Imported Debian version 2.5.3~trusty1
[deb_ffmpeg.git] / ffmpeg / 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 FFmpeg.
7 *
8 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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 "h264chroma.h"
39 #include "h263dsp.h"
40 #include "hpeldsp.h"
41 #include "idctdsp.h"
42 #include "me_cmp.h"
43 #include "mpegvideodsp.h"
44 #include "mpegvideoencdsp.h"
45 #include "pixblockdsp.h"
46 #include "put_bits.h"
47 #include "ratecontrol.h"
48 #include "parser.h"
49 #include "mpeg12data.h"
50 #include "qpeldsp.h"
51 #include "rl.h"
52 #include "thread.h"
53 #include "videodsp.h"
54
55 #include "libavutil/opt.h"
56 #include "libavutil/timecode.h"
57
58 #define FRAME_SKIPPED 100 ///< return value for header parsers if frame is not coded
59
60 enum OutputFormat {
61 FMT_MPEG1,
62 FMT_H261,
63 FMT_H263,
64 FMT_MJPEG,
65 };
66
67 #define MAX_FCODE 7
68 #define MAX_MV 4096
69
70 #define MAX_THREADS 32
71 #define MAX_PICTURE_COUNT 36
72
73 #define MAX_B_FRAMES 16
74
75 #define ME_MAP_SIZE 64
76
77 #define MAX_MB_BYTES (30*16*16*3/8 + 120)
78
79 #define INPLACE_OFFSET 16
80
81 #define EDGE_WIDTH 16
82
83 /* Start codes. */
84 #define SEQ_END_CODE 0x000001b7
85 #define SEQ_START_CODE 0x000001b3
86 #define GOP_START_CODE 0x000001b8
87 #define PICTURE_START_CODE 0x00000100
88 #define SLICE_MIN_START_CODE 0x00000101
89 #define SLICE_MAX_START_CODE 0x000001af
90 #define EXT_START_CODE 0x000001b5
91 #define USER_START_CODE 0x000001b2
92
93 /* encoding scans */
94 extern const uint8_t ff_alternate_horizontal_scan[64];
95 extern const uint8_t ff_alternate_vertical_scan[64];
96
97 struct MpegEncContext;
98
99 /**
100 * Picture.
101 */
102 typedef struct Picture{
103 struct AVFrame *f;
104 ThreadFrame tf;
105
106 AVBufferRef *qscale_table_buf;
107 int8_t *qscale_table;
108
109 AVBufferRef *motion_val_buf[2];
110 int16_t (*motion_val[2])[2];
111
112 AVBufferRef *mb_type_buf;
113 uint32_t *mb_type; ///< types and macros are defined in mpegutils.h
114
115 AVBufferRef *mbskip_table_buf;
116 uint8_t *mbskip_table;
117
118 AVBufferRef *ref_index_buf[2];
119 int8_t *ref_index[2];
120
121 AVBufferRef *mb_var_buf;
122 uint16_t *mb_var; ///< Table for MB variances
123
124 AVBufferRef *mc_mb_var_buf;
125 uint16_t *mc_mb_var; ///< Table for motion compensated MB variances
126
127 int alloc_mb_width; ///< mb_width used to allocate tables
128 int alloc_mb_height; ///< mb_height used to allocate tables
129
130 AVBufferRef *mb_mean_buf;
131 uint8_t *mb_mean; ///< Table for MB luminance
132
133 AVBufferRef *hwaccel_priv_buf;
134 /**
135 * hardware accelerator private data
136 */
137 void *hwaccel_picture_private;
138
139 int field_picture; ///< whether or not the picture was encoded in separate fields
140
141 int64_t mb_var_sum; ///< sum of MB variance for current frame
142 int64_t mc_mb_var_sum; ///< motion compensated MB variance for current frame
143
144 int b_frame_score;
145 int needs_realloc; ///< Picture needs to be reallocated (eg due to a frame size change)
146
147 int reference;
148 int shared;
149
150 uint64_t error[AV_NUM_DATA_POINTERS];
151 } Picture;
152
153 /**
154 * Motion estimation context.
155 */
156 typedef struct MotionEstContext{
157 AVCodecContext *avctx;
158 int skip; ///< set if ME is skipped for the current MB
159 int co_located_mv[4][2]; ///< mv from last P-frame for direct mode ME
160 int direct_basis_mv[4][2];
161 uint8_t *scratchpad; ///< data area for the ME algo, so that the ME does not need to malloc/free
162 uint8_t *best_mb;
163 uint8_t *temp_mb[2];
164 uint8_t *temp;
165 int best_bits;
166 uint32_t *map; ///< map to avoid duplicate evaluations
167 uint32_t *score_map; ///< map to store the scores
168 unsigned map_generation;
169 int pre_penalty_factor;
170 int penalty_factor; /**< an estimate of the bits required to
171 code a given mv value, e.g. (1,0) takes
172 more bits than (0,0). We have to
173 estimate whether any reduction in
174 residual is worth the extra bits. */
175 int sub_penalty_factor;
176 int mb_penalty_factor;
177 int flags;
178 int sub_flags;
179 int mb_flags;
180 int pre_pass; ///< = 1 for the pre pass
181 int dia_size;
182 int xmin;
183 int xmax;
184 int ymin;
185 int ymax;
186 int pred_x;
187 int pred_y;
188 uint8_t *src[4][4];
189 uint8_t *ref[4][4];
190 int stride;
191 int uvstride;
192 /* temp variables for picture complexity calculation */
193 int64_t mc_mb_var_sum_temp;
194 int64_t mb_var_sum_temp;
195 int scene_change_score;
196 /* cmp, chroma_cmp;*/
197 op_pixels_func (*hpel_put)[4];
198 op_pixels_func (*hpel_avg)[4];
199 qpel_mc_func (*qpel_put)[16];
200 qpel_mc_func (*qpel_avg)[16];
201 uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
202 uint8_t *current_mv_penalty;
203 int (*sub_motion_search)(struct MpegEncContext * s,
204 int *mx_ptr, int *my_ptr, int dmin,
205 int src_index, int ref_index,
206 int size, int h);
207 }MotionEstContext;
208
209 /**
210 * MpegEncContext.
211 */
212 typedef struct MpegEncContext {
213 AVClass *class;
214
215 int y_dc_scale, c_dc_scale;
216 int ac_pred;
217 int block_last_index[12]; ///< last non zero coefficient in block
218 int h263_aic; ///< Advanded INTRA Coding (AIC)
219
220 /* scantables */
221 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
222 ScanTable intra_scantable;
223 ScanTable intra_h_scantable;
224 ScanTable intra_v_scantable;
225
226 /* WARNING: changes above this line require updates to hardcoded
227 * offsets used in asm. */
228
229 struct AVCodecContext *avctx;
230 /* the following parameters must be initialized before encoding */
231 int width, height;///< picture size. must be a multiple of 16
232 int gop_size;
233 int intra_only; ///< if true, only intra pictures are generated
234 int bit_rate; ///< wanted bit rate
235 enum OutputFormat out_format; ///< output format
236 int h263_pred; ///< use mpeg4/h263 ac/dc predictions
237 int pb_frame; ///< PB frame mode (0 = none, 1 = base, 2 = improved)
238
239 /* the following codec id fields are deprecated in favor of codec_id */
240 int h263_plus; ///< h263 plus headers
241 int h263_flv; ///< use flv h263 header
242
243 enum AVCodecID codec_id; /* see AV_CODEC_ID_xxx */
244 int fixed_qscale; ///< fixed qscale if non zero
245 int encoding; ///< true if we are encoding (vs decoding)
246 int flags; ///< AVCodecContext.flags (HQ, MV4, ...)
247 int flags2; ///< AVCodecContext.flags2
248 int max_b_frames; ///< max number of b-frames for encoding
249 int luma_elim_threshold;
250 int chroma_elim_threshold;
251 int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
252 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
253 int codec_tag; ///< internal codec_tag upper case converted from avctx codec_tag
254 int stream_codec_tag; ///< internal stream_codec_tag upper case converted from avctx stream_codec_tag
255 /* the following fields are managed internally by the encoder */
256
257 /* sequence parameters */
258 int context_initialized;
259 int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
260 int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
261 int picture_number; //FIXME remove, unclear definition
262 int picture_in_gop_number; ///< 0-> first pic in gop, ...
263 int mb_width, mb_height; ///< number of MBs horizontally & vertically
264 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
265 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
266 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
267 int mb_num; ///< number of MBs of a picture
268 ptrdiff_t linesize; ///< line size, in bytes, may be different from width
269 ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
270 Picture *picture; ///< main picture buffer
271 Picture **input_picture; ///< next pictures on display order for encoding
272 Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
273
274 int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_encode_video2()
275 /**
276 * pts difference between the first and second input frame, used for
277 * calculating dts of the first frame when there's a delay */
278 int64_t dts_delta;
279 /**
280 * reordered pts to be used as dts for the next output frame when there's
281 * a delay */
282 int64_t reordered_pts;
283
284 /** bit output */
285 PutBitContext pb;
286
287 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
288 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
289 struct MpegEncContext *thread_context[MAX_THREADS];
290 int slice_context_count; ///< number of used thread_contexts
291
292 /**
293 * copy of the previous picture structure.
294 * note, linesize & data, might not match the previous picture (for field pictures)
295 */
296 Picture last_picture;
297
298 /**
299 * copy of the next picture structure.
300 * note, linesize & data, might not match the next picture (for field pictures)
301 */
302 Picture next_picture;
303
304 /**
305 * copy of the source picture structure for encoding.
306 * note, linesize & data, might not match the source picture (for field pictures)
307 */
308 Picture new_picture;
309
310 /**
311 * copy of the current picture structure.
312 * note, linesize & data, might not match the current picture (for field pictures)
313 */
314 Picture current_picture; ///< buffer to store the decompressed current picture
315
316 Picture *last_picture_ptr; ///< pointer to the previous picture.
317 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
318 Picture *current_picture_ptr; ///< pointer to the current picture
319 int last_dc[3]; ///< last DC values for MPEG1
320 int16_t *dc_val_base;
321 int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
322 const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
323 const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
324 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
325 uint8_t *coded_block_base;
326 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
327 int16_t (*ac_val_base)[16];
328 int16_t (*ac_val[3])[16]; ///< used for mpeg4 AC prediction, all 3 arrays must be continuous
329 int mb_skipped; ///< MUST BE SET only during DECODING
330 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
331 and used for b-frame encoding & decoding (contains skip table of next P Frame) */
332 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
333 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
334 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
335 uint8_t *edge_emu_buffer; ///< temporary buffer for if MVs point to out-of-frame data
336 uint8_t *rd_scratchpad; ///< scratchpad for rate distortion mb decision
337 uint8_t *obmc_scratchpad;
338 uint8_t *b_scratchpad; ///< scratchpad used for writing into write only buffers
339
340 int qscale; ///< QP
341 int chroma_qscale; ///< chroma QP
342 unsigned int lambda; ///< lagrange multipler used in rate distortion
343 unsigned int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
344 int *lambda_table;
345 int adaptive_quant; ///< use adaptive quantization
346 int dquant; ///< qscale difference to prev qscale
347 int closed_gop; ///< MPEG1/2 GOP is closed
348 int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ...
349 int vbv_delay;
350 int last_pict_type; //FIXME removes
351 int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
352 int droppable;
353 int frame_rate_index;
354 AVRational mpeg2_frame_rate_ext;
355 int last_lambda_for[5]; ///< last lambda for a specific pict type
356 int skipdct; ///< skip dct and code zero residual
357
358 /* motion compensation */
359 int unrestricted_mv; ///< mv can point outside of the coded picture
360 int h263_long_vectors; ///< use horrible h263v1 long vector mode
361
362 BlockDSPContext bdsp;
363 FDCTDSPContext fdsp;
364 H264ChromaContext h264chroma;
365 HpelDSPContext hdsp;
366 IDCTDSPContext idsp;
367 MECmpContext mecc;
368 MpegVideoDSPContext mdsp;
369 MpegvideoEncDSPContext mpvencdsp;
370 PixblockDSPContext pdsp;
371 QpelDSPContext qdsp;
372 VideoDSPContext vdsp;
373 H263DSPContext h263dsp;
374 int f_code; ///< forward MV resolution
375 int b_code; ///< backward MV resolution for B Frames (mpeg4)
376 int16_t (*p_mv_table_base)[2];
377 int16_t (*b_forw_mv_table_base)[2];
378 int16_t (*b_back_mv_table_base)[2];
379 int16_t (*b_bidir_forw_mv_table_base)[2];
380 int16_t (*b_bidir_back_mv_table_base)[2];
381 int16_t (*b_direct_mv_table_base)[2];
382 int16_t (*p_field_mv_table_base[2][2])[2];
383 int16_t (*b_field_mv_table_base[2][2][2])[2];
384 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
385 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
386 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
387 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
388 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
389 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
390 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
391 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
392 uint8_t (*p_field_select_table[2]);
393 uint8_t (*b_field_select_table[2][2]);
394 int me_method; ///< ME algorithm
395 int mv_dir;
396 #define MV_DIR_FORWARD 1
397 #define MV_DIR_BACKWARD 2
398 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
399 int mv_type;
400 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
401 #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
402 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
403 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
404 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
405 /**motion vectors for a macroblock
406 first coordinate : 0 = forward 1 = backward
407 second " : depend on type
408 third " : 0 = x, 1 = y
409 */
410 int mv[2][4][2];
411 int field_select[2][2];
412 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
413 uint8_t *fcode_tab; ///< smallest fcode needed for each MV
414 int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
415
416 MotionEstContext me;
417
418 int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
419 for b-frames rounding mode is always 0 */
420
421 /* macroblock layer */
422 int mb_x, mb_y;
423 int mb_skip_run;
424 int mb_intra;
425 uint16_t *mb_type; ///< Table for candidate MB types for encoding (defines in mpegutils.h)
426
427 int block_index[6]; ///< index to current MB in block based arrays with edges
428 int block_wrap[6];
429 uint8_t *dest[3];
430
431 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
432
433 /** matrix transmitted in the bitstream */
434 uint16_t intra_matrix[64];
435 uint16_t chroma_intra_matrix[64];
436 uint16_t inter_matrix[64];
437 uint16_t chroma_inter_matrix[64];
438
439 int intra_quant_bias; ///< bias for the quantizer
440 int inter_quant_bias; ///< bias for the quantizer
441 int min_qcoeff; ///< minimum encodable coefficient
442 int max_qcoeff; ///< maximum encodable coefficient
443 int ac_esc_length; ///< num of bits needed to encode the longest esc
444 uint8_t *intra_ac_vlc_length;
445 uint8_t *intra_ac_vlc_last_length;
446 uint8_t *inter_ac_vlc_length;
447 uint8_t *inter_ac_vlc_last_length;
448 uint8_t *luma_dc_vlc_length;
449 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
450
451 int coded_score[12];
452
453 /** precomputed matrix (combine qscale and DCT renorm) */
454 int (*q_intra_matrix)[64];
455 int (*q_chroma_intra_matrix)[64];
456 int (*q_inter_matrix)[64];
457 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
458 uint16_t (*q_intra_matrix16)[2][64];
459 uint16_t (*q_chroma_intra_matrix16)[2][64];
460 uint16_t (*q_inter_matrix16)[2][64];
461
462 /* noise reduction */
463 int (*dct_error_sum)[64];
464 int dct_count[2];
465 uint16_t (*dct_offset)[64];
466
467 /* bit rate control */
468 int64_t total_bits;
469 int frame_bits; ///< bits used for the current frame
470 int stuffing_bits; ///< bits used for stuffing
471 int next_lambda; ///< next lambda used for retrying to encode a frame
472 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
473
474 /* statistics, used for 2-pass encoding */
475 int mv_bits;
476 int header_bits;
477 int i_tex_bits;
478 int p_tex_bits;
479 int i_count;
480 int f_count;
481 int b_count;
482 int skip_count;
483 int misc_bits; ///< cbp, mb_type
484 int last_bits; ///< temp var used for calculating the above vars
485
486 /* error concealment / resync */
487 int resync_mb_x; ///< x position of last resync marker
488 int resync_mb_y; ///< y position of last resync marker
489 GetBitContext last_resync_gb; ///< used to search for the next resync marker
490 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
491 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
492 int err_recognition;
493
494 ParseContext parse_context;
495
496 /* H.263 specific */
497 int gob_index;
498 int obmc; ///< overlapped block motion compensation
499 int mb_info; ///< interval for outputting info about mb offsets as side data
500 int prev_mb_info, last_mb_info;
501 uint8_t *mb_info_ptr;
502 int mb_info_size;
503 int ehc_mode;
504
505 /* H.263+ specific */
506 int umvplus; ///< == H263+ && unrestricted_mv
507 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
508 int h263_slice_structured;
509 int alt_inter_vlc; ///< alternative inter vlc
510 int modified_quant;
511 int loop_filter;
512 int custom_pcf;
513
514 /* mpeg4 specific */
515 ///< number of bits to represent the fractional part of time (encoder only)
516 int time_increment_bits;
517 int last_time_base;
518 int time_base; ///< time in seconds of last I,P,S Frame
519 int64_t time; ///< time of current frame
520 int64_t last_non_b_time;
521 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
522 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
523 uint16_t pp_field_time;
524 uint16_t pb_field_time; ///< like above, just for interlaced
525 int real_sprite_warping_points;
526 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
527 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
528 int mcsel;
529 int quant_precision;
530 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
531 int aspect_ratio_info; //FIXME remove
532 int sprite_warping_accuracy;
533 int data_partitioning; ///< data partitioning flag from header
534 int partitioned_frame; ///< is current frame partitioned
535 int low_delay; ///< no reordering needed / has no b-frames
536 int vo_type;
537 PutBitContext tex_pb; ///< used for data partitioned VOPs
538 PutBitContext pb2; ///< used for data partitioned VOPs
539 int mpeg_quant;
540 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
541
542 /* divx specific, used to workaround (many) bugs in divx5 */
543 int divx_packed;
544 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
545 int bitstream_buffer_size;
546 unsigned int allocated_bitstream_buffer_size;
547
548 /* RV10 specific */
549 int rv10_version; ///< RV10 version: 0 or 3
550 int rv10_first_dc_coded[3];
551
552 /* MJPEG specific */
553 struct MJpegContext *mjpeg_ctx;
554 int esc_pos;
555
556 /* MSMPEG4 specific */
557 int mv_table_index;
558 int rl_table_index;
559 int rl_chroma_table_index;
560 int dc_table_index;
561 int use_skip_mb_code;
562 int slice_height; ///< in macroblocks
563 int first_slice_line; ///< used in mpeg4 too to handle resync markers
564 int flipflop_rounding;
565 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
566 int per_mb_rl_table;
567 int esc3_level_length;
568 int esc3_run_length;
569 /** [mb_intra][isChroma][level][run][last] */
570 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
571 int inter_intra_pred;
572 int mspel;
573
574 /* decompression specific */
575 GetBitContext gb;
576
577 /* Mpeg1 specific */
578 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
579 int last_mv_dir; ///< last mv_dir, used for b frame encoding
580 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
581
582 /* MPEG-2-specific - I wished not to have to support this mess. */
583 int progressive_sequence;
584 int mpeg_f_code[2][2];
585
586 // picture structure defines are loaded from mpegutils.h
587 int picture_structure;
588
589 int intra_dc_precision;
590 int frame_pred_frame_dct;
591 int top_field_first;
592 int concealment_motion_vectors;
593 int q_scale_type;
594 int intra_vlc_format;
595 int alternate_scan;
596 int seq_disp_ext;
597 int repeat_first_field;
598 int chroma_420_type;
599 int chroma_format;
600 #define CHROMA_420 1
601 #define CHROMA_422 2
602 #define CHROMA_444 3
603 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
604 int chroma_y_shift;
605
606 int progressive_frame;
607 int full_pel[2];
608 int interlaced_dct;
609 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
610 int drop_frame_timecode; ///< timecode is in drop frame format.
611 int scan_offset; ///< reserve space for SVCD scan offset user data.
612
613 /* RTP specific */
614 int rtp_mode;
615
616 char *tc_opt_str; ///< timecode option string
617 AVTimecode tc; ///< timecode context
618
619 uint8_t *ptr_lastgob;
620 int swap_uv; //vcr2 codec is an MPEG-2 variant with U and V swapped
621 int pack_pblocks; //xvmc needs to keep blocks without gaps.
622 int16_t (*pblocks[12])[64];
623
624 int16_t (*block)[64]; ///< points to one of the following blocks
625 int16_t (*blocks)[12][64]; // for HQ mode we need to keep the best block
626 int (*decode_mb)(struct MpegEncContext *s, int16_t block[6][64]); // used by some codecs to avoid a switch()
627 #define SLICE_OK 0
628 #define SLICE_ERROR -1
629 #define SLICE_END -2 ///<end marker found
630 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
631
632 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
633 int16_t *block/*align 16*/, int n, int qscale);
634 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
635 int16_t *block/*align 16*/, int n, int qscale);
636 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
637 int16_t *block/*align 16*/, int n, int qscale);
638 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
639 int16_t *block/*align 16*/, int n, int qscale);
640 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
641 int16_t *block/*align 16*/, int n, int qscale);
642 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
643 int16_t *block/*align 16*/, int n, int qscale);
644 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
645 int16_t *block/*align 16*/, int n, int qscale);
646 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
647 int16_t *block/*align 16*/, int n, int qscale);
648 int (*dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
649 int (*fast_dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
650 void (*denoise_dct)(struct MpegEncContext *s, int16_t *block);
651
652 int mpv_flags; ///< flags set by private options
653 int quantizer_noise_shaping;
654
655 /**
656 * ratecontrol qmin qmax limiting method
657 * 0-> clipping, 1-> use a nice continuous function to limit qscale within qmin/qmax.
658 */
659 float rc_qsquish;
660 float rc_qmod_amp;
661 int rc_qmod_freq;
662 float rc_initial_cplx;
663 float rc_buffer_aggressivity;
664 float border_masking;
665 int lmin, lmax;
666
667 char *rc_eq;
668
669 /* temp buffers for rate control */
670 float *cplx_tab, *bits_tab;
671
672 /* flag to indicate a reinitialization is required, e.g. after
673 * a frame size change */
674 int context_reinit;
675
676 ERContext er;
677
678 int error_rate;
679
680 /* temporary frames used by b_frame_strategy = 2 */
681 AVFrame *tmp_frames[MAX_B_FRAMES + 2];
682 } MpegEncContext;
683
684 #define REBASE_PICTURE(pic, new_ctx, old_ctx) \
685 ((pic && pic >= old_ctx->picture && \
686 pic < old_ctx->picture + MAX_PICTURE_COUNT) ? \
687 &new_ctx->picture[pic - old_ctx->picture] : NULL)
688
689 /* mpegvideo_enc common options */
690 #define FF_MPV_FLAG_SKIP_RD 0x0001
691 #define FF_MPV_FLAG_STRICT_GOP 0x0002
692 #define FF_MPV_FLAG_QP_RD 0x0004
693 #define FF_MPV_FLAG_CBP_RD 0x0008
694 #define FF_MPV_FLAG_NAQ 0x0010
695 #define FF_MPV_FLAG_MV0 0x0020
696
697 #ifndef FF_MPV_OFFSET
698 #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x)
699 #endif
700 #define FF_MPV_OPT_FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
701 #define FF_MPV_COMMON_OPTS \
702 { "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" },\
703 { "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" },\
704 { "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" },\
705 { "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" },\
706 { "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" },\
707 { "naq", "normalize adaptive quantization", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_NAQ }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
708 { "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" },\
709 { "luma_elim_threshold", "single coefficient elimination threshold for luminance (negative values also consider dc coefficient)",\
710 FF_MPV_OFFSET(luma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
711 { "chroma_elim_threshold", "single coefficient elimination threshold for chrominance (negative values also consider dc coefficient)",\
712 FF_MPV_OFFSET(chroma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
713 { "quantizer_noise_shaping", NULL, FF_MPV_OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
714 { "error_rate", "Simulate errors in the bitstream to test error concealment.", \
715 FF_MPV_OFFSET(error_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
716 {"qsquish", "how to keep quantizer between qmin and qmax (0 = clip, 1 = use differentiable function)", \
717 FF_MPV_OFFSET(rc_qsquish), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, 0, 99, FF_MPV_OPT_FLAGS}, \
718 {"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}, \
719 {"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}, \
720 {"rc_eq", "Set rate control equation. When computing the expression, besides the standard functions " \
721 "defined in the section 'Expression Evaluation', the following functions are available: " \
722 "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv " \
723 "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.", \
724 FF_MPV_OFFSET(rc_eq), AV_OPT_TYPE_STRING, .flags = FF_MPV_OPT_FLAGS }, \
725 {"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}, \
726 {"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}, \
727 {"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}, \
728 {"lmin", "minimum Lagrange factor (VBR)", FF_MPV_OFFSET(lmin), AV_OPT_TYPE_INT, {.i64 = 2*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
729 {"lmax", "maximum Lagrange factor (VBR)", FF_MPV_OFFSET(lmax), AV_OPT_TYPE_INT, {.i64 = 31*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
730
731 extern const AVOption ff_mpv_generic_options[];
732
733 #define FF_MPV_GENERIC_CLASS(name) \
734 static const AVClass name ## _class = {\
735 .class_name = #name " encoder",\
736 .item_name = av_default_item_name,\
737 .option = ff_mpv_generic_options,\
738 .version = LIBAVUTIL_VERSION_INT,\
739 };
740
741 /**
742 * Set the given MpegEncContext to common defaults (same for encoding
743 * and decoding). The changed fields will not depend upon the prior
744 * state of the MpegEncContext.
745 */
746 void ff_mpv_common_defaults(MpegEncContext *s);
747
748 void ff_dct_encode_init_x86(MpegEncContext *s);
749
750 int ff_mpv_common_init(MpegEncContext *s);
751 void ff_mpv_common_init_arm(MpegEncContext *s);
752 void ff_mpv_common_init_axp(MpegEncContext *s);
753 void ff_mpv_common_init_neon(MpegEncContext *s);
754 void ff_mpv_common_init_ppc(MpegEncContext *s);
755 void ff_mpv_common_init_x86(MpegEncContext *s);
756
757 int ff_mpv_common_frame_size_change(MpegEncContext *s);
758 void ff_mpv_common_end(MpegEncContext *s);
759
760 void ff_mpv_decode_defaults(MpegEncContext *s);
761 void ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx);
762 void ff_mpv_decode_mb(MpegEncContext *s, int16_t block[12][64]);
763 void ff_mpv_report_decode_progress(MpegEncContext *s);
764
765 int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx);
766 void ff_mpv_frame_end(MpegEncContext *s);
767
768 int ff_mpv_lowest_referenced_row(MpegEncContext *s, int dir);
769
770 int ff_mpv_encode_init(AVCodecContext *avctx);
771 void ff_mpv_encode_init_x86(MpegEncContext *s);
772
773 int ff_mpv_encode_end(AVCodecContext *avctx);
774 int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
775 const AVFrame *frame, int *got_packet);
776
777 void ff_clean_intra_table_entries(MpegEncContext *s);
778 void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h);
779 void ff_mpeg_flush(AVCodecContext *avctx);
780
781 void ff_print_debug_info(MpegEncContext *s, Picture *p, AVFrame *pict);
782 void ff_print_debug_info2(AVCodecContext *avctx, AVFrame *pict, uint8_t *mbskip_table,
783 uint32_t *mbtype_table, int8_t *qscale_table, int16_t (*motion_val[2])[2],
784 int *low_delay,
785 int mb_width, int mb_height, int mb_stride, int quarter_sample);
786
787 int ff_mpv_export_qp_table(MpegEncContext *s, AVFrame *f, Picture *p, int qp_type);
788
789 void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
790 int ff_find_unused_picture(MpegEncContext *s, int shared);
791 void ff_denoise_dct(MpegEncContext *s, int16_t *block);
792 int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
793 int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
794 void ff_set_qscale(MpegEncContext * s, int qscale);
795
796 void ff_mpv_idct_init(MpegEncContext *s);
797 int ff_dct_encode_init(MpegEncContext *s);
798 void ff_convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16)[2][64],
799 const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra);
800 int ff_dct_quantize_c(MpegEncContext *s, int16_t *block, int n, int qscale, int *overflow);
801
802 void ff_init_block_index(MpegEncContext *s);
803
804 void ff_mpv_motion(MpegEncContext *s,
805 uint8_t *dest_y, uint8_t *dest_cb,
806 uint8_t *dest_cr, int dir,
807 uint8_t **ref_picture,
808 op_pixels_func (*pix_op)[4],
809 qpel_mc_func (*qpix_op)[16]);
810
811 /**
812 * Allocate a Picture.
813 * The pixels are allocated/set by calling get_buffer() if shared = 0.
814 */
815 int ff_alloc_picture(MpegEncContext *s, Picture *pic, int shared);
816
817 /**
818 * permute block according to permuatation.
819 * @param last last non zero element in scantable order
820 */
821 void ff_block_permute(int16_t *block, uint8_t *permutation, const uint8_t *scantable, int last);
822
823 static inline void ff_update_block_index(MpegEncContext *s){
824 const int block_size= 8 >> s->avctx->lowres;
825
826 s->block_index[0]+=2;
827 s->block_index[1]+=2;
828 s->block_index[2]+=2;
829 s->block_index[3]+=2;
830 s->block_index[4]++;
831 s->block_index[5]++;
832 s->dest[0]+= 2*block_size;
833 s->dest[1]+= block_size;
834 s->dest[2]+= block_size;
835 }
836
837 static inline int get_bits_diff(MpegEncContext *s){
838 const int bits= put_bits_count(&s->pb);
839 const int last= s->last_bits;
840
841 s->last_bits = bits;
842
843 return bits - last;
844 }
845
846 static inline int ff_h263_round_chroma(int x){
847 static const uint8_t h263_chroma_roundtab[16] = {
848 // 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
849 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1,
850 };
851 return h263_chroma_roundtab[x & 0xf] + (x >> 3);
852 }
853
854 /* motion_est.c */
855 void ff_estimate_p_frame_motion(MpegEncContext * s,
856 int mb_x, int mb_y);
857 void ff_estimate_b_frame_motion(MpegEncContext * s,
858 int mb_x, int mb_y);
859 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
860 void ff_fix_long_p_mvs(MpegEncContext * s);
861 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
862 int16_t (*mv_table)[2], int f_code, int type, int truncate);
863 int ff_init_me(MpegEncContext *s);
864 int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
865 int ff_epzs_motion_search(MpegEncContext * s, int *mx_ptr, int *my_ptr,
866 int P[10][2], int src_index, int ref_index, int16_t (*last_mv)[2],
867 int ref_mv_scale, int size, int h);
868 int ff_get_mb_score(MpegEncContext * s, int mx, int my, int src_index,
869 int ref_index, int size, int h, int add_rate);
870
871 /* mpeg12.c */
872 extern const uint8_t ff_mpeg1_dc_scale_table[128];
873 extern const uint8_t * const ff_mpeg2_dc_scale_table[4];
874
875 void ff_mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
876 void ff_mpeg1_encode_mb(MpegEncContext *s,
877 int16_t block[8][64],
878 int motion_x, int motion_y);
879 void ff_mpeg1_encode_init(MpegEncContext *s);
880 void ff_mpeg1_encode_slice_header(MpegEncContext *s);
881
882 extern const uint8_t ff_aic_dc_scale_table[32];
883 extern const uint8_t ff_h263_chroma_qscale_table[32];
884
885 /* rv10.c */
886 void ff_rv10_encode_picture_header(MpegEncContext *s, int picture_number);
887 int ff_rv_decode_dc(MpegEncContext *s, int n);
888 void ff_rv20_encode_picture_header(MpegEncContext *s, int picture_number);
889
890
891 /* msmpeg4.c */
892 void ff_msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number);
893 void ff_msmpeg4_encode_ext_header(MpegEncContext * s);
894 void ff_msmpeg4_encode_mb(MpegEncContext * s,
895 int16_t block[6][64],
896 int motion_x, int motion_y);
897 int ff_msmpeg4_decode_picture_header(MpegEncContext * s);
898 int ff_msmpeg4_decode_ext_header(MpegEncContext * s, int buf_size);
899 int ff_msmpeg4_decode_init(AVCodecContext *avctx);
900 void ff_msmpeg4_encode_init(MpegEncContext *s);
901 int ff_wmv2_decode_picture_header(MpegEncContext * s);
902 int ff_wmv2_decode_secondary_picture_header(MpegEncContext * s);
903 void ff_wmv2_add_mb(MpegEncContext *s, int16_t block[6][64], uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr);
904 void ff_mspel_motion(MpegEncContext *s,
905 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
906 uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
907 int motion_x, int motion_y, int h);
908 int ff_wmv2_encode_picture_header(MpegEncContext * s, int picture_number);
909 void ff_wmv2_encode_mb(MpegEncContext * s,
910 int16_t block[6][64],
911 int motion_x, int motion_y);
912
913 int ff_mpeg_ref_picture(MpegEncContext *s, Picture *dst, Picture *src);
914 void ff_mpeg_unref_picture(MpegEncContext *s, Picture *picture);
915 void ff_free_picture_tables(Picture *pic);
916
917
918 #endif /* AVCODEC_MPEGVIDEO_H */