mpegvideo: Move tables to a separate file
[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 file is part of Libav.
7 *
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 /**
24 * @file
25 * mpegvideo header.
26 */
27
28 #ifndef AVCODEC_MPEGVIDEO_H
29 #define AVCODEC_MPEGVIDEO_H
30
31 #include <float.h>
32
33 #include "avcodec.h"
34 #include "blockdsp.h"
35 #include "error_resilience.h"
36 #include "fdctdsp.h"
37 #include "get_bits.h"
38 #include "h263dsp.h"
39 #include "hpeldsp.h"
40 #include "idctdsp.h"
41 #include "me_cmp.h"
42 #include "mpegvideodsp.h"
43 #include "mpegvideoencdsp.h"
44 #include "pixblockdsp.h"
45 #include "put_bits.h"
46 #include "ratecontrol.h"
47 #include "parser.h"
48 #include "mpeg12data.h"
49 #include "qpeldsp.h"
50 #include "thread.h"
51 #include "videodsp.h"
52
53 #include "libavutil/opt.h"
54
55 #define FRAME_SKIPPED 100 ///< return value for header parsers if frame is not coded
56
57 enum OutputFormat {
58 FMT_MPEG1,
59 FMT_H261,
60 FMT_H263,
61 FMT_MJPEG,
62 };
63
64 #define MAX_FCODE 7
65 #define MAX_MV 2048
66
67 #define MAX_THREADS 16
68
69 #define MAX_PICTURE_COUNT 32
70
71 #define MAX_B_FRAMES 16
72
73 #define ME_MAP_SIZE 64
74
75 #define MAX_MB_BYTES (30*16*16*3/8 + 120)
76
77 #define INPLACE_OFFSET 16
78
79 #define EDGE_WIDTH 16
80
81 /* Start codes. */
82 #define SEQ_END_CODE 0x000001b7
83 #define SEQ_START_CODE 0x000001b3
84 #define GOP_START_CODE 0x000001b8
85 #define PICTURE_START_CODE 0x00000100
86 #define SLICE_MIN_START_CODE 0x00000101
87 #define SLICE_MAX_START_CODE 0x000001af
88 #define EXT_START_CODE 0x000001b5
89 #define USER_START_CODE 0x000001b2
90
91 struct MpegEncContext;
92
93 /**
94 * Picture.
95 */
96 typedef struct Picture{
97 struct AVFrame *f;
98 ThreadFrame tf;
99
100 AVBufferRef *qscale_table_buf;
101 int8_t *qscale_table;
102
103 AVBufferRef *motion_val_buf[2];
104 int16_t (*motion_val[2])[2];
105
106 AVBufferRef *mb_type_buf;
107 uint32_t *mb_type; ///< types and macros are defined in mpegutils.h
108
109 AVBufferRef *mbskip_table_buf;
110 uint8_t *mbskip_table;
111
112 AVBufferRef *ref_index_buf[2];
113 int8_t *ref_index[2];
114
115 AVBufferRef *mb_var_buf;
116 uint16_t *mb_var; ///< Table for MB variances
117
118 AVBufferRef *mc_mb_var_buf;
119 uint16_t *mc_mb_var; ///< Table for motion compensated MB variances
120
121 AVBufferRef *mb_mean_buf;
122 uint8_t *mb_mean; ///< Table for MB luminance
123
124 AVBufferRef *hwaccel_priv_buf;
125 /**
126 * hardware accelerator private data
127 */
128 void *hwaccel_picture_private;
129
130 int field_picture; ///< whether or not the picture was encoded in separate fields
131
132 int mb_var_sum; ///< sum of MB variance for current frame
133 int mc_mb_var_sum; ///< motion compensated MB variance for current frame
134
135 int b_frame_score; /* */
136 int needs_realloc; ///< Picture needs to be reallocated (eg due to a frame size change)
137
138 int reference;
139 int shared;
140 } Picture;
141
142 /**
143 * Motion estimation context.
144 */
145 typedef struct MotionEstContext{
146 AVCodecContext *avctx;
147 int skip; ///< set if ME is skipped for the current MB
148 int co_located_mv[4][2]; ///< mv from last P-frame for direct mode ME
149 int direct_basis_mv[4][2];
150 uint8_t *scratchpad; ///< data area for the ME algo, so that the ME does not need to malloc/free
151 uint8_t *best_mb;
152 uint8_t *temp_mb[2];
153 uint8_t *temp;
154 int best_bits;
155 uint32_t *map; ///< map to avoid duplicate evaluations
156 uint32_t *score_map; ///< map to store the scores
157 unsigned map_generation;
158 int pre_penalty_factor;
159 int penalty_factor; /**< an estimate of the bits required to
160 code a given mv value, e.g. (1,0) takes
161 more bits than (0,0). We have to
162 estimate whether any reduction in
163 residual is worth the extra bits. */
164 int sub_penalty_factor;
165 int mb_penalty_factor;
166 int flags;
167 int sub_flags;
168 int mb_flags;
169 int pre_pass; ///< = 1 for the pre pass
170 int dia_size;
171 int xmin;
172 int xmax;
173 int ymin;
174 int ymax;
175 int pred_x;
176 int pred_y;
177 uint8_t *src[4][4];
178 uint8_t *ref[4][4];
179 int stride;
180 int uvstride;
181 /* temp variables for picture complexity calculation */
182 int mc_mb_var_sum_temp;
183 int mb_var_sum_temp;
184 int scene_change_score;
185 /* cmp, chroma_cmp;*/
186 op_pixels_func (*hpel_put)[4];
187 op_pixels_func (*hpel_avg)[4];
188 qpel_mc_func (*qpel_put)[16];
189 qpel_mc_func (*qpel_avg)[16];
190 uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
191 uint8_t *current_mv_penalty;
192 int (*sub_motion_search)(struct MpegEncContext * s,
193 int *mx_ptr, int *my_ptr, int dmin,
194 int src_index, int ref_index,
195 int size, int h);
196 }MotionEstContext;
197
198 /**
199 * MpegEncContext.
200 */
201 typedef struct MpegEncContext {
202 AVClass *class;
203
204 int y_dc_scale, c_dc_scale;
205 int ac_pred;
206 int block_last_index[12]; ///< last non zero coefficient in block
207 int h263_aic; ///< Advanded INTRA Coding (AIC)
208
209 /* scantables */
210 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
211 ScanTable intra_scantable;
212 ScanTable intra_h_scantable;
213 ScanTable intra_v_scantable;
214
215 /* WARNING: changes above this line require updates to hardcoded
216 * offsets used in asm. */
217
218 struct AVCodecContext *avctx;
219 /* the following parameters must be initialized before encoding */
220 int width, height;///< picture size. must be a multiple of 16
221 int gop_size;
222 int intra_only; ///< if true, only intra pictures are generated
223 int bit_rate; ///< wanted bit rate
224 enum OutputFormat out_format; ///< output format
225 int h263_pred; ///< use mpeg4/h263 ac/dc predictions
226 int pb_frame; ///< PB frame mode (0 = none, 1 = base, 2 = improved)
227
228 /* the following codec id fields are deprecated in favor of codec_id */
229 int h263_plus; ///< h263 plus headers
230 int h263_flv; ///< use flv h263 header
231
232 enum AVCodecID codec_id; /* see AV_CODEC_ID_xxx */
233 int fixed_qscale; ///< fixed qscale if non zero
234 int encoding; ///< true if we are encoding (vs decoding)
235 int max_b_frames; ///< max number of b-frames for encoding
236 int luma_elim_threshold;
237 int chroma_elim_threshold;
238 int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
239 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
240 int codec_tag; ///< internal codec_tag upper case converted from avctx codec_tag
241 /* the following fields are managed internally by the encoder */
242
243 /* sequence parameters */
244 int context_initialized;
245 int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
246 int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
247 int picture_number; //FIXME remove, unclear definition
248 int picture_in_gop_number; ///< 0-> first pic in gop, ...
249 int mb_width, mb_height; ///< number of MBs horizontally & vertically
250 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
251 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
252 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
253 int mb_num; ///< number of MBs of a picture
254 ptrdiff_t linesize; ///< line size, in bytes, may be different from width
255 ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
256 Picture *picture; ///< main picture buffer
257 Picture **input_picture; ///< next pictures on display order for encoding
258 Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
259
260 int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_encode_video2()
261 /**
262 * pts difference between the first and second input frame, used for
263 * calculating dts of the first frame when there's a delay */
264 int64_t dts_delta;
265 /**
266 * reordered pts to be used as dts for the next output frame when there's
267 * a delay */
268 int64_t reordered_pts;
269
270 /** bit output */
271 PutBitContext pb;
272
273 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
274 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
275 struct MpegEncContext *thread_context[MAX_THREADS];
276 int slice_context_count; ///< number of used thread_contexts
277
278 /**
279 * copy of the previous picture structure.
280 * note, linesize & data, might not match the previous picture (for field pictures)
281 */
282 Picture last_picture;
283
284 /**
285 * copy of the next picture structure.
286 * note, linesize & data, might not match the next picture (for field pictures)
287 */
288 Picture next_picture;
289
290 /**
291 * copy of the source picture structure for encoding.
292 * note, linesize & data, might not match the source picture (for field pictures)
293 */
294 Picture new_picture;
295
296 /**
297 * copy of the current picture structure.
298 * note, linesize & data, might not match the current picture (for field pictures)
299 */
300 Picture current_picture; ///< buffer to store the decompressed current picture
301
302 Picture *last_picture_ptr; ///< pointer to the previous picture.
303 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
304 Picture *current_picture_ptr; ///< pointer to the current picture
305 int last_dc[3]; ///< last DC values for MPEG1
306 int16_t *dc_val_base;
307 int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
308 const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
309 const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
310 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
311 uint8_t *coded_block_base;
312 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
313 int16_t (*ac_val_base)[16];
314 int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
315 int mb_skipped; ///< MUST BE SET only during DECODING
316 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
317 and used for b-frame encoding & decoding (contains skip table of next P Frame) */
318 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
319 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
320 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
321 uint8_t *edge_emu_buffer; ///< temporary buffer for if MVs point to out-of-frame data
322 uint8_t *rd_scratchpad; ///< scratchpad for rate distortion mb decision
323 uint8_t *obmc_scratchpad;
324 uint8_t *b_scratchpad; ///< scratchpad used for writing into write only buffers
325
326 int qscale; ///< QP
327 int chroma_qscale; ///< chroma QP
328 unsigned int lambda; ///< lagrange multipler used in rate distortion
329 unsigned int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
330 int *lambda_table;
331 int adaptive_quant; ///< use adaptive quantization
332 int dquant; ///< qscale difference to prev qscale
333 int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ...
334 int last_pict_type; //FIXME removes
335 int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
336 int droppable;
337 int frame_rate_index;
338 int last_lambda_for[5]; ///< last lambda for a specific pict type
339 int skipdct; ///< skip dct and code zero residual
340
341 /* motion compensation */
342 int unrestricted_mv; ///< mv can point outside of the coded picture
343 int h263_long_vectors; ///< use horrible h263v1 long vector mode
344
345 BlockDSPContext bdsp;
346 FDCTDSPContext fdsp;
347 HpelDSPContext hdsp;
348 IDCTDSPContext idsp;
349 MECmpContext mecc;
350 MpegVideoDSPContext mdsp;
351 MpegvideoEncDSPContext mpvencdsp;
352 PixblockDSPContext pdsp;
353 QpelDSPContext qdsp;
354 VideoDSPContext vdsp;
355 H263DSPContext h263dsp;
356 int f_code; ///< forward MV resolution
357 int b_code; ///< backward MV resolution for B Frames (mpeg4)
358 int16_t (*p_mv_table_base)[2];
359 int16_t (*b_forw_mv_table_base)[2];
360 int16_t (*b_back_mv_table_base)[2];
361 int16_t (*b_bidir_forw_mv_table_base)[2];
362 int16_t (*b_bidir_back_mv_table_base)[2];
363 int16_t (*b_direct_mv_table_base)[2];
364 int16_t (*p_field_mv_table_base[2][2])[2];
365 int16_t (*b_field_mv_table_base[2][2][2])[2];
366 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
367 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
368 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
369 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
370 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
371 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
372 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
373 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
374 uint8_t (*p_field_select_table[2]);
375 uint8_t (*b_field_select_table[2][2]);
376 int me_method; ///< ME algorithm
377 int mv_dir;
378 #define MV_DIR_FORWARD 1
379 #define MV_DIR_BACKWARD 2
380 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
381 int mv_type;
382 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
383 #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
384 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
385 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
386 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
387 /**motion vectors for a macroblock
388 first coordinate : 0 = forward 1 = backward
389 second " : depend on type
390 third " : 0 = x, 1 = y
391 */
392 int mv[2][4][2];
393 int field_select[2][2];
394 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
395 uint8_t *fcode_tab; ///< smallest fcode needed for each MV
396 int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
397
398 MotionEstContext me;
399
400 int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
401 for b-frames rounding mode is always 0 */
402
403 /* macroblock layer */
404 int mb_x, mb_y;
405 int mb_skip_run;
406 int mb_intra;
407 uint16_t *mb_type; ///< Table for candidate MB types for encoding (defines in mpegutils.h)
408
409 int block_index[6]; ///< index to current MB in block based arrays with edges
410 int block_wrap[6];
411 uint8_t *dest[3];
412
413 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
414
415 /** matrix transmitted in the bitstream */
416 uint16_t intra_matrix[64];
417 uint16_t chroma_intra_matrix[64];
418 uint16_t inter_matrix[64];
419 uint16_t chroma_inter_matrix[64];
420
421 int intra_quant_bias; ///< bias for the quantizer
422 int inter_quant_bias; ///< bias for the quantizer
423 int min_qcoeff; ///< minimum encodable coefficient
424 int max_qcoeff; ///< maximum encodable coefficient
425 int ac_esc_length; ///< num of bits needed to encode the longest esc
426 uint8_t *intra_ac_vlc_length;
427 uint8_t *intra_ac_vlc_last_length;
428 uint8_t *inter_ac_vlc_length;
429 uint8_t *inter_ac_vlc_last_length;
430 uint8_t *luma_dc_vlc_length;
431 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
432
433 int coded_score[8];
434
435 /** precomputed matrix (combine qscale and DCT renorm) */
436 int (*q_intra_matrix)[64];
437 int (*q_inter_matrix)[64];
438 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
439 uint16_t (*q_intra_matrix16)[2][64];
440 uint16_t (*q_inter_matrix16)[2][64];
441
442 /* noise reduction */
443 int (*dct_error_sum)[64];
444 int dct_count[2];
445 uint16_t (*dct_offset)[64];
446
447 /* bit rate control */
448 int64_t total_bits;
449 int frame_bits; ///< bits used for the current frame
450 int next_lambda; ///< next lambda used for retrying to encode a frame
451 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
452
453 /* statistics, used for 2-pass encoding */
454 int mv_bits;
455 int header_bits;
456 int i_tex_bits;
457 int p_tex_bits;
458 int i_count;
459 int f_count;
460 int b_count;
461 int skip_count;
462 int misc_bits; ///< cbp, mb_type
463 int last_bits; ///< temp var used for calculating the above vars
464
465 /* error concealment / resync */
466 int resync_mb_x; ///< x position of last resync marker
467 int resync_mb_y; ///< y position of last resync marker
468 GetBitContext last_resync_gb; ///< used to search for the next resync marker
469 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
470 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
471
472 ParseContext parse_context;
473
474 /* H.263 specific */
475 int gob_index;
476 int obmc; ///< overlapped block motion compensation
477 int mb_info; ///< interval for outputting info about mb offsets as side data
478 int prev_mb_info, last_mb_info;
479 uint8_t *mb_info_ptr;
480 int mb_info_size;
481
482 /* H.263+ specific */
483 int umvplus; ///< == H263+ && unrestricted_mv
484 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
485 int h263_slice_structured;
486 int alt_inter_vlc; ///< alternative inter vlc
487 int modified_quant;
488 int loop_filter;
489 int custom_pcf;
490
491 /* mpeg4 specific */
492 ///< number of bits to represent the fractional part of time (encoder only)
493 int time_increment_bits;
494 int last_time_base;
495 int time_base; ///< time in seconds of last I,P,S Frame
496 int64_t time; ///< time of current frame
497 int64_t last_non_b_time;
498 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
499 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
500 uint16_t pp_field_time;
501 uint16_t pb_field_time; ///< like above, just for interlaced
502 int real_sprite_warping_points;
503 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
504 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
505 int mcsel;
506 int quant_precision;
507 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
508 int aspect_ratio_info; //FIXME remove
509 int sprite_warping_accuracy;
510 int data_partitioning; ///< data partitioning flag from header
511 int partitioned_frame; ///< is current frame partitioned
512 int low_delay; ///< no reordering needed / has no b-frames
513 int vo_type;
514 PutBitContext tex_pb; ///< used for data partitioned VOPs
515 PutBitContext pb2; ///< used for data partitioned VOPs
516 int mpeg_quant;
517 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
518
519 /* divx specific, used to workaround (many) bugs in divx5 */
520 int divx_packed;
521 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
522 int bitstream_buffer_size;
523 unsigned int allocated_bitstream_buffer_size;
524
525 /* RV10 specific */
526 int rv10_version; ///< RV10 version: 0 or 3
527 int rv10_first_dc_coded[3];
528
529 /* MJPEG specific */
530 struct MJpegContext *mjpeg_ctx;
531
532 /* MSMPEG4 specific */
533 int mv_table_index;
534 int rl_table_index;
535 int rl_chroma_table_index;
536 int dc_table_index;
537 int use_skip_mb_code;
538 int slice_height; ///< in macroblocks
539 int first_slice_line; ///< used in mpeg4 too to handle resync markers
540 int flipflop_rounding;
541 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
542 int per_mb_rl_table;
543 int esc3_level_length;
544 int esc3_run_length;
545 /** [mb_intra][isChroma][level][run][last] */
546 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
547 int inter_intra_pred;
548 int mspel;
549
550 /* decompression specific */
551 GetBitContext gb;
552
553 /* Mpeg1 specific */
554 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
555 int last_mv_dir; ///< last mv_dir, used for b frame encoding
556 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
557
558 /* MPEG-2-specific - I wished not to have to support this mess. */
559 int progressive_sequence;
560 int mpeg_f_code[2][2];
561
562 // picture structure defines are loaded from mpegutils.h
563 int picture_structure;
564
565 int intra_dc_precision;
566 int frame_pred_frame_dct;
567 int top_field_first;
568 int concealment_motion_vectors;
569 int q_scale_type;
570 int intra_vlc_format;
571 int alternate_scan;
572 int repeat_first_field;
573 int chroma_420_type;
574 int chroma_format;
575 #define CHROMA_420 1
576 #define CHROMA_422 2
577 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
578 int chroma_y_shift;
579
580 int progressive_frame;
581 int full_pel[2];
582 int interlaced_dct;
583 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
584 int drop_frame_timecode; ///< timecode is in drop frame format.
585 int scan_offset; ///< reserve space for SVCD scan offset user data.
586
587 /* RTP specific */
588 int rtp_mode;
589
590 uint8_t *ptr_lastgob;
591 int16_t (*pblocks[12])[64];
592
593 int16_t (*block)[64]; ///< points to one of the following blocks
594 int16_t (*blocks)[12][64]; // for HQ mode we need to keep the best block
595 int (*decode_mb)(struct MpegEncContext *s, int16_t block[6][64]); // used by some codecs to avoid a switch()
596 #define SLICE_OK 0
597 #define SLICE_ERROR -1
598 #define SLICE_END -2 ///<end marker found
599 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
600
601 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
602 int16_t *block/*align 16*/, int n, int qscale);
603 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
604 int16_t *block/*align 16*/, int n, int qscale);
605 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
606 int16_t *block/*align 16*/, int n, int qscale);
607 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
608 int16_t *block/*align 16*/, int n, int qscale);
609 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
610 int16_t *block/*align 16*/, int n, int qscale);
611 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
612 int16_t *block/*align 16*/, int n, int qscale);
613 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
614 int16_t *block/*align 16*/, int n, int qscale);
615 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
616 int16_t *block/*align 16*/, int n, int qscale);
617 int (*dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
618 int (*fast_dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
619 void (*denoise_dct)(struct MpegEncContext *s, int16_t *block);
620
621 int mpv_flags; ///< flags set by private options
622 int quantizer_noise_shaping;
623
624 /**
625 * ratecontrol qmin qmax limiting method
626 * 0-> clipping, 1-> use a nice continuous function to limit qscale wthin qmin/qmax.
627 */
628 float rc_qsquish;
629 float rc_qmod_amp;
630 int rc_qmod_freq;
631 float rc_initial_cplx;
632 float rc_buffer_aggressivity;
633 float border_masking;
634 int lmin, lmax;
635
636 char *rc_eq;
637
638 /* temp buffers for rate control */
639 float *cplx_tab, *bits_tab;
640
641 /* flag to indicate a reinitialization is required, e.g. after
642 * a frame size change */
643 int context_reinit;
644
645 ERContext er;
646
647 int error_rate;
648
649 /* temporary frames used by b_frame_strategy = 2 */
650 AVFrame *tmp_frames[MAX_B_FRAMES + 2];
651 } MpegEncContext;
652
653 /* mpegvideo_enc common options */
654 #define FF_MPV_FLAG_SKIP_RD 0x0001
655 #define FF_MPV_FLAG_STRICT_GOP 0x0002
656 #define FF_MPV_FLAG_QP_RD 0x0004
657 #define FF_MPV_FLAG_CBP_RD 0x0008
658 #define FF_MPV_FLAG_NAQ 0x0010
659 #define FF_MPV_FLAG_MV0 0x0020
660
661 #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x)
662 #define FF_MPV_OPT_FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
663 #define FF_MPV_COMMON_OPTS \
664 { "mpv_flags", "Flags common for all mpegvideo-based encoders.", FF_MPV_OFFSET(mpv_flags), AV_OPT_TYPE_FLAGS, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "mpv_flags" },\
665 { "skip_rd", "RD optimal MB level residual skipping", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_SKIP_RD }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
666 { "strict_gop", "Strictly enforce gop size", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_STRICT_GOP }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
667 { "qp_rd", "Use rate distortion optimization for qp selection", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_QP_RD }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
668 { "cbp_rd", "use rate distortion optimization for CBP", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_CBP_RD }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
669 { "naq", "normalize adaptive quantization", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_NAQ }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
670 { "mv0", "always try a mb with mv=<0,0>", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_MV0 }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
671 { "luma_elim_threshold", "single coefficient elimination threshold for luminance (negative values also consider dc coefficient)",\
672 FF_MPV_OFFSET(luma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
673 { "chroma_elim_threshold", "single coefficient elimination threshold for chrominance (negative values also consider dc coefficient)",\
674 FF_MPV_OFFSET(chroma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
675 { "quantizer_noise_shaping", NULL, FF_MPV_OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
676 { "error_rate", "Simulate errors in the bitstream to test error concealment.", \
677 FF_MPV_OFFSET(error_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
678 {"qsquish", "how to keep quantizer between qmin and qmax (0 = clip, 1 = use differentiable function)", \
679 FF_MPV_OFFSET(rc_qsquish), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, 0, 99, FF_MPV_OPT_FLAGS}, \
680 {"rc_qmod_amp", "experimental quantizer modulation", FF_MPV_OFFSET(rc_qmod_amp), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, -FLT_MAX, FLT_MAX, FF_MPV_OPT_FLAGS}, \
681 {"rc_qmod_freq", "experimental quantizer modulation", FF_MPV_OFFSET(rc_qmod_freq), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS}, \
682 {"rc_eq", "Set rate control equation. When computing the expression, besides the standard functions " \
683 "defined in the section 'Expression Evaluation', the following functions are available: " \
684 "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv " \
685 "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.", \
686 FF_MPV_OFFSET(rc_eq), AV_OPT_TYPE_STRING, .flags = FF_MPV_OPT_FLAGS }, \
687 {"rc_init_cplx", "initial complexity for 1-pass encoding", FF_MPV_OFFSET(rc_initial_cplx), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, -FLT_MAX, FLT_MAX, FF_MPV_OPT_FLAGS}, \
688 {"rc_buf_aggressivity", "currently useless", FF_MPV_OFFSET(rc_buffer_aggressivity), AV_OPT_TYPE_FLOAT, {.dbl = 1.0 }, -FLT_MAX, FLT_MAX, FF_MPV_OPT_FLAGS}, \
689 {"border_mask", "increase the quantizer for macroblocks close to borders", FF_MPV_OFFSET(border_masking), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, -FLT_MAX, FLT_MAX, FF_MPV_OPT_FLAGS}, \
690 {"lmin", "minimum Lagrange factor (VBR)", FF_MPV_OFFSET(lmin), AV_OPT_TYPE_INT, {.i64 = 2*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
691 {"lmax", "maximum Lagrange factor (VBR)", FF_MPV_OFFSET(lmax), AV_OPT_TYPE_INT, {.i64 = 31*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
692
693 extern const AVOption ff_mpv_generic_options[];
694
695 #define FF_MPV_GENERIC_CLASS(name) \
696 static const AVClass name ## _class = {\
697 .class_name = #name " encoder",\
698 .item_name = av_default_item_name,\
699 .option = ff_mpv_generic_options,\
700 .version = LIBAVUTIL_VERSION_INT,\
701 };
702
703 /**
704 * Set the given MpegEncContext to common defaults (same for encoding
705 * and decoding). The changed fields will not depend upon the prior
706 * state of the MpegEncContext.
707 */
708 void ff_mpv_common_defaults(MpegEncContext *s);
709
710 int ff_mpv_common_init(MpegEncContext *s);
711 void ff_mpv_common_init_arm(MpegEncContext *s);
712 void ff_mpv_common_init_neon(MpegEncContext *s);
713 void ff_mpv_common_init_ppc(MpegEncContext *s);
714 void ff_mpv_common_init_x86(MpegEncContext *s);
715
716 int ff_mpv_common_frame_size_change(MpegEncContext *s);
717 void ff_mpv_common_end(MpegEncContext *s);
718
719 void ff_mpv_decode_defaults(MpegEncContext *s);
720 void ff_mpv_decode_mb(MpegEncContext *s, int16_t block[12][64]);
721 void ff_mpv_report_decode_progress(MpegEncContext *s);
722
723 int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx);
724 void ff_mpv_frame_end(MpegEncContext *s);
725
726 int ff_mpv_lowest_referenced_row(MpegEncContext *s, int dir);
727
728 int ff_mpv_encode_init(AVCodecContext *avctx);
729 void ff_mpv_encode_init_x86(MpegEncContext *s);
730
731 int ff_mpv_encode_end(AVCodecContext *avctx);
732 int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
733 const AVFrame *frame, int *got_packet);
734
735 void ff_clean_intra_table_entries(MpegEncContext *s);
736 void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h);
737 void ff_mpeg_flush(AVCodecContext *avctx);
738 void ff_print_debug_info(MpegEncContext *s, Picture *p);
739 void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
740
741 int ff_find_unused_picture(AVCodecContext *avctx, Picture *picture, int shared);
742 int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
743 int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
744 void ff_set_qscale(MpegEncContext * s, int qscale);
745
746 void ff_mpv_idct_init(MpegEncContext *s);
747 void ff_convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16)[2][64],
748 const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra);
749 int ff_dct_quantize_c(MpegEncContext *s, int16_t *block, int n, int qscale, int *overflow);
750
751 void ff_init_block_index(MpegEncContext *s);
752
753 void ff_mpv_motion(MpegEncContext *s,
754 uint8_t *dest_y, uint8_t *dest_cb,
755 uint8_t *dest_cr, int dir,
756 uint8_t **ref_picture,
757 op_pixels_func (*pix_op)[4],
758 qpel_mc_func (*qpix_op)[16]);
759
760 /**
761 * Allocate a Picture.
762 * The pixels are allocated/set by calling get_buffer() if shared = 0.
763 */
764 int ff_alloc_picture(MpegEncContext *s, Picture *pic, int shared);
765
766 /**
767 * permute block according to permuatation.
768 * @param last last non zero element in scantable order
769 */
770 void ff_block_permute(int16_t *block, uint8_t *permutation, const uint8_t *scantable, int last);
771
772 static inline void ff_update_block_index(MpegEncContext *s){
773 const int block_size = 8;
774
775 s->block_index[0]+=2;
776 s->block_index[1]+=2;
777 s->block_index[2]+=2;
778 s->block_index[3]+=2;
779 s->block_index[4]++;
780 s->block_index[5]++;
781 s->dest[0]+= 2*block_size;
782 s->dest[1]+= block_size;
783 s->dest[2]+= block_size;
784 }
785
786 static inline int get_bits_diff(MpegEncContext *s){
787 const int bits= put_bits_count(&s->pb);
788 const int last= s->last_bits;
789
790 s->last_bits = bits;
791
792 return bits - last;
793 }
794
795 static inline int ff_h263_round_chroma(int x){
796 static const uint8_t h263_chroma_roundtab[16] = {
797 // 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
798 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1,
799 };
800 return h263_chroma_roundtab[x & 0xf] + (x >> 3);
801 }
802
803 /* motion_est.c */
804 void ff_estimate_p_frame_motion(MpegEncContext * s,
805 int mb_x, int mb_y);
806 void ff_estimate_b_frame_motion(MpegEncContext * s,
807 int mb_x, int mb_y);
808 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
809 void ff_fix_long_p_mvs(MpegEncContext * s);
810 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
811 int16_t (*mv_table)[2], int f_code, int type, int truncate);
812 int ff_init_me(MpegEncContext *s);
813 int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
814 int ff_epzs_motion_search(MpegEncContext * s, int *mx_ptr, int *my_ptr,
815 int P[10][2], int src_index, int ref_index, int16_t (*last_mv)[2],
816 int ref_mv_scale, int size, int h);
817 int ff_get_mb_score(MpegEncContext * s, int mx, int my, int src_index,
818 int ref_index, int size, int h, int add_rate);
819
820 void ff_mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
821 void ff_mpeg1_encode_mb(MpegEncContext *s,
822 int16_t block[8][64],
823 int motion_x, int motion_y);
824 void ff_mpeg1_encode_init(MpegEncContext *s);
825 void ff_mpeg1_encode_slice_header(MpegEncContext *s);
826
827 extern const uint8_t ff_aic_dc_scale_table[32];
828 extern const uint8_t ff_h263_chroma_qscale_table[32];
829
830 /* rv10.c */
831 int ff_rv10_encode_picture_header(MpegEncContext *s, int picture_number);
832 int ff_rv_decode_dc(MpegEncContext *s, int n);
833 void ff_rv20_encode_picture_header(MpegEncContext *s, int picture_number);
834
835 int ff_mpeg_ref_picture(AVCodecContext *avctx, Picture *dst, Picture *src);
836 void ff_mpeg_unref_picture(AVCodecContext *avctx, Picture *picture);
837 void ff_free_picture_tables(Picture *pic);
838
839 #endif /* AVCODEC_MPEGVIDEO_H */