interpolated_dts,
532              p_vop->i_pts, p_vop->i_dts );
533 #endif
534
535     if( p_dec->p_sys->i_fps_num < 5 && /* Work-around buggy streams */
536         p_dec->fmt_in.video.i_frame_rate > 0 &&
537         p_dec->fmt_in.video.i_frame_rate_base > 0 )
538     {
539         p_sys->i_interpolated_pts += I64C(1000000) *
540         p_dec->fmt_in.video.i_frame_rate_base *
541         p_vop->i_rate / INPUT_RATE_DEFAULT /
542         p_dec->fmt_in.video.i_frame_rate;
543     }
544     else if( p_dec->p_sys->i_fps_num )
545         p_sys->i_interpolated_pts +=
546             ( I64C(1000000) * (i_time_ref + i_time_increment -
547               p_sys->i_last_time - p_sys->i_last_timeincr) *
548               p_vop->i_rate / INPUT_RATE_DEFAULT /
549               p_dec->p_sys->i_fps_num );
550
551     p_sys->i_last_time = i_time_ref;
552     p_sys->i_last_timeincr = i_time_increment;
553
554     /* Correct interpolated dts when we receive a new pts/dts */
555     if( p_vop->i_pts > 0 )
556         p_sys->i_interpolated_pts = p_vop->i_pts;
557     if( p_vop->i_dts > 0 )
558         p_sys->i_interpolated_dts = p_vop->i_dts;
559
560     if( (p_sys->i_flags & BLOCK_FLAG_TYPE_B) || !p_sys->b_frame )
561     {
562         /* Trivial case (DTS == PTS) */
563
564         p_sys->i_interpolated_dts = p_sys->i_interpolated_pts;
565
566         if( p_vop->i_pts > 0 )
567             p_sys->i_interpolated_dts = p_vop->i_pts;
568         if( p_vop->i_dts > 0 )
569             p_sys->i_interpolated_dts = p_vop->i_dts;
570
571         p_sys->i_interpolated_pts = p_sys->i_interpolated_dts;
572     }
573     else
574     {
575         if( p_sys->i_last_ref_pts > 0 )
576             p_sys->i_interpolated_dts = p_sys->i_last_ref_pts;
577
578         p_sys->i_last_ref_pts = p_sys->i_interpolated_pts;
579     }
580
581     return VLC_SUCCESS;
582 }
583
584 /* look at ffmpeg av_log2 ;) */
585 static int vlc_log2( unsigned int v )
586 {
587     int n = 0;
588     static const int vlc_log2_table[16] =
589     {
590         0,0,1,1,2,2,2,2, 3,3,3,3,3,3,3,3
591     };
592
593     if( v&0xffff0000 )
594     {
595         v >>= 16;
596         n += 16;
597     }
598     if( v&0xff00 )
599     {
600         v >>= 8;
601         n += 8;
602     }
603     if( v&0xf0 )
604     {
605         v >>= 4;
606         n += 4;
607     }
608     n += vlc_log2_table[v];
609
610     return n;
611 }