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