various security fixes and precautionary checks
[libav.git] / libavcodec / mpegvideo.h
1 /*
2 * Generic DCT based hybrid video encoder
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
4 * Copyright (c) 2002-2004 Michael Niedermayer
5 *
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
10 *
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19 */
20
21 /**
22 * @file mpegvideo.h
23 * mpegvideo header.
24 */
25
26 #ifndef AVCODEC_MPEGVIDEO_H
27 #define AVCODEC_MPEGVIDEO_H
28
29 #include "dsputil.h"
30 #include "bitstream.h"
31
32 #define FRAME_SKIPED 100 ///< return value for header parsers if frame is not coded
33
34 enum OutputFormat {
35 FMT_MPEG1,
36 FMT_H261,
37 FMT_H263,
38 FMT_MJPEG,
39 FMT_H264,
40 };
41
42 #define EDGE_WIDTH 16
43
44 #define MPEG_BUF_SIZE (16 * 1024)
45
46 #define QMAT_SHIFT_MMX 16
47 #define QMAT_SHIFT 22
48
49 #define MAX_FCODE 7
50 #define MAX_MV 2048
51
52 #define MAX_THREADS 8
53
54 #define MAX_PICTURE_COUNT 32
55
56 #define ME_MAP_SIZE 64
57 #define ME_MAP_SHIFT 3
58 #define ME_MAP_MV_BITS 11
59
60 /* run length table */
61 #define MAX_RUN 64
62 #define MAX_LEVEL 64
63
64 #define I_TYPE FF_I_TYPE ///< Intra
65 #define P_TYPE FF_P_TYPE ///< Predicted
66 #define B_TYPE FF_B_TYPE ///< Bi-dir predicted
67 #define S_TYPE FF_S_TYPE ///< S(GMC)-VOP MPEG4
68 #define SI_TYPE FF_SI_TYPE ///< Switching Intra
69 #define SP_TYPE FF_SP_TYPE ///< Switching Predicted
70
71 #define MAX_MB_BYTES (30*16*16*3/8 + 120)
72
73 typedef struct Predictor{
74 double coeff;
75 double count;
76 double decay;
77 } Predictor;
78
79 typedef struct RateControlEntry{
80 int pict_type;
81 float qscale;
82 int mv_bits;
83 int i_tex_bits;
84 int p_tex_bits;
85 int misc_bits;
86 uint64_t expected_bits;
87 int new_pict_type;
88 float new_qscale;
89 int mc_mb_var_sum;
90 int mb_var_sum;
91 int i_count;
92 int f_code;
93 int b_code;
94 }RateControlEntry;
95
96 /**
97 * rate control context.
98 */
99 typedef struct RateControlContext{
100 FILE *stats_file;
101 int num_entries; ///< number of RateControlEntries
102 RateControlEntry *entry;
103 double buffer_index; ///< amount of bits in the video/audio buffer
104 Predictor pred[5];
105 double short_term_qsum; ///< sum of recent qscales
106 double short_term_qcount; ///< count of recent qscales
107 double pass1_rc_eq_output_sum;///< sum of the output of the rc equation, this is used for normalization
108 double pass1_wanted_bits; ///< bits which should have been outputed by the pass1 code (including complexity init)
109 double last_qscale;
110 double last_qscale_for[5]; ///< last qscale for a specific pict type, used for max_diff & ipb factor stuff
111 int last_mc_mb_var_sum;
112 int last_mb_var_sum;
113 uint64_t i_cplx_sum[5];
114 uint64_t p_cplx_sum[5];
115 uint64_t mv_bits_sum[5];
116 uint64_t qscale_sum[5];
117 int frame_count[5];
118 int last_non_b_pict_type;
119 }RateControlContext;
120
121 /**
122 * Scantable.
123 */
124 typedef struct ScanTable{
125 const uint8_t *scantable;
126 uint8_t permutated[64];
127 uint8_t raster_end[64];
128 #ifdef ARCH_POWERPC
129 /** Used by dct_quantise_alitvec to find last-non-zero */
130 uint8_t __align8 inverse[64];
131 #endif
132 } ScanTable;
133
134 /**
135 * Picture.
136 */
137 typedef struct Picture{
138 FF_COMMON_FRAME
139
140 /**
141 * halfpel luma planes.
142 */
143 uint8_t *interpolated[3];
144 int16_t (*motion_val_base[2])[2];
145 uint32_t *mb_type_base;
146 #define MB_TYPE_INTRA MB_TYPE_INTRA4x4 //default mb_type if theres just one type
147 #define IS_INTRA4x4(a) ((a)&MB_TYPE_INTRA4x4)
148 #define IS_INTRA16x16(a) ((a)&MB_TYPE_INTRA16x16)
149 #define IS_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
150 #define IS_INTRA(a) ((a)&7)
151 #define IS_INTER(a) ((a)&(MB_TYPE_16x16|MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8))
152 #define IS_SKIP(a) ((a)&MB_TYPE_SKIP)
153 #define IS_INTRA_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
154 #define IS_INTERLACED(a) ((a)&MB_TYPE_INTERLACED)
155 #define IS_DIRECT(a) ((a)&MB_TYPE_DIRECT2)
156 #define IS_GMC(a) ((a)&MB_TYPE_GMC)
157 #define IS_16X16(a) ((a)&MB_TYPE_16x16)
158 #define IS_16X8(a) ((a)&MB_TYPE_16x8)
159 #define IS_8X16(a) ((a)&MB_TYPE_8x16)
160 #define IS_8X8(a) ((a)&MB_TYPE_8x8)
161 #define IS_SUB_8X8(a) ((a)&MB_TYPE_16x16) //note reused
162 #define IS_SUB_8X4(a) ((a)&MB_TYPE_16x8) //note reused
163 #define IS_SUB_4X8(a) ((a)&MB_TYPE_8x16) //note reused
164 #define IS_SUB_4X4(a) ((a)&MB_TYPE_8x8) //note reused
165 #define IS_ACPRED(a) ((a)&MB_TYPE_ACPRED)
166 #define IS_QUANT(a) ((a)&MB_TYPE_QUANT)
167 #define IS_DIR(a, part, list) ((a) & (MB_TYPE_P0L0<<((part)+2*(list))))
168 #define USES_LIST(a, list) ((a) & ((MB_TYPE_P0L0|MB_TYPE_P1L0)<<(2*(list)))) ///< does this mb use listX, note doesnt work if subMBs
169 #define HAS_CBP(a) ((a)&MB_TYPE_CBP)
170
171 int field_poc[2]; ///< h264 top/bottom POC
172 int poc; ///< h264 frame POC
173 int frame_num; ///< h264 frame_num
174 int pic_id; ///< h264 pic_num or long_term_pic_idx
175 int long_ref; ///< 1->long term reference 0->short term reference
176
177 int mb_var_sum; ///< sum of MB variance for current frame
178 int mc_mb_var_sum; ///< motion compensated MB variance for current frame
179 uint16_t *mb_var; ///< Table for MB variances
180 uint16_t *mc_mb_var; ///< Table for motion compensated MB variances
181 uint8_t *mb_mean; ///< Table for MB luminance
182 int32_t *mb_cmp_score; ///< Table for MB cmp scores, for mb decission FIXME remove
183 int b_frame_score; /* */
184 } Picture;
185
186 typedef struct ParseContext{
187 uint8_t *buffer;
188 int index;
189 int last_index;
190 int buffer_size;
191 uint32_t state; ///< contains the last few bytes in MSB order
192 int frame_start_found;
193 int overread; ///< the number of bytes which where irreversibly read from the next frame
194 int overread_index; ///< the index into ParseContext.buffer of the overreaded bytes
195 } ParseContext;
196
197 struct MpegEncContext;
198
199 /**
200 * Motion estimation context.
201 */
202 typedef struct MotionEstContext{
203 AVCodecContext *avctx;
204 int skip; ///< set if ME is skiped for the current MB
205 int co_located_mv[4][2]; ///< mv from last p frame for direct mode ME
206 int direct_basis_mv[4][2];
207 uint8_t *scratchpad; ///< data area for the me algo, so that the ME doesnt need to malloc/free
208 uint8_t *best_mb;
209 uint8_t *temp_mb[2];
210 uint8_t *temp;
211 int best_bits;
212 uint32_t *map; ///< map to avoid duplicate evaluations
213 uint32_t *score_map; ///< map to store the scores
214 int map_generation;
215 int pre_penalty_factor;
216 int penalty_factor;
217 int sub_penalty_factor;
218 int mb_penalty_factor;
219 int flags;
220 int sub_flags;
221 int mb_flags;
222 int pre_pass; ///< = 1 for the pre pass
223 int dia_size;
224 int xmin;
225 int xmax;
226 int ymin;
227 int ymax;
228 int pred_x;
229 int pred_y;
230 uint8_t *src[4][4];
231 uint8_t *ref[4][4];
232 int stride;
233 int uvstride;
234 /* temp variables for picture complexity calculation */
235 int mc_mb_var_sum_temp;
236 int mb_var_sum_temp;
237 int scene_change_score;
238 /* cmp, chroma_cmp;*/
239 op_pixels_func (*hpel_put)[4];
240 op_pixels_func (*hpel_avg)[4];
241 qpel_mc_func (*qpel_put)[16];
242 qpel_mc_func (*qpel_avg)[16];
243 uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
244 uint8_t *current_mv_penalty;
245 int (*sub_motion_search)(struct MpegEncContext * s,
246 int *mx_ptr, int *my_ptr, int dmin,
247 int src_index, int ref_index,
248 int size, int h);
249 }MotionEstContext;
250
251 /**
252 * MpegEncContext.
253 */
254 typedef struct MpegEncContext {
255 struct AVCodecContext *avctx;
256 /* the following parameters must be initialized before encoding */
257 int width, height;///< picture size. must be a multiple of 16
258 int gop_size;
259 int intra_only; ///< if true, only intra pictures are generated
260 int bit_rate; ///< wanted bit rate
261 enum OutputFormat out_format; ///< output format
262 int h263_pred; ///< use mpeg4/h263 ac/dc predictions
263
264 /* the following codec id fields are deprecated in favor of codec_id */
265 int h263_plus; ///< h263 plus headers
266 int h263_msmpeg4; ///< generate MSMPEG4 compatible stream (deprecated, use msmpeg4_version instead)
267 int h263_flv; ///< use flv h263 header
268
269 enum CodecID codec_id; /* see CODEC_ID_xxx */
270 int fixed_qscale; ///< fixed qscale if non zero
271 int encoding; ///< true if we are encoding (vs decoding)
272 int flags; ///< AVCodecContext.flags (HQ, MV4, ...)
273 int flags2; ///< AVCodecContext.flags2
274 int max_b_frames; ///< max number of b-frames for encoding
275 int luma_elim_threshold;
276 int chroma_elim_threshold;
277 int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
278 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
279 /* the following fields are managed internally by the encoder */
280
281 /** bit output */
282 PutBitContext pb;
283
284 /* sequence parameters */
285 int context_initialized;
286 int input_picture_number; ///< used to set pic->display_picture_number, shouldnt be used for/by anything else
287 int coded_picture_number; ///< used to set pic->coded_picture_number, shouldnt be used for/by anything else
288 int picture_number; //FIXME remove, unclear definition
289 int picture_in_gop_number; ///< 0-> first pic in gop, ...
290 int b_frames_since_non_b; ///< used for encoding, relative to not yet reordered input
291 int64_t user_specified_pts;///< last non zero pts from AVFrame which was passed into avcodec_encode_video()
292 int mb_width, mb_height; ///< number of MBs horizontally & vertically
293 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressng of left & top MBs withoutt sig11
294 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressng
295 int b4_stride; ///< 4*mb_width+1 used for some 4x4 block arrays to allow simple addressng
296 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replicateion)
297 int mb_num; ///< number of MBs of a picture
298 int linesize; ///< line size, in bytes, may be different from width
299 int uvlinesize; ///< line size, for chroma in bytes, may be different from width
300 Picture *picture; ///< main picture buffer
301 Picture **input_picture; ///< next pictures on display order for encoding
302 Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
303
304 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
305 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
306 struct MpegEncContext *thread_context[MAX_THREADS];
307
308 /**
309 * copy of the previous picture structure.
310 * note, linesize & data, might not match the previous picture (for field pictures)
311 */
312 Picture last_picture;
313
314 /**
315 * copy of the next picture structure.
316 * note, linesize & data, might not match the next picture (for field pictures)
317 */
318 Picture next_picture;
319
320 /**
321 * copy of the source picture structure for encoding.
322 * note, linesize & data, might not match the source picture (for field pictures)
323 */
324 Picture new_picture;
325
326 /**
327 * copy of the current picture structure.
328 * note, linesize & data, might not match the current picture (for field pictures)
329 */
330 Picture current_picture; ///< buffer to store the decompressed current picture
331
332 Picture *last_picture_ptr; ///< pointer to the previous picture.
333 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
334 Picture *current_picture_ptr; ///< pointer to the current picture
335 uint8_t *visualization_buffer[3]; //< temporary buffer vor MV visualization
336 int last_dc[3]; ///< last DC values for MPEG1
337 int16_t *dc_val_base;
338 int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
339 int16_t dc_cache[4*5];
340 int y_dc_scale, c_dc_scale;
341 uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
342 uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
343 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
344 uint8_t *coded_block_base;
345 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
346 int16_t (*ac_val_base)[16];
347 int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
348 int ac_pred;
349 uint8_t *prev_pict_types; ///< previous picture types in bitstream order, used for mb skip
350 #define PREV_PICT_TYPES_BUFFER_SIZE 256
351 int mb_skiped; ///< MUST BE SET only during DECODING
352 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
353 and used for b-frame encoding & decoding (contains skip table of next P Frame) */
354 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
355 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
356 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
357 uint8_t *allocated_edge_emu_buffer;
358 uint8_t *edge_emu_buffer; ///< points into the middle of allocated_edge_emu_buffer
359 uint8_t *rd_scratchpad; ///< scartchpad for rate distortion mb decission
360 uint8_t *obmc_scratchpad;
361 uint8_t *b_scratchpad; ///< scratchpad used for writing into write only buffers
362
363 int qscale; ///< QP
364 int chroma_qscale; ///< chroma QP
365 int lambda; ///< lagrange multipler used in rate distortion
366 int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
367 int *lambda_table;
368 int adaptive_quant; ///< use adaptive quantization
369 int dquant; ///< qscale difference to prev qscale
370 int pict_type; ///< I_TYPE, P_TYPE, B_TYPE, ...
371 int last_pict_type; //FIXME removes
372 int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
373 int dropable;
374 int frame_rate_index;
375
376 /* motion compensation */
377 int unrestricted_mv; ///< mv can point outside of the coded picture
378 int h263_long_vectors; ///< use horrible h263v1 long vector mode
379 int decode; ///< if 0 then decoding will be skiped (for encoding b frames for example)
380
381 DSPContext dsp; ///< pointers for accelerated dsp fucntions
382 int f_code; ///< forward MV resolution
383 int b_code; ///< backward MV resolution for B Frames (mpeg4)
384 int16_t (*p_mv_table_base)[2];
385 int16_t (*b_forw_mv_table_base)[2];
386 int16_t (*b_back_mv_table_base)[2];
387 int16_t (*b_bidir_forw_mv_table_base)[2];
388 int16_t (*b_bidir_back_mv_table_base)[2];
389 int16_t (*b_direct_mv_table_base)[2];
390 int16_t (*p_field_mv_table_base[2][2])[2];
391 int16_t (*b_field_mv_table_base[2][2][2])[2];
392 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
393 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
394 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
395 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
396 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
397 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
398 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
399 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
400 uint8_t (*p_field_select_table[2]);
401 uint8_t (*b_field_select_table[2][2]);
402 int me_method; ///< ME algorithm
403 int mv_dir;
404 #define MV_DIR_BACKWARD 1
405 #define MV_DIR_FORWARD 2
406 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
407 int mv_type;
408 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
409 #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
410 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
411 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
412 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
413 /**motion vectors for a macroblock
414 first coordinate : 0 = forward 1 = backward
415 second " : depend on type
416 third " : 0 = x, 1 = y
417 */
418 int mv[2][4][2];
419 int field_select[2][2];
420 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
421 uint8_t *fcode_tab; ///< smallest fcode needed for each MV
422
423 MotionEstContext me;
424
425 int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
426 for b-frames rounding mode is allways 0 */
427
428 int hurry_up; /**< when set to 1 during decoding, b frames will be skiped
429 when set to 2 idct/dequant will be skipped too */
430
431 /* macroblock layer */
432 int mb_x, mb_y;
433 int mb_skip_run;
434 int mb_intra;
435 uint16_t *mb_type; ///< Table for candidate MB types for encoding
436 #define CANDIDATE_MB_TYPE_INTRA 0x01
437 #define CANDIDATE_MB_TYPE_INTER 0x02
438 #define CANDIDATE_MB_TYPE_INTER4V 0x04
439 #define CANDIDATE_MB_TYPE_SKIPED 0x08
440 //#define MB_TYPE_GMC 0x10
441
442 #define CANDIDATE_MB_TYPE_DIRECT 0x10
443 #define CANDIDATE_MB_TYPE_FORWARD 0x20
444 #define CANDIDATE_MB_TYPE_BACKWARD 0x40
445 #define CANDIDATE_MB_TYPE_BIDIR 0x80
446
447 #define CANDIDATE_MB_TYPE_INTER_I 0x100
448 #define CANDIDATE_MB_TYPE_FORWARD_I 0x200
449 #define CANDIDATE_MB_TYPE_BACKWARD_I 0x400
450 #define CANDIDATE_MB_TYPE_BIDIR_I 0x800
451
452 int block_index[6]; ///< index to current MB in block based arrays with edges
453 int block_wrap[6];
454 uint8_t *dest[3];
455
456 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
457
458 /** matrix transmitted in the bitstream */
459 uint16_t intra_matrix[64];
460 uint16_t chroma_intra_matrix[64];
461 uint16_t inter_matrix[64];
462 uint16_t chroma_inter_matrix[64];
463 #define QUANT_BIAS_SHIFT 8
464 int intra_quant_bias; ///< bias for the quantizer
465 int inter_quant_bias; ///< bias for the quantizer
466 int min_qcoeff; ///< minimum encodable coefficient
467 int max_qcoeff; ///< maximum encodable coefficient
468 int ac_esc_length; ///< num of bits needed to encode the longest esc
469 uint8_t *intra_ac_vlc_length;
470 uint8_t *intra_ac_vlc_last_length;
471 uint8_t *inter_ac_vlc_length;
472 uint8_t *inter_ac_vlc_last_length;
473 uint8_t *luma_dc_vlc_length;
474 uint8_t *chroma_dc_vlc_length;
475 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
476
477 int coded_score[6];
478
479 /** precomputed matrix (combine qscale and DCT renorm) */
480 int (*q_intra_matrix)[64];
481 int (*q_inter_matrix)[64];
482 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
483 uint16_t (*q_intra_matrix16)[2][64];
484 uint16_t (*q_inter_matrix16)[2][64];
485 int block_last_index[12]; ///< last non zero coefficient in block
486 /* scantables */
487 ScanTable __align8 intra_scantable;
488 ScanTable intra_h_scantable;
489 ScanTable intra_v_scantable;
490 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
491
492 /* noise reduction */
493 int (*dct_error_sum)[64];
494 int dct_count[2];
495 uint16_t (*dct_offset)[64];
496
497 void *opaque; ///< private data for the user
498
499 /* bit rate control */
500 int64_t wanted_bits;
501 int64_t total_bits;
502 int frame_bits; ///< bits used for the current frame
503 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
504
505 /* statistics, used for 2-pass encoding */
506 int mv_bits;
507 int header_bits;
508 int i_tex_bits;
509 int p_tex_bits;
510 int i_count;
511 int f_count;
512 int b_count;
513 int skip_count;
514 int misc_bits; ///< cbp, mb_type
515 int last_bits; ///< temp var used for calculating the above vars
516
517 /* error concealment / resync */
518 int error_count;
519 uint8_t *error_status_table; ///< table of the error status of each MB
520 #define VP_START 1 ///< current MB is the first after a resync marker
521 #define AC_ERROR 2
522 #define DC_ERROR 4
523 #define MV_ERROR 8
524 #define AC_END 16
525 #define DC_END 32
526 #define MV_END 64
527 //FIXME some prefix?
528
529 int resync_mb_x; ///< x position of last resync marker
530 int resync_mb_y; ///< y position of last resync marker
531 GetBitContext last_resync_gb; ///< used to search for the next resync marker
532 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
533 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
534 int error_resilience;
535
536 ParseContext parse_context;
537
538 /* H.263 specific */
539 int gob_index;
540 int obmc; ///< overlapped block motion compensation
541
542 /* H.263+ specific */
543 int umvplus; ///< == H263+ && unrestricted_mv
544 int h263_aic; ///< Advanded INTRA Coding (AIC)
545 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
546 int h263_slice_structured;
547 int alt_inter_vlc; ///< alternative inter vlc
548 int modified_quant;
549 int loop_filter;
550 int custom_pcf;
551
552 /* mpeg4 specific */
553 int time_increment_resolution;
554 int time_increment_bits; ///< number of bits to represent the fractional part of time
555 int last_time_base;
556 int time_base; ///< time in seconds of last I,P,S Frame
557 int64_t time; ///< time of current frame
558 int64_t last_non_b_time;
559 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
560 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
561 uint16_t pp_field_time;
562 uint16_t pb_field_time; ///< like above, just for interlaced
563 int shape;
564 int vol_sprite_usage;
565 int sprite_width;
566 int sprite_height;
567 int sprite_left;
568 int sprite_top;
569 int sprite_brightness_change;
570 int num_sprite_warping_points;
571 int real_sprite_warping_points;
572 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
573 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
574 int sprite_shift[2]; ///< sprite shift [isChroma]
575 int mcsel;
576 int quant_precision;
577 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
578 int scalability;
579 int hierachy_type;
580 int enhancement_type;
581 int new_pred;
582 int reduced_res_vop;
583 int aspect_ratio_info; //FIXME remove
584 int sprite_warping_accuracy;
585 int low_latency_sprite;
586 int data_partitioning; ///< data partitioning flag from header
587 int partitioned_frame; ///< is current frame partitioned
588 int rvlc; ///< reversible vlc
589 int resync_marker; ///< could this stream contain resync markers
590 int low_delay; ///< no reordering needed / has no b-frames
591 int vo_type;
592 int vol_control_parameters; ///< does the stream contain the low_delay flag, used to workaround buggy encoders
593 int intra_dc_threshold; ///< QP above whch the ac VLC should be used for intra dc
594 PutBitContext tex_pb; ///< used for data partitioned VOPs
595 PutBitContext pb2; ///< used for data partitioned VOPs
596 int mpeg_quant;
597 int t_frame; ///< time distance of first I -> B, used for interlaced b frames
598 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
599
600 /* divx specific, used to workaround (many) bugs in divx5 */
601 int divx_version;
602 int divx_build;
603 int divx_packed;
604 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
605 int bitstream_buffer_size;
606 int allocated_bitstream_buffer_size;
607
608 int xvid_build;
609
610 /* lavc specific stuff, used to workaround bugs in libavcodec */
611 int ffmpeg_version;
612 int lavc_build;
613
614 /* RV10 specific */
615 int rv10_version; ///< RV10 version: 0 or 3
616 int rv10_first_dc_coded[3];
617
618 /* MJPEG specific */
619 struct MJpegContext *mjpeg_ctx;
620 int mjpeg_vsample[3]; ///< vertical sampling factors, default = {2, 1, 1}
621 int mjpeg_hsample[3]; ///< horizontal sampling factors, default = {2, 1, 1}
622 int mjpeg_write_tables; ///< do we want to have quantisation- and huffmantables in the jpeg file ?
623 int mjpeg_data_only_frames; ///< frames only with SOI, SOS and EOI markers
624
625 /* MSMPEG4 specific */
626 int mv_table_index;
627 int rl_table_index;
628 int rl_chroma_table_index;
629 int dc_table_index;
630 int use_skip_mb_code;
631 int slice_height; ///< in macroblocks
632 int first_slice_line; ///< used in mpeg4 too to handle resync markers
633 int flipflop_rounding;
634 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
635 int per_mb_rl_table;
636 int esc3_level_length;
637 int esc3_run_length;
638 /** [mb_intra][isChroma][level][run][last] */
639 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
640 int inter_intra_pred;
641 int mspel;
642
643 /* decompression specific */
644 GetBitContext gb;
645
646 /* Mpeg1 specific */
647 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
648 int last_mv_dir; ///< last mv_dir, used for b frame encoding
649 int broken_link; ///< no_output_of_prior_pics_flag
650 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
651
652 /* MPEG2 specific - I wish I had not to support this mess. */
653 int progressive_sequence;
654 int mpeg_f_code[2][2];
655 int picture_structure;
656 /* picture type */
657 #define PICT_TOP_FIELD 1
658 #define PICT_BOTTOM_FIELD 2
659 #define PICT_FRAME 3
660
661 int intra_dc_precision;
662 int frame_pred_frame_dct;
663 int top_field_first;
664 int concealment_motion_vectors;
665 int q_scale_type;
666 int intra_vlc_format;
667 int alternate_scan;
668 int repeat_first_field;
669 int chroma_420_type;
670 int chroma_format;
671 #define CHROMA_420 1
672 #define CHROMA_422 2
673 #define CHROMA_444 3
674 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
675 int chroma_y_shift;
676
677 int progressive_frame;
678 int full_pel[2];
679 int interlaced_dct;
680 int first_slice;
681 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
682
683 /* RTP specific */
684 int rtp_mode;
685
686 uint8_t *ptr_lastgob;
687 int swap_uv;//vcr2 codec is mpeg2 varint with UV swaped
688 short * pblocks[12];
689
690 DCTELEM (*block)[64]; ///< points to one of the following blocks
691 DCTELEM (*blocks)[6][64]; // for HQ mode we need to keep the best block
692 int (*decode_mb)(struct MpegEncContext *s, DCTELEM block[6][64]); // used by some codecs to avoid a switch()
693 #define SLICE_OK 0
694 #define SLICE_ERROR -1
695 #define SLICE_END -2 ///<end marker found
696 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
697
698 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
699 DCTELEM *block/*align 16*/, int n, int qscale);
700 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
701 DCTELEM *block/*align 16*/, int n, int qscale);
702 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
703 DCTELEM *block/*align 16*/, int n, int qscale);
704 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
705 DCTELEM *block/*align 16*/, int n, int qscale);
706 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
707 DCTELEM *block/*align 16*/, int n, int qscale);
708 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
709 DCTELEM *block/*align 16*/, int n, int qscale);
710 void (*dct_unquantize_h261_intra)(struct MpegEncContext *s,
711 DCTELEM *block/*align 16*/, int n, int qscale);
712 void (*dct_unquantize_h261_inter)(struct MpegEncContext *s,
713 DCTELEM *block/*align 16*/, int n, int qscale);
714 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
715 DCTELEM *block/*align 16*/, int n, int qscale);
716 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
717 DCTELEM *block/*align 16*/, int n, int qscale);
718 int (*dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
719 int (*fast_dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
720 void (*denoise_dct)(struct MpegEncContext *s, DCTELEM *block);
721 } MpegEncContext;
722
723
724 int DCT_common_init(MpegEncContext *s);
725 void MPV_decode_defaults(MpegEncContext *s);
726 int MPV_common_init(MpegEncContext *s);
727 void MPV_common_end(MpegEncContext *s);
728 void MPV_decode_mb(MpegEncContext *s, DCTELEM block[12][64]);
729 int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx);
730 void MPV_frame_end(MpegEncContext *s);
731 int MPV_encode_init(AVCodecContext *avctx);
732 int MPV_encode_end(AVCodecContext *avctx);
733 int MPV_encode_picture(AVCodecContext *avctx, unsigned char *buf, int buf_size, void *data);
734 #ifdef HAVE_MMX
735 void MPV_common_init_mmx(MpegEncContext *s);
736 #endif
737 #ifdef ARCH_ALPHA
738 void MPV_common_init_axp(MpegEncContext *s);
739 #endif
740 #ifdef HAVE_MLIB
741 void MPV_common_init_mlib(MpegEncContext *s);
742 #endif
743 #ifdef HAVE_MMI
744 void MPV_common_init_mmi(MpegEncContext *s);
745 #endif
746 #ifdef ARCH_ARMV4L
747 void MPV_common_init_armv4l(MpegEncContext *s);
748 #endif
749 #ifdef ARCH_POWERPC
750 void MPV_common_init_ppc(MpegEncContext *s);
751 #endif
752 extern void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
753 void ff_copy_bits(PutBitContext *pb, uint8_t *src, int length);
754 void ff_clean_intra_table_entries(MpegEncContext *s);
755 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
756 void ff_draw_horiz_band(MpegEncContext *s, int y, int h);
757 void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize, int block_w, int block_h,
758 int src_x, int src_y, int w, int h);
759 #define END_NOT_FOUND -100
760 int ff_combine_frame(ParseContext *pc, int next, uint8_t **buf, int *buf_size);
761 void ff_parse_close(AVCodecParserContext *s);
762 void ff_mpeg_flush(AVCodecContext *avctx);
763 void ff_print_debug_info(MpegEncContext *s, AVFrame *pict);
764 void ff_write_quant_matrix(PutBitContext *pb, int16_t *matrix);
765 int ff_find_unused_picture(MpegEncContext *s, int shared);
766 void ff_denoise_dct(MpegEncContext *s, DCTELEM *block);
767 void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
768
769 void ff_er_frame_start(MpegEncContext *s);
770 void ff_er_frame_end(MpegEncContext *s);
771 void ff_er_add_slice(MpegEncContext *s, int startx, int starty, int endx, int endy, int status);
772
773
774 extern enum PixelFormat ff_yuv420p_list[2];
775
776 void ff_init_block_index(MpegEncContext *s);
777
778 static inline void ff_update_block_index(MpegEncContext *s){
779 const int block_size= 8>>s->avctx->lowres;
780
781 s->block_index[0]+=2;
782 s->block_index[1]+=2;
783 s->block_index[2]+=2;
784 s->block_index[3]+=2;
785 s->block_index[4]++;
786 s->block_index[5]++;
787 s->dest[0]+= 2*block_size;
788 s->dest[1]+= block_size;
789 s->dest[2]+= block_size;
790 }
791
792 static inline int get_bits_diff(MpegEncContext *s){
793 const int bits= put_bits_count(&s->pb);
794 const int last= s->last_bits;
795
796 s->last_bits = bits;
797
798 return bits - last;
799 }
800
801 /* motion_est.c */
802 void ff_estimate_p_frame_motion(MpegEncContext * s,
803 int mb_x, int mb_y);
804 void ff_estimate_b_frame_motion(MpegEncContext * s,
805 int mb_x, int mb_y);
806 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
807 void ff_fix_long_p_mvs(MpegEncContext * s);
808 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
809 int16_t (*mv_table)[2], int f_code, int type, int truncate);
810 void ff_init_me(MpegEncContext *s);
811 int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
812 inline int ff_epzs_motion_search(MpegEncContext * s, int *mx_ptr, int *my_ptr,
813 int P[10][2], int src_index, int ref_index, int16_t (*last_mv)[2],
814 int ref_mv_scale, int size, int h);
815 int inline ff_get_mb_score(MpegEncContext * s, int mx, int my, int src_index,
816 int ref_index, int size, int h, int add_rate);
817
818 /* mpeg12.c */
819 extern const int16_t ff_mpeg1_default_intra_matrix[64];
820 extern const int16_t ff_mpeg1_default_non_intra_matrix[64];
821 extern uint8_t ff_mpeg1_dc_scale_table[128];
822
823 void mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
824 void mpeg1_encode_mb(MpegEncContext *s,
825 DCTELEM block[6][64],
826 int motion_x, int motion_y);
827 void ff_mpeg1_encode_init(MpegEncContext *s);
828 void ff_mpeg1_encode_slice_header(MpegEncContext *s);
829 void ff_mpeg1_clean_buffers(MpegEncContext *s);
830 int ff_mpeg1_find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size);
831
832
833 /** RLTable. */
834 typedef struct RLTable {
835 int n; ///< number of entries of table_vlc minus 1
836 int last; ///< number of values for last = 0
837 const uint16_t (*table_vlc)[2];
838 const int8_t *table_run;
839 const int8_t *table_level;
840 uint8_t *index_run[2]; ///< encoding only
841 int8_t *max_level[2]; ///< encoding & decoding
842 int8_t *max_run[2]; ///< encoding & decoding
843 VLC vlc; ///< decoding only deprected FIXME remove
844 RL_VLC_ELEM *rl_vlc[32]; ///< decoding only
845 } RLTable;
846
847 void init_rl(RLTable *rl, int use_static);
848 void init_vlc_rl(RLTable *rl, int use_static);
849
850 static inline int get_rl_index(const RLTable *rl, int last, int run, int level)
851 {
852 int index;
853 index = rl->index_run[last][run];
854 if (index >= rl->n)
855 return rl->n;
856 if (level > rl->max_level[last][run])
857 return rl->n;
858 return index + level - 1;
859 }
860
861 extern uint8_t ff_mpeg4_y_dc_scale_table[32];
862 extern uint8_t ff_mpeg4_c_dc_scale_table[32];
863 extern uint8_t ff_aic_dc_scale_table[32];
864 extern const int16_t ff_mpeg4_default_intra_matrix[64];
865 extern const int16_t ff_mpeg4_default_non_intra_matrix[64];
866 extern const uint8_t ff_h263_chroma_qscale_table[32];
867 extern const uint8_t ff_h263_loop_filter_strength[32];
868
869 /* h261.c */
870 void ff_h261_loop_filter(MpegEncContext *s);
871 void ff_h261_reorder_mb_index(MpegEncContext* s);
872 void ff_h261_encode_mb(MpegEncContext *s,
873 DCTELEM block[6][64],
874 int motion_x, int motion_y);
875 void ff_h261_encode_picture_header(MpegEncContext * s, int picture_number);
876 void ff_h261_encode_init(MpegEncContext *s);
877
878
879 /* h263.c, h263dec.c */
880 int ff_h263_decode_init(AVCodecContext *avctx);
881 int ff_h263_decode_frame(AVCodecContext *avctx,
882 void *data, int *data_size,
883 uint8_t *buf, int buf_size);
884 int ff_h263_decode_end(AVCodecContext *avctx);
885 void h263_encode_mb(MpegEncContext *s,
886 DCTELEM block[6][64],
887 int motion_x, int motion_y);
888 void mpeg4_encode_mb(MpegEncContext *s,
889 DCTELEM block[6][64],
890 int motion_x, int motion_y);
891 void h263_encode_picture_header(MpegEncContext *s, int picture_number);
892 void ff_flv_encode_picture_header(MpegEncContext *s, int picture_number);
893 void h263_encode_gob_header(MpegEncContext * s, int mb_line);
894 int16_t *h263_pred_motion(MpegEncContext * s, int block, int dir,
895 int *px, int *py);
896 void mpeg4_pred_ac(MpegEncContext * s, DCTELEM *block, int n,
897 int dir);
898 void ff_set_mpeg4_time(MpegEncContext * s, int picture_number);
899 void mpeg4_encode_picture_header(MpegEncContext *s, int picture_number);
900 void h263_encode_init(MpegEncContext *s);
901 void h263_decode_init_vlc(MpegEncContext *s);
902 int h263_decode_picture_header(MpegEncContext *s);
903 int ff_h263_decode_gob_header(MpegEncContext *s);
904 int ff_mpeg4_decode_picture_header(MpegEncContext * s, GetBitContext *gb);
905 void ff_h263_update_motion_val(MpegEncContext * s);
906 void ff_h263_loop_filter(MpegEncContext * s);
907 void ff_set_qscale(MpegEncContext * s, int qscale);
908 int ff_h263_decode_mba(MpegEncContext *s);
909 void ff_h263_encode_mba(MpegEncContext *s);
910
911 int intel_h263_decode_picture_header(MpegEncContext *s);
912 int flv_h263_decode_picture_header(MpegEncContext *s);
913 int ff_h263_decode_mb(MpegEncContext *s,
914 DCTELEM block[6][64]);
915 int ff_mpeg4_decode_mb(MpegEncContext *s,
916 DCTELEM block[6][64]);
917 int h263_get_picture_format(int width, int height);
918 void ff_mpeg4_encode_video_packet_header(MpegEncContext *s);
919 void ff_mpeg4_clean_buffers(MpegEncContext *s);
920 void ff_mpeg4_stuffing(PutBitContext * pbc);
921 void ff_mpeg4_init_partitions(MpegEncContext *s);
922 void ff_mpeg4_merge_partitions(MpegEncContext *s);
923 void ff_clean_mpeg4_qscales(MpegEncContext *s);
924 void ff_clean_h263_qscales(MpegEncContext *s);
925 int ff_mpeg4_decode_partitions(MpegEncContext *s);
926 int ff_mpeg4_get_video_packet_prefix_length(MpegEncContext *s);
927 int ff_h263_resync(MpegEncContext *s);
928 int ff_h263_get_gob_height(MpegEncContext *s);
929 int ff_mpeg4_set_direct_mv(MpegEncContext *s, int mx, int my);
930 int ff_h263_round_chroma(int x);
931 void ff_h263_encode_motion(MpegEncContext * s, int val, int f_code);
932 int ff_mpeg4_find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size);
933
934
935 /* rv10.c */
936 void rv10_encode_picture_header(MpegEncContext *s, int picture_number);
937 int rv_decode_dc(MpegEncContext *s, int n);
938 void rv20_encode_picture_header(MpegEncContext *s, int picture_number);
939
940
941 /* msmpeg4.c */
942 void msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number);
943 void msmpeg4_encode_ext_header(MpegEncContext * s);
944 void msmpeg4_encode_mb(MpegEncContext * s,
945 DCTELEM block[6][64],
946 int motion_x, int motion_y);
947 int msmpeg4_decode_picture_header(MpegEncContext * s);
948 int msmpeg4_decode_ext_header(MpegEncContext * s, int buf_size);
949 int ff_msmpeg4_decode_init(MpegEncContext *s);
950 void ff_msmpeg4_encode_init(MpegEncContext *s);
951 int ff_wmv2_decode_picture_header(MpegEncContext * s);
952 int ff_wmv2_decode_secondary_picture_header(MpegEncContext * s);
953 void ff_wmv2_add_mb(MpegEncContext *s, DCTELEM block[6][64], uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr);
954 void ff_mspel_motion(MpegEncContext *s,
955 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
956 uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
957 int motion_x, int motion_y, int h);
958 int ff_wmv2_encode_picture_header(MpegEncContext * s, int picture_number);
959 void ff_wmv2_encode_mb(MpegEncContext * s,
960 DCTELEM block[6][64],
961 int motion_x, int motion_y);
962
963 /* mjpeg.c */
964 int mjpeg_init(MpegEncContext *s);
965 void mjpeg_close(MpegEncContext *s);
966 void mjpeg_encode_mb(MpegEncContext *s,
967 DCTELEM block[6][64]);
968 void mjpeg_picture_header(MpegEncContext *s);
969 void mjpeg_picture_trailer(MpegEncContext *s);
970 void ff_mjpeg_stuffing(PutBitContext * pbc);
971
972
973 /* rate control */
974 int ff_rate_control_init(MpegEncContext *s);
975 float ff_rate_estimate_qscale(MpegEncContext *s);
976 void ff_write_pass1_stats(MpegEncContext *s);
977 void ff_rate_control_uninit(MpegEncContext *s);
978 double ff_eval(char *s, double *const_value, const char **const_name,
979 double (**func1)(void *, double), const char **func1_name,
980 double (**func2)(void *, double, double), char **func2_name,
981 void *opaque);
982 int ff_vbv_update(MpegEncContext *s, int frame_size);
983
984
985 #endif /* AVCODEC_MPEGVIDEO_H */