48c9cffc89f38afcaba8fc891e1fc982d21fbaa1
[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 /* encoding scans */
92 extern const uint8_t ff_alternate_horizontal_scan[64];
93 extern const uint8_t ff_alternate_vertical_scan[64];
94
95 struct MpegEncContext;
96
97 /**
98 * Picture.
99 */
100 typedef struct Picture{
101 struct AVFrame *f;
102 ThreadFrame tf;
103
104 AVBufferRef *qscale_table_buf;
105 int8_t *qscale_table;
106
107 AVBufferRef *motion_val_buf[2];
108 int16_t (*motion_val[2])[2];
109
110 AVBufferRef *mb_type_buf;
111 uint32_t *mb_type; ///< types and macros are defined in mpegutils.h
112
113 AVBufferRef *mbskip_table_buf;
114 uint8_t *mbskip_table;
115
116 AVBufferRef *ref_index_buf[2];
117 int8_t *ref_index[2];
118
119 AVBufferRef *mb_var_buf;
120 uint16_t *mb_var; ///< Table for MB variances
121
122 AVBufferRef *mc_mb_var_buf;
123 uint16_t *mc_mb_var; ///< Table for motion compensated MB variances
124
125 AVBufferRef *mb_mean_buf;
126 uint8_t *mb_mean; ///< Table for MB luminance
127
128 AVBufferRef *hwaccel_priv_buf;
129 /**
130 * hardware accelerator private data
131 */
132 void *hwaccel_picture_private;
133
134 int field_picture; ///< whether or not the picture was encoded in separate fields
135
136 int mb_var_sum; ///< sum of MB variance for current frame
137 int mc_mb_var_sum; ///< motion compensated MB variance for current frame
138
139 int b_frame_score; /* */
140 int needs_realloc; ///< Picture needs to be reallocated (eg due to a frame size change)
141
142 int reference;
143 int shared;
144 } Picture;
145
146 /**
147 * Motion estimation context.
148 */
149 typedef struct MotionEstContext{
150 AVCodecContext *avctx;
151 int skip; ///< set if ME is skipped for the current MB
152 int co_located_mv[4][2]; ///< mv from last P-frame for direct mode ME
153 int direct_basis_mv[4][2];
154 uint8_t *scratchpad; ///< data area for the ME algo, so that the ME does not need to malloc/free
155 uint8_t *best_mb;
156 uint8_t *temp_mb[2];
157 uint8_t *temp;
158 int best_bits;
159 uint32_t *map; ///< map to avoid duplicate evaluations
160 uint32_t *score_map; ///< map to store the scores
161 unsigned map_generation;
162 int pre_penalty_factor;
163 int penalty_factor; /**< an estimate of the bits required to
164 code a given mv value, e.g. (1,0) takes
165 more bits than (0,0). We have to
166 estimate whether any reduction in
167 residual is worth the extra bits. */
168 int sub_penalty_factor;
169 int mb_penalty_factor;
170 int flags;
171 int sub_flags;
172 int mb_flags;
173 int pre_pass; ///< = 1 for the pre pass
174 int dia_size;
175 int xmin;
176 int xmax;
177 int ymin;
178 int ymax;
179 int pred_x;
180 int pred_y;
181 uint8_t *src[4][4];
182 uint8_t *ref[4][4];
183 int stride;
184 int uvstride;
185 /* temp variables for picture complexity calculation */
186 int mc_mb_var_sum_temp;
187 int mb_var_sum_temp;
188 int scene_change_score;
189 /* cmp, chroma_cmp;*/
190 op_pixels_func (*hpel_put)[4];
191 op_pixels_func (*hpel_avg)[4];
192 qpel_mc_func (*qpel_put)[16];
193 qpel_mc_func (*qpel_avg)[16];
194 uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
195 uint8_t *current_mv_penalty;
196 int (*sub_motion_search)(struct MpegEncContext * s,
197 int *mx_ptr, int *my_ptr, int dmin,
198 int src_index, int ref_index,
199 int size, int h);
200 }MotionEstContext;
201
202 /**
203 * MpegEncContext.
204 */
205 typedef struct MpegEncContext {
206 AVClass *class;
207
208 int y_dc_scale, c_dc_scale;
209 int ac_pred;
210 int block_last_index[12]; ///< last non zero coefficient in block
211 int h263_aic; ///< Advanded INTRA Coding (AIC)
212
213 /* scantables */
214 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
215 ScanTable intra_scantable;
216 ScanTable intra_h_scantable;
217 ScanTable intra_v_scantable;
218
219 /* WARNING: changes above this line require updates to hardcoded
220 * offsets used in asm. */
221
222 struct AVCodecContext *avctx;
223 /* the following parameters must be initialized before encoding */
224 int width, height;///< picture size. must be a multiple of 16
225 int gop_size;
226 int intra_only; ///< if true, only intra pictures are generated
227 int bit_rate; ///< wanted bit rate
228 enum OutputFormat out_format; ///< output format
229 int h263_pred; ///< use mpeg4/h263 ac/dc predictions
230 int pb_frame; ///< PB frame mode (0 = none, 1 = base, 2 = improved)
231
232 /* the following codec id fields are deprecated in favor of codec_id */
233 int h263_plus; ///< h263 plus headers
234 int h263_flv; ///< use flv h263 header
235
236 enum AVCodecID codec_id; /* see AV_CODEC_ID_xxx */
237 int fixed_qscale; ///< fixed qscale if non zero
238 int encoding; ///< true if we are encoding (vs decoding)
239 int max_b_frames; ///< max number of b-frames for encoding
240 int luma_elim_threshold;
241 int chroma_elim_threshold;
242 int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
243 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
244 int codec_tag; ///< internal codec_tag upper case converted from avctx codec_tag
245 /* the following fields are managed internally by the encoder */
246
247 /* sequence parameters */
248 int context_initialized;
249 int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
250 int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
251 int picture_number; //FIXME remove, unclear definition
252 int picture_in_gop_number; ///< 0-> first pic in gop, ...
253 int mb_width, mb_height; ///< number of MBs horizontally & vertically
254 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
255 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
256 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
257 int mb_num; ///< number of MBs of a picture
258 ptrdiff_t linesize; ///< line size, in bytes, may be different from width
259 ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
260 Picture *picture; ///< main picture buffer
261 Picture **input_picture; ///< next pictures on display order for encoding
262 Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
263
264 int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_encode_video2()
265 /**
266 * pts difference between the first and second input frame, used for
267 * calculating dts of the first frame when there's a delay */
268 int64_t dts_delta;
269 /**
270 * reordered pts to be used as dts for the next output frame when there's
271 * a delay */
272 int64_t reordered_pts;
273
274 /** bit output */
275 PutBitContext pb;
276
277 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
278 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
279 struct MpegEncContext *thread_context[MAX_THREADS];
280 int slice_context_count; ///< number of used thread_contexts
281
282 /**
283 * copy of the previous picture structure.
284 * note, linesize & data, might not match the previous picture (for field pictures)
285 */
286 Picture last_picture;
287
288 /**
289 * copy of the next picture structure.
290 * note, linesize & data, might not match the next picture (for field pictures)
291 */
292 Picture next_picture;
293
294 /**
295 * copy of the source picture structure for encoding.
296 * note, linesize & data, might not match the source picture (for field pictures)
297 */
298 Picture new_picture;
299
300 /**
301 * copy of the current picture structure.
302 * note, linesize & data, might not match the current picture (for field pictures)
303 */
304 Picture current_picture; ///< buffer to store the decompressed current picture
305
306 Picture *last_picture_ptr; ///< pointer to the previous picture.
307 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
308 Picture *current_picture_ptr; ///< pointer to the current picture
309 int last_dc[3]; ///< last DC values for MPEG1
310 int16_t *dc_val_base;
311 int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
312 const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
313 const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
314 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
315 uint8_t *coded_block_base;
316 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
317 int16_t (*ac_val_base)[16];
318 int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
319 int mb_skipped; ///< MUST BE SET only during DECODING
320 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
321 and used for b-frame encoding & decoding (contains skip table of next P Frame) */
322 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
323 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
324 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
325 uint8_t *edge_emu_buffer; ///< temporary buffer for if MVs point to out-of-frame data
326 uint8_t *rd_scratchpad; ///< scratchpad for rate distortion mb decision
327 uint8_t *obmc_scratchpad;
328 uint8_t *b_scratchpad; ///< scratchpad used for writing into write only buffers
329
330 int qscale; ///< QP
331 int chroma_qscale; ///< chroma QP
332 unsigned int lambda; ///< lagrange multipler used in rate distortion
333 unsigned int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
334 int *lambda_table;
335 int adaptive_quant; ///< use adaptive quantization
336 int dquant; ///< qscale difference to prev qscale
337 int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ...
338 int last_pict_type; //FIXME removes
339 int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
340 int droppable;
341 int frame_rate_index;
342 int last_lambda_for[5]; ///< last lambda for a specific pict type
343 int skipdct; ///< skip dct and code zero residual
344
345 /* motion compensation */
346 int unrestricted_mv; ///< mv can point outside of the coded picture
347 int h263_long_vectors; ///< use horrible h263v1 long vector mode
348
349 BlockDSPContext bdsp;
350 FDCTDSPContext fdsp;
351 HpelDSPContext hdsp;
352 IDCTDSPContext idsp;
353 MECmpContext mecc;
354 MpegVideoDSPContext mdsp;
355 MpegvideoEncDSPContext mpvencdsp;
356 PixblockDSPContext pdsp;
357 QpelDSPContext qdsp;
358 VideoDSPContext vdsp;
359 H263DSPContext h263dsp;
360 int f_code; ///< forward MV resolution
361 int b_code; ///< backward MV resolution for B Frames (mpeg4)
362 int16_t (*p_mv_table_base)[2];
363 int16_t (*b_forw_mv_table_base)[2];
364 int16_t (*b_back_mv_table_base)[2];
365 int16_t (*b_bidir_forw_mv_table_base)[2];
366 int16_t (*b_bidir_back_mv_table_base)[2];
367 int16_t (*b_direct_mv_table_base)[2];
368 int16_t (*p_field_mv_table_base[2][2])[2];
369 int16_t (*b_field_mv_table_base[2][2][2])[2];
370 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
371 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
372 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
373 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
374 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
375 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
376 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
377 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
378 uint8_t (*p_field_select_table[2]);
379 uint8_t (*b_field_select_table[2][2]);
380 int me_method; ///< ME algorithm
381 int mv_dir;
382 #define MV_DIR_FORWARD 1
383 #define MV_DIR_BACKWARD 2
384 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
385 int mv_type;
386 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
387 #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
388 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
389 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
390 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
391 /**motion vectors for a macroblock
392 first coordinate : 0 = forward 1 = backward
393 second " : depend on type
394 third " : 0 = x, 1 = y
395 */
396 int mv[2][4][2];
397 int field_select[2][2];
398 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
399 uint8_t *fcode_tab; ///< smallest fcode needed for each MV
400 int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
401
402 MotionEstContext me;
403
404 int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
405 for b-frames rounding mode is always 0 */
406
407 /* macroblock layer */
408 int mb_x, mb_y;
409 int mb_skip_run;
410 int mb_intra;
411 uint16_t *mb_type; ///< Table for candidate MB types for encoding (defines in mpegutils.h)
412
413 int block_index[6]; ///< index to current MB in block based arrays with edges
414 int block_wrap[6];
415 uint8_t *dest[3];
416
417 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
418
419 /** matrix transmitted in the bitstream */
420 uint16_t intra_matrix[64];
421 uint16_t chroma_intra_matrix[64];
422 uint16_t inter_matrix[64];
423 uint16_t chroma_inter_matrix[64];
424
425 int intra_quant_bias; ///< bias for the quantizer
426 int inter_quant_bias; ///< bias for the quantizer
427 int min_qcoeff; ///< minimum encodable coefficient
428 int max_qcoeff; ///< maximum encodable coefficient
429 int ac_esc_length; ///< num of bits needed to encode the longest esc
430 uint8_t *intra_ac_vlc_length;
431 uint8_t *intra_ac_vlc_last_length;
432 uint8_t *inter_ac_vlc_length;
433 uint8_t *inter_ac_vlc_last_length;
434 uint8_t *luma_dc_vlc_length;
435 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
436
437 int coded_score[8];
438
439 /** precomputed matrix (combine qscale and DCT renorm) */
440 int (*q_intra_matrix)[64];
441 int (*q_inter_matrix)[64];
442 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
443 uint16_t (*q_intra_matrix16)[2][64];
444 uint16_t (*q_inter_matrix16)[2][64];
445
446 /* noise reduction */
447 int (*dct_error_sum)[64];
448 int dct_count[2];
449 uint16_t (*dct_offset)[64];
450
451 /* bit rate control */
452 int64_t total_bits;
453 int frame_bits; ///< bits used for the current frame
454 int next_lambda; ///< next lambda used for retrying to encode a frame
455 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
456
457 /* statistics, used for 2-pass encoding */
458 int mv_bits;
459 int header_bits;
460 int i_tex_bits;
461 int p_tex_bits;
462 int i_count;
463 int f_count;
464 int b_count;
465 int skip_count;
466 int misc_bits; ///< cbp, mb_type
467 int last_bits; ///< temp var used for calculating the above vars
468
469 /* error concealment / resync */
470 int resync_mb_x; ///< x position of last resync marker
471 int resync_mb_y; ///< y position of last resync marker
472 GetBitContext last_resync_gb; ///< used to search for the next resync marker
473 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
474 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
475
476 ParseContext parse_context;
477
478 /* H.263 specific */
479 int gob_index;
480 int obmc; ///< overlapped block motion compensation
481 int mb_info; ///< interval for outputting info about mb offsets as side data
482 int prev_mb_info, last_mb_info;
483 uint8_t *mb_info_ptr;
484 int mb_info_size;
485
486 /* H.263+ specific */
487 int umvplus; ///< == H263+ && unrestricted_mv
488 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
489 int h263_slice_structured;
490 int alt_inter_vlc; ///< alternative inter vlc
491 int modified_quant;
492 int loop_filter;
493 int custom_pcf;
494
495 /* mpeg4 specific */
496 ///< number of bits to represent the fractional part of time (encoder only)
497 int time_increment_bits;
498 int last_time_base;
499 int time_base; ///< time in seconds of last I,P,S Frame
500 int64_t time; ///< time of current frame
501 int64_t last_non_b_time;
502 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
503 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
504 uint16_t pp_field_time;
505 uint16_t pb_field_time; ///< like above, just for interlaced
506 int real_sprite_warping_points;
507 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
508 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
509 int mcsel;
510 int quant_precision;
511 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
512 int aspect_ratio_info; //FIXME remove
513 int sprite_warping_accuracy;
514 int data_partitioning; ///< data partitioning flag from header
515 int partitioned_frame; ///< is current frame partitioned
516 int low_delay; ///< no reordering needed / has no b-frames
517 int vo_type;
518 PutBitContext tex_pb; ///< used for data partitioned VOPs
519 PutBitContext pb2; ///< used for data partitioned VOPs
520 int mpeg_quant;
521 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
522
523 /* divx specific, used to workaround (many) bugs in divx5 */
524 int divx_packed;
525 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
526 int bitstream_buffer_size;
527 unsigned int allocated_bitstream_buffer_size;
528
529 /* RV10 specific */
530 int rv10_version; ///< RV10 version: 0 or 3
531 int rv10_first_dc_coded[3];
532
533 /* MJPEG specific */
534 struct MJpegContext *mjpeg_ctx;
535
536 /* MSMPEG4 specific */
537 int mv_table_index;
538 int rl_table_index;
539 int rl_chroma_table_index;
540 int dc_table_index;
541 int use_skip_mb_code;
542 int slice_height; ///< in macroblocks
543 int first_slice_line; ///< used in mpeg4 too to handle resync markers
544 int flipflop_rounding;
545 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
546 int per_mb_rl_table;
547 int esc3_level_length;
548 int esc3_run_length;
549 /** [mb_intra][isChroma][level][run][last] */
550 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
551 int inter_intra_pred;
552 int mspel;
553
554 /* decompression specific */
555 GetBitContext gb;
556
557 /* Mpeg1 specific */
558 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
559 int last_mv_dir; ///< last mv_dir, used for b frame encoding
560 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
561
562 /* MPEG-2-specific - I wished not to have to support this mess. */
563 int progressive_sequence;
564 int mpeg_f_code[2][2];
565
566 // picture structure defines are loaded from mpegutils.h
567 int picture_structure;
568
569 int intra_dc_precision;
570 int frame_pred_frame_dct;
571 int top_field_first;
572 int concealment_motion_vectors;
573 int q_scale_type;
574 int intra_vlc_format;
575 int alternate_scan;
576 int repeat_first_field;
577 int chroma_420_type;
578 int chroma_format;
579 #define CHROMA_420 1
580 #define CHROMA_422 2
581 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
582 int chroma_y_shift;
583
584 int progressive_frame;
585 int full_pel[2];
586 int interlaced_dct;
587 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
588 int drop_frame_timecode; ///< timecode is in drop frame format.
589 int scan_offset; ///< reserve space for SVCD scan offset user data.
590
591 /* RTP specific */
592 int rtp_mode;
593
594 uint8_t *ptr_lastgob;
595 int16_t (*pblocks[12])[64];
596
597 int16_t (*block)[64]; ///< points to one of the following blocks
598 int16_t (*blocks)[12][64]; // for HQ mode we need to keep the best block
599 int (*decode_mb)(struct MpegEncContext *s, int16_t block[6][64]); // used by some codecs to avoid a switch()
600 #define SLICE_OK 0
601 #define SLICE_ERROR -1
602 #define SLICE_END -2 ///<end marker found
603 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
604
605 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
606 int16_t *block/*align 16*/, int n, int qscale);
607 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
608 int16_t *block/*align 16*/, int n, int qscale);
609 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
610 int16_t *block/*align 16*/, int n, int qscale);
611 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
612 int16_t *block/*align 16*/, int n, int qscale);
613 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
614 int16_t *block/*align 16*/, int n, int qscale);
615 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
616 int16_t *block/*align 16*/, int n, int qscale);
617 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
618 int16_t *block/*align 16*/, int n, int qscale);
619 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
620 int16_t *block/*align 16*/, int n, int qscale);
621 int (*dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
622 int (*fast_dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
623 void (*denoise_dct)(struct MpegEncContext *s, int16_t *block);
624
625 int mpv_flags; ///< flags set by private options
626 int quantizer_noise_shaping;
627
628 /**
629 * ratecontrol qmin qmax limiting method
630 * 0-> clipping, 1-> use a nice continuous function to limit qscale wthin qmin/qmax.
631 */
632 float rc_qsquish;
633 float rc_qmod_amp;
634 int rc_qmod_freq;
635 float rc_initial_cplx;
636 float rc_buffer_aggressivity;
637 float border_masking;
638 int lmin, lmax;
639
640 char *rc_eq;
641
642 /* temp buffers for rate control */
643 float *cplx_tab, *bits_tab;
644
645 /* flag to indicate a reinitialization is required, e.g. after
646 * a frame size change */
647 int context_reinit;
648
649 ERContext er;
650
651 int error_rate;
652
653 /* temporary frames used by b_frame_strategy = 2 */
654 AVFrame *tmp_frames[MAX_B_FRAMES + 2];
655 } MpegEncContext;
656
657 /* mpegvideo_enc common options */
658 #define FF_MPV_FLAG_SKIP_RD 0x0001
659 #define FF_MPV_FLAG_STRICT_GOP 0x0002
660 #define FF_MPV_FLAG_QP_RD 0x0004
661 #define FF_MPV_FLAG_CBP_RD 0x0008
662 #define FF_MPV_FLAG_NAQ 0x0010
663 #define FF_MPV_FLAG_MV0 0x0020
664
665 #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x)
666 #define FF_MPV_OPT_FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
667 #define FF_MPV_COMMON_OPTS \
668 { "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" },\
669 { "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" },\
670 { "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" },\
671 { "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" },\
672 { "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" },\
673 { "naq", "normalize adaptive quantization", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_NAQ }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
674 { "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" },\
675 { "luma_elim_threshold", "single coefficient elimination threshold for luminance (negative values also consider dc coefficient)",\
676 FF_MPV_OFFSET(luma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
677 { "chroma_elim_threshold", "single coefficient elimination threshold for chrominance (negative values also consider dc coefficient)",\
678 FF_MPV_OFFSET(chroma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
679 { "quantizer_noise_shaping", NULL, FF_MPV_OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
680 { "error_rate", "Simulate errors in the bitstream to test error concealment.", \
681 FF_MPV_OFFSET(error_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
682 {"qsquish", "how to keep quantizer between qmin and qmax (0 = clip, 1 = use differentiable function)", \
683 FF_MPV_OFFSET(rc_qsquish), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, 0, 99, FF_MPV_OPT_FLAGS}, \
684 {"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}, \
685 {"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}, \
686 {"rc_eq", "Set rate control equation. When computing the expression, besides the standard functions " \
687 "defined in the section 'Expression Evaluation', the following functions are available: " \
688 "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv " \
689 "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.", \
690 FF_MPV_OFFSET(rc_eq), AV_OPT_TYPE_STRING, .flags = FF_MPV_OPT_FLAGS }, \
691 {"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}, \
692 {"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}, \
693 {"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}, \
694 {"lmin", "minimum Lagrange factor (VBR)", FF_MPV_OFFSET(lmin), AV_OPT_TYPE_INT, {.i64 = 2*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
695 {"lmax", "maximum Lagrange factor (VBR)", FF_MPV_OFFSET(lmax), AV_OPT_TYPE_INT, {.i64 = 31*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
696
697 extern const AVOption ff_mpv_generic_options[];
698
699 #define FF_MPV_GENERIC_CLASS(name) \
700 static const AVClass name ## _class = {\
701 .class_name = #name " encoder",\
702 .item_name = av_default_item_name,\
703 .option = ff_mpv_generic_options,\
704 .version = LIBAVUTIL_VERSION_INT,\
705 };
706
707 /**
708 * Set the given MpegEncContext to common defaults (same for encoding
709 * and decoding). The changed fields will not depend upon the prior
710 * state of the MpegEncContext.
711 */
712 void ff_mpv_common_defaults(MpegEncContext *s);
713
714 int ff_mpv_common_init(MpegEncContext *s);
715 void ff_mpv_common_init_arm(MpegEncContext *s);
716 void ff_mpv_common_init_neon(MpegEncContext *s);
717 void ff_mpv_common_init_ppc(MpegEncContext *s);
718 void ff_mpv_common_init_x86(MpegEncContext *s);
719
720 int ff_mpv_common_frame_size_change(MpegEncContext *s);
721 void ff_mpv_common_end(MpegEncContext *s);
722
723 void ff_mpv_decode_defaults(MpegEncContext *s);
724 void ff_mpv_decode_mb(MpegEncContext *s, int16_t block[12][64]);
725 void ff_mpv_report_decode_progress(MpegEncContext *s);
726
727 int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx);
728 void ff_mpv_frame_end(MpegEncContext *s);
729
730 int ff_mpv_lowest_referenced_row(MpegEncContext *s, int dir);
731
732 int ff_mpv_encode_init(AVCodecContext *avctx);
733 void ff_mpv_encode_init_x86(MpegEncContext *s);
734
735 int ff_mpv_encode_end(AVCodecContext *avctx);
736 int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
737 const AVFrame *frame, int *got_packet);
738
739 void ff_clean_intra_table_entries(MpegEncContext *s);
740 void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h);
741 void ff_mpeg_flush(AVCodecContext *avctx);
742 void ff_print_debug_info(MpegEncContext *s, Picture *p);
743 void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
744
745 int ff_find_unused_picture(AVCodecContext *avctx, Picture *picture, int shared);
746 int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
747 int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
748 void ff_set_qscale(MpegEncContext * s, int qscale);
749
750 void ff_mpv_idct_init(MpegEncContext *s);
751 void ff_convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16)[2][64],
752 const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra);
753 int ff_dct_quantize_c(MpegEncContext *s, int16_t *block, int n, int qscale, int *overflow);
754
755 void ff_init_block_index(MpegEncContext *s);
756
757 void ff_mpv_motion(MpegEncContext *s,
758 uint8_t *dest_y, uint8_t *dest_cb,
759 uint8_t *dest_cr, int dir,
760 uint8_t **ref_picture,
761 op_pixels_func (*pix_op)[4],
762 qpel_mc_func (*qpix_op)[16]);
763
764 /**
765 * Allocate a Picture.
766 * The pixels are allocated/set by calling get_buffer() if shared = 0.
767 */
768 int ff_alloc_picture(MpegEncContext *s, Picture *pic, int shared);
769
770 /**
771 * permute block according to permuatation.
772 * @param last last non zero element in scantable order
773 */
774 void ff_block_permute(int16_t *block, uint8_t *permutation, const uint8_t *scantable, int last);
775
776 static inline void ff_update_block_index(MpegEncContext *s){
777 const int block_size = 8;
778
779 s->block_index[0]+=2;
780 s->block_index[1]+=2;
781 s->block_index[2]+=2;
782 s->block_index[3]+=2;
783 s->block_index[4]++;
784 s->block_index[5]++;
785 s->dest[0]+= 2*block_size;
786 s->dest[1]+= block_size;
787 s->dest[2]+= block_size;
788 }
789
790 static inline int get_bits_diff(MpegEncContext *s){
791 const int bits= put_bits_count(&s->pb);
792 const int last= s->last_bits;
793
794 s->last_bits = bits;
795
796 return bits - last;
797 }
798
799 static inline int ff_h263_round_chroma(int x){
800 static const uint8_t h263_chroma_roundtab[16] = {
801 // 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
802 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1,
803 };
804 return h263_chroma_roundtab[x & 0xf] + (x >> 3);
805 }
806
807 /* motion_est.c */
808 void ff_estimate_p_frame_motion(MpegEncContext * s,
809 int mb_x, int mb_y);
810 void ff_estimate_b_frame_motion(MpegEncContext * s,
811 int mb_x, int mb_y);
812 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
813 void ff_fix_long_p_mvs(MpegEncContext * s);
814 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
815 int16_t (*mv_table)[2], int f_code, int type, int truncate);
816 int ff_init_me(MpegEncContext *s);
817 int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
818 int ff_epzs_motion_search(MpegEncContext * s, int *mx_ptr, int *my_ptr,
819 int P[10][2], int src_index, int ref_index, int16_t (*last_mv)[2],
820 int ref_mv_scale, int size, int h);
821 int ff_get_mb_score(MpegEncContext * s, int mx, int my, int src_index,
822 int ref_index, int size, int h, int add_rate);
823
824 /* mpeg12.c */
825 extern const uint8_t ff_mpeg1_dc_scale_table[128];
826 extern const uint8_t * const ff_mpeg2_dc_scale_table[4];
827
828 void ff_mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
829 void ff_mpeg1_encode_mb(MpegEncContext *s,
830 int16_t block[8][64],
831 int motion_x, int motion_y);
832 void ff_mpeg1_encode_init(MpegEncContext *s);
833 void ff_mpeg1_encode_slice_header(MpegEncContext *s);
834
835 extern const uint8_t ff_aic_dc_scale_table[32];
836 extern const uint8_t ff_h263_chroma_qscale_table[32];
837
838 /* rv10.c */
839 int ff_rv10_encode_picture_header(MpegEncContext *s, int picture_number);
840 int ff_rv_decode_dc(MpegEncContext *s, int n);
841 void ff_rv20_encode_picture_header(MpegEncContext *s, int picture_number);
842
843 int ff_mpeg_ref_picture(AVCodecContext *avctx, Picture *dst, Picture *src);
844 void ff_mpeg_unref_picture(AVCodecContext *avctx, Picture *picture);
845 void ff_free_picture_tables(Picture *pic);
846
847 #endif /* AVCODEC_MPEGVIDEO_H */