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 "internal.h"
43 #include "me_cmp.h"
44 #include "motion_est.h"
45 #include "mpegpicture.h"
46 #include "mpegvideodsp.h"
47 #include "mpegvideoencdsp.h"
48 #include "mpegvideodata.h"
49 #include "pixblockdsp.h"
50 #include "put_bits.h"
51 #include "ratecontrol.h"
52 #include "parser.h"
53 #include "mpegutils.h"
54 #include "mpeg12data.h"
55 #include "qpeldsp.h"
56 #include "thread.h"
57 #include "videodsp.h"
58
59 #include "libavutil/opt.h"
60 #include "libavutil/timecode.h"
61
62 #define MAX_THREADS 32
63
64 #define MAX_B_FRAMES 16
65
66 /* Start codes. */
67 #define SEQ_END_CODE 0x000001b7
68 #define SEQ_START_CODE 0x000001b3
69 #define GOP_START_CODE 0x000001b8
70 #define PICTURE_START_CODE 0x00000100
71 #define SLICE_MIN_START_CODE 0x00000101
72 #define SLICE_MAX_START_CODE 0x000001af
73 #define EXT_START_CODE 0x000001b5
74 #define USER_START_CODE 0x000001b2
75 #define SLICE_START_CODE 0x000001b7
76
77
78 /**
79 * MpegEncContext.
80 */
81 typedef struct MpegEncContext {
82 AVClass *class;
83
84 int y_dc_scale, c_dc_scale;
85 int ac_pred;
86 int block_last_index[12]; ///< last non zero coefficient in block
87 int h263_aic; ///< Advanced INTRA Coding (AIC)
88
89 /* scantables */
90 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce the cache usage
91 ScanTable intra_scantable;
92 ScanTable intra_h_scantable;
93 ScanTable intra_v_scantable;
94
95 /* WARNING: changes above this line require updates to hardcoded
96 * offsets used in ASM. */
97
98 struct AVCodecContext *avctx;
99 /* the following parameters must be initialized before encoding */
100 int width, height;///< picture size. must be a multiple of 16
101 int gop_size;
102 int intra_only; ///< if true, only intra pictures are generated
103 int64_t bit_rate; ///< wanted bit rate
104 enum OutputFormat out_format; ///< output format
105 int h263_pred; ///< use MPEG-4/H.263 ac/dc predictions
106 int pb_frame; ///< PB-frame mode (0 = none, 1 = base, 2 = improved)
107
108 /* the following codec id fields are deprecated in favor of codec_id */
109 int h263_plus; ///< H.263+ headers
110 int h263_flv; ///< use flv H.263 header
111
112 enum AVCodecID codec_id; /* see AV_CODEC_ID_xxx */
113 int fixed_qscale; ///< fixed qscale if non zero
114 int encoding; ///< true if we are encoding (vs decoding)
115 int max_b_frames; ///< max number of B-frames for encoding
116 int luma_elim_threshold;
117 int chroma_elim_threshold;
118 int strict_std_compliance; ///< strictly follow the std (MPEG-4, ...)
119 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
120 int codec_tag; ///< internal codec_tag upper case converted from avctx codec_tag
121 /* the following fields are managed internally by the encoder */
122
123 /* sequence parameters */
124 int context_initialized;
125 int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
126 int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
127 int picture_number; //FIXME remove, unclear definition
128 int picture_in_gop_number; ///< 0-> first pic in gop, ...
129 int mb_width, mb_height; ///< number of MBs horizontally & vertically
130 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
131 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
132 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
133 int mb_num; ///< number of MBs of a picture
134 ptrdiff_t linesize; ///< line size, in bytes, may be different from width
135 ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
136 Picture *picture; ///< main picture buffer
137 Picture **input_picture; ///< next pictures on display order for encoding
138 Picture **reordered_input_picture; ///< pointer to the next pictures in coded order for encoding
139
140 int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_encode_video2()
141 /**
142 * pts difference between the first and second input frame, used for
143 * calculating dts of the first frame when there's a delay */
144 int64_t dts_delta;
145 /**
146 * reordered pts to be used as dts for the next output frame when there's
147 * a delay */
148 int64_t reordered_pts;
149
150 /** bit output */
151 PutBitContext pb;
152
153 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
154 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
155 struct MpegEncContext *thread_context[MAX_THREADS];
156 int slice_context_count; ///< number of used thread_contexts
157
158 /**
159 * copy of the previous picture structure.
160 * note, linesize & data, might not match the previous picture (for field pictures)
161 */
162 Picture last_picture;
163
164 /**
165 * copy of the next picture structure.
166 * note, linesize & data, might not match the next picture (for field pictures)
167 */
168 Picture next_picture;
169
170 /**
171 * copy of the source picture structure for encoding.
172 * note, linesize & data, might not match the source picture (for field pictures)
173 */
174 Picture new_picture;
175
176 /**
177 * copy of the current picture structure.
178 * note, linesize & data, might not match the current picture (for field pictures)
179 */
180 Picture current_picture; ///< buffer to store the decompressed current picture
181
182 Picture *last_picture_ptr; ///< pointer to the previous picture.
183 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
184 Picture *current_picture_ptr; ///< pointer to the current picture
185 int last_dc[3]; ///< last DC values for MPEG-1
186 int16_t *dc_val_base;
187 int16_t *dc_val[3]; ///< used for MPEG-4 DC prediction, all 3 arrays must be continuous
188 const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
189 const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
190 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (H.263)
191 uint8_t *coded_block_base;
192 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
193 int16_t (*ac_val_base)[16];
194 int16_t (*ac_val[3])[16]; ///< used for MPEG-4 AC prediction, all 3 arrays must be continuous
195 int mb_skipped; ///< MUST BE SET only during DECODING
196 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
197 and used for B-frame encoding & decoding (contains skip table of next P-frame) */
198 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
199 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
200 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
201
202 ScratchpadContext sc;
203
204 int qscale; ///< QP
205 int chroma_qscale; ///< chroma QP
206 unsigned int lambda; ///< Lagrange multiplier used in rate distortion
207 unsigned int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
208 int *lambda_table;
209 int adaptive_quant; ///< use adaptive quantization
210 int dquant; ///< qscale difference to prev qscale
211 int closed_gop; ///< MPEG1/2 GOP is closed
212 int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ...
213 int vbv_delay;
214 int last_pict_type; //FIXME removes
215 int last_non_b_pict_type; ///< used for MPEG-4 gmc B-frames & ratecontrol
216 int droppable;
217 int frame_rate_index;
218 AVRational mpeg2_frame_rate_ext;
219 int last_lambda_for[5]; ///< last lambda for a specific pict type
220 int skipdct; ///< skip dct and code zero residual
221
222 /* motion compensation */
223 int unrestricted_mv; ///< mv can point outside of the coded picture
224 int h263_long_vectors; ///< use horrible H.263v1 long vector mode
225
226 BlockDSPContext bdsp;
227 FDCTDSPContext fdsp;
228 H264ChromaContext h264chroma;
229 HpelDSPContext hdsp;
230 IDCTDSPContext idsp;
231 MECmpContext mecc;
232 MpegVideoDSPContext mdsp;
233 MpegvideoEncDSPContext mpvencdsp;
234 PixblockDSPContext pdsp;
235 QpelDSPContext qdsp;
236 VideoDSPContext vdsp;
237 H263DSPContext h263dsp;
238 int f_code; ///< forward MV resolution
239 int b_code; ///< backward MV resolution for B-frames (MPEG-4)
240 int16_t (*p_mv_table_base)[2];
241 int16_t (*b_forw_mv_table_base)[2];
242 int16_t (*b_back_mv_table_base)[2];
243 int16_t (*b_bidir_forw_mv_table_base)[2];
244 int16_t (*b_bidir_back_mv_table_base)[2];
245 int16_t (*b_direct_mv_table_base)[2];
246 int16_t (*p_field_mv_table_base[2][2])[2];
247 int16_t (*b_field_mv_table_base[2][2][2])[2];
248 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) P-frame encoding
249 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode B-frame encoding
250 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode B-frame encoding
251 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode B-frame encoding
252 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode B-frame encoding
253 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode B-frame encoding
254 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced P-frame encoding
255 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced B-frame encoding
256 uint8_t (*p_field_select_table[2]);
257 uint8_t (*b_field_select_table[2][2]);
258 int motion_est; ///< ME algorithm
259 int me_penalty_compensation;
260 int me_pre; ///< prepass for motion estimation
261 int mv_dir;
262 #define MV_DIR_FORWARD 1
263 #define MV_DIR_BACKWARD 2
264 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (MPEG-4)
265 int mv_type;
266 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
267 #define MV_TYPE_8X8 1 ///< 4 vectors (H.263, MPEG-4 4MV)
268 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
269 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
270 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
271 /**motion vectors for a macroblock
272 first coordinate : 0 = forward 1 = backward
273 second " : depend on type
274 third " : 0 = x, 1 = y
275 */
276 int mv[2][4][2];
277 int field_select[2][2];
278 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG-1 & B-frame MPEG-4
279 const uint8_t *fcode_tab; ///< smallest fcode needed for each MV
280 int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
281
282 MotionEstContext me;
283
284 int no_rounding; /**< apply no rounding to motion compensation (MPEG-4, msmpeg4, ...)
285 for B-frames rounding mode is always 0 */
286
287 /* macroblock layer */
288 int mb_x, mb_y;
289 int mb_skip_run;
290 int mb_intra;
291 uint16_t *mb_type; ///< Table for candidate MB types for encoding (defines in mpegutils.h)
292
293 int block_index[6]; ///< index to current MB in block based arrays with edges
294 int block_wrap[6];
295 uint8_t *dest[3];
296
297 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
298
299 /** matrix transmitted in the bitstream */
300 uint16_t intra_matrix[64];
301 uint16_t chroma_intra_matrix[64];
302 uint16_t inter_matrix[64];
303 uint16_t chroma_inter_matrix[64];
304 int force_duplicated_matrix; ///< Force duplication of mjpeg matrices, useful for rtp streaming
305
306 int intra_quant_bias; ///< bias for the quantizer
307 int inter_quant_bias; ///< bias for the quantizer
308 int min_qcoeff; ///< minimum encodable coefficient
309 int max_qcoeff; ///< maximum encodable coefficient
310 int ac_esc_length; ///< num of bits needed to encode the longest esc
311 uint8_t *intra_ac_vlc_length;
312 uint8_t *intra_ac_vlc_last_length;
313 uint8_t *intra_chroma_ac_vlc_length;
314 uint8_t *intra_chroma_ac_vlc_last_length;
315 uint8_t *inter_ac_vlc_length;
316 uint8_t *inter_ac_vlc_last_length;
317 uint8_t *luma_dc_vlc_length;
318 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
319
320 int coded_score[12];
321
322 /** precomputed matrix (combine qscale and DCT renorm) */
323 int (*q_intra_matrix)[64];
324 int (*q_chroma_intra_matrix)[64];
325 int (*q_inter_matrix)[64];
326 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
327 uint16_t (*q_intra_matrix16)[2][64];
328 uint16_t (*q_chroma_intra_matrix16)[2][64];
329 uint16_t (*q_inter_matrix16)[2][64];
330
331 /* noise reduction */
332 int (*dct_error_sum)[64];
333 int dct_count[2];
334 uint16_t (*dct_offset)[64];
335
336 /* bit rate control */
337 int64_t total_bits;
338 int frame_bits; ///< bits used for the current frame
339 int stuffing_bits; ///< bits used for stuffing
340 int next_lambda; ///< next lambda used for retrying to encode a frame
341 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
342
343 /* statistics, used for 2-pass encoding */
344 int mv_bits;
345 int header_bits;
346 int i_tex_bits;
347 int p_tex_bits;
348 int i_count;
349 int f_count;
350 int b_count;
351 int skip_count;
352 int misc_bits; ///< cbp, mb_type
353 int last_bits; ///< temp var used for calculating the above vars
354
355 /* error concealment / resync */
356 int resync_mb_x; ///< x position of last resync marker
357 int resync_mb_y; ///< y position of last resync marker
358 GetBitContext last_resync_gb; ///< used to search for the next resync marker
359 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
360 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed B-frames
361
362 ParseContext parse_context;
363
364 /* H.263 specific */
365 int gob_index;
366 int obmc; ///< overlapped block motion compensation
367 int mb_info; ///< interval for outputting info about mb offsets as side data
368 int prev_mb_info, last_mb_info;
369 uint8_t *mb_info_ptr;
370 int mb_info_size;
371 int ehc_mode;
372 #if FF_API_MPV_RC_STRATEGY
373 int rc_strategy; ///< deprecated
374 #endif
375
376 /* H.263+ specific */
377 int umvplus; ///< == H.263+ && unrestricted_mv
378 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
379 int h263_slice_structured;
380 int alt_inter_vlc; ///< alternative inter vlc
381 int modified_quant;
382 int loop_filter;
383 int custom_pcf;
384
385 /* MPEG-4 specific */
386 int studio_profile;
387 int dct_precision;
388 ///< number of bits to represent the fractional part of time (encoder only)
389 int time_increment_bits;
390 int last_time_base;
391 int time_base; ///< time in seconds of last I,P,S Frame
392 int64_t time; ///< time of current frame
393 int64_t last_non_b_time;
394 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
395 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
396 uint16_t pp_field_time;
397 uint16_t pb_field_time; ///< like above, just for interlaced
398 int real_sprite_warping_points;
399 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
400 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
401 int mcsel;
402 int quant_precision;
403 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
404 int aspect_ratio_info; //FIXME remove
405 int sprite_warping_accuracy;
406 int data_partitioning; ///< data partitioning flag from header
407 int partitioned_frame; ///< is current frame partitioned
408 int low_delay; ///< no reordering needed / has no B-frames
409 int vo_type;
410 PutBitContext tex_pb; ///< used for data partitioned VOPs
411 PutBitContext pb2; ///< used for data partitioned VOPs
412 int mpeg_quant;
413 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG-4
414
415 /* divx specific, used to workaround (many) bugs in divx5 */
416 int divx_packed;
417 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
418 int bitstream_buffer_size;
419 unsigned int allocated_bitstream_buffer_size;
420
421 /* RV10 specific */
422 int rv10_version; ///< RV10 version: 0 or 3
423 int rv10_first_dc_coded[3];
424
425 /* MJPEG specific */
426 struct MJpegContext *mjpeg_ctx;
427 int esc_pos;
428 int pred;
429 int huffman;
430
431 /* MSMPEG4 specific */
432 int mv_table_index;
433 int rl_table_index;
434 int rl_chroma_table_index;
435 int dc_table_index;
436 int use_skip_mb_code;
437 int slice_height; ///< in macroblocks
438 int first_slice_line; ///< used in MPEG-4 too to handle resync markers
439 int flipflop_rounding;
440 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
441 int per_mb_rl_table;
442 int esc3_level_length;
443 int esc3_run_length;
444 /** [mb_intra][isChroma][level][run][last] */
445 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
446 int inter_intra_pred;
447 int mspel;
448
449 /* SpeedHQ specific */
450 int slice_start;
451
452 /* decompression specific */
453 GetBitContext gb;
454
455 /* MPEG-1 specific */
456 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & MPEG-1 specific
457 int last_mv_dir; ///< last mv_dir, used for B-frame encoding
458 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
459
460 /* MPEG-2-specific - I wished not to have to support this mess. */
461 int progressive_sequence;
462 int mpeg_f_code[2][2];
463 int a53_cc;
464
465 // picture structure defines are loaded from mpegutils.h
466 int picture_structure;
467
468 int64_t timecode_frame_start; ///< GOP timecode frame start number, in non drop frame format
469 int intra_dc_precision;
470 int frame_pred_frame_dct;
471 int top_field_first;
472 int concealment_motion_vectors;
473 int q_scale_type;
474 int brd_scale;
475 int intra_vlc_format;
476 int alternate_scan;
477 int seq_disp_ext;
478 int video_format;
479 #define VIDEO_FORMAT_COMPONENT 0
480 #define VIDEO_FORMAT_PAL 1
481 #define VIDEO_FORMAT_NTSC 2
482 #define VIDEO_FORMAT_SECAM 3
483 #define VIDEO_FORMAT_MAC 4
484 #define VIDEO_FORMAT_UNSPECIFIED 5
485 int repeat_first_field;
486 int chroma_420_type;
487 int chroma_format;
488 #define CHROMA_420 1
489 #define CHROMA_422 2
490 #define CHROMA_444 3
491 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
492 int chroma_y_shift;
493
494 int progressive_frame;
495 int full_pel[2];
496 int interlaced_dct;
497 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
498 int drop_frame_timecode; ///< timecode is in drop frame format.
499 int scan_offset; ///< reserve space for SVCD scan offset user data.
500
501 /* RTP specific */
502 int rtp_mode;
503 int rtp_payload_size;
504
505 char *tc_opt_str; ///< timecode option string
506 AVTimecode tc; ///< timecode context
507
508 uint8_t *ptr_lastgob;
509 int swap_uv; //vcr2 codec is an MPEG-2 variant with U and V swapped
510 int pack_pblocks; //xvmc needs to keep blocks without gaps.
511 int16_t (*pblocks[12])[64];
512
513 int16_t (*block)[64]; ///< points to one of the following blocks
514 int16_t (*blocks)[12][64]; // for HQ mode we need to keep the best block
515 int (*decode_mb)(struct MpegEncContext *s, int16_t block[12][64]); // used by some codecs to avoid a switch()
516
517 int32_t (*block32)[12][64];
518 int dpcm_direction; // 0 = DCT, 1 = DPCM top to bottom scan, -1 = DPCM bottom to top scan
519 int16_t (*dpcm_macroblock)[3][256];
520
521 #define SLICE_OK 0
522 #define SLICE_ERROR -1
523 #define SLICE_END -2 ///<end marker found
524 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
525
526 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
527 int16_t *block/*align 16*/, int n, int qscale);
528 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
529 int16_t *block/*align 16*/, int n, int qscale);
530 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
531 int16_t *block/*align 16*/, int n, int qscale);
532 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
533 int16_t *block/*align 16*/, int n, int qscale);
534 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
535 int16_t *block/*align 16*/, int n, int qscale);
536 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
537 int16_t *block/*align 16*/, int n, int qscale);
538 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (MPEG-4 can use both)
539 int16_t *block/*align 16*/, int n, int qscale);
540 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (MPEG-4 can use both)
541 int16_t *block/*align 16*/, int n, int qscale);
542 int (*dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
543 int (*fast_dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
544 void (*denoise_dct)(struct MpegEncContext *s, int16_t *block);
545
546 int mpv_flags; ///< flags set by private options
547 int quantizer_noise_shaping;
548
549 /**
550 * ratecontrol qmin qmax limiting method
551 * 0-> clipping, 1-> use a nice continuous function to limit qscale within qmin/qmax.
552 */
553 float rc_qsquish;
554 float rc_qmod_amp;
555 int rc_qmod_freq;
556 float rc_initial_cplx;
557 float rc_buffer_aggressivity;
558 float border_masking;
559 int lmin, lmax;
560 int vbv_ignore_qmax;
561
562 char *rc_eq;
563
564 /* temp buffers for rate control */
565 float *cplx_tab, *bits_tab;
566
567 /* flag to indicate a reinitialization is required, e.g. after
568 * a frame size change */
569 int context_reinit;
570
571 ERContext er;
572
573 int error_rate;
574
575 /* temporary frames used by b_frame_strategy = 2 */
576 AVFrame *tmp_frames[MAX_B_FRAMES + 2];
577 int b_frame_strategy;
578 int b_sensitivity;
579
580 /* frame skip options for encoding */
581 int frame_skip_threshold;
582 int frame_skip_factor;
583 int frame_skip_exp;
584 int frame_skip_cmp;
585
586 int scenechange_threshold;
587 int noise_reduction;
588
589 int intra_penalty;
590 } MpegEncContext;
591
592 /* mpegvideo_enc common options */
593 #define FF_MPV_FLAG_SKIP_RD 0x0001
594 #define FF_MPV_FLAG_STRICT_GOP 0x0002
595 #define FF_MPV_FLAG_QP_RD 0x0004
596 #define FF_MPV_FLAG_CBP_RD 0x0008
597 #define FF_MPV_FLAG_NAQ 0x0010
598 #define FF_MPV_FLAG_MV0 0x0020
599
600 #define FF_MPV_OPT_CMP_FUNC \
601 { "sad", "Sum of absolute differences, fast", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_SAD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
602 { "sse", "Sum of squared errors", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_SSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
603 { "satd", "Sum of absolute Hadamard transformed differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_SATD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
604 { "dct", "Sum of absolute DCT transformed differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCT }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
605 { "psnr", "Sum of squared quantization errors, low quality", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_PSNR }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
606 { "bit", "Number of bits needed for the block", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_BIT }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
607 { "rd", "Rate distortion optimal, slow", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_RD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
608 { "zero", "Zero", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_ZERO }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
609 { "vsad", "Sum of absolute vertical differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_VSAD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
610 { "vsse", "Sum of squared vertical differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_VSSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
611 { "nsse", "Noise preserving sum of squared differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_NSSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
612 { "dct264", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCT264 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
613 { "dctmax", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCTMAX }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
614 { "chroma", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_CHROMA }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
615 { "msad", "Sum of absolute differences, median predicted", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_MEDIAN_SAD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }
616
617 #ifndef FF_MPV_OFFSET
618 #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x)
619 #endif
620 #if FF_API_MPV_RC_STRATEGY
621 #define FF_MPV_RC_STRATEGY_OPTS \
622 {"rc_strategy", "ratecontrol method", FF_MPV_OFFSET(rc_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 1, FF_MPV_OPT_FLAGS | AV_OPT_FLAG_DEPRECATED, "rc_strategy" }, \
623 { "ffmpeg", "deprecated, does nothing", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, FF_MPV_OPT_FLAGS | AV_OPT_FLAG_DEPRECATED, "rc_strategy" }, \
624 { "xvid", "deprecated, does nothing", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, FF_MPV_OPT_FLAGS | AV_OPT_FLAG_DEPRECATED, "rc_strategy" },
625 #else
626 #define FF_MPV_RC_STRATEGY_OPTS
627 #endif
628 #define FF_MPV_OPT_FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
629 #define FF_MPV_COMMON_OPTS \
630 FF_MPV_OPT_CMP_FUNC, \
631 { "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" },\
632 { "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" },\
633 { "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" },\
634 { "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" },\
635 { "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" },\
636 { "naq", "normalize adaptive quantization", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_NAQ }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
637 { "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" },\
638 { "luma_elim_threshold", "single coefficient elimination threshold for luminance (negative values also consider dc coefficient)",\
639 FF_MPV_OFFSET(luma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
640 { "chroma_elim_threshold", "single coefficient elimination threshold for chrominance (negative values also consider dc coefficient)",\
641 FF_MPV_OFFSET(chroma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
642 { "quantizer_noise_shaping", NULL, FF_MPV_OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
643 { "error_rate", "Simulate errors in the bitstream to test error concealment.", \
644 FF_MPV_OFFSET(error_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
645 {"qsquish", "how to keep quantizer between qmin and qmax (0 = clip, 1 = use differentiable function)", \
646 FF_MPV_OFFSET(rc_qsquish), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, 0, 99, FF_MPV_OPT_FLAGS}, \
647 {"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}, \
648 {"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}, \
649 {"rc_eq", "Set rate control equation. When computing the expression, besides the standard functions " \
650 "defined in the section 'Expression Evaluation', the following functions are available: " \
651 "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv " \
652 "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.", \
653 FF_MPV_OFFSET(rc_eq), AV_OPT_TYPE_STRING, .flags = FF_MPV_OPT_FLAGS }, \
654 {"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}, \
655 {"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}, \
656 {"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}, \
657 {"lmin", "minimum Lagrange factor (VBR)", FF_MPV_OFFSET(lmin), AV_OPT_TYPE_INT, {.i64 = 2*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
658 {"lmax", "maximum Lagrange factor (VBR)", FF_MPV_OFFSET(lmax), AV_OPT_TYPE_INT, {.i64 = 31*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
659 {"ibias", "intra quant bias", FF_MPV_OFFSET(intra_quant_bias), AV_OPT_TYPE_INT, {.i64 = FF_DEFAULT_QUANT_BIAS }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
660 {"pbias", "inter quant bias", FF_MPV_OFFSET(inter_quant_bias), AV_OPT_TYPE_INT, {.i64 = FF_DEFAULT_QUANT_BIAS }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
661 {"motion_est", "motion estimation algorithm", FF_MPV_OFFSET(motion_est), AV_OPT_TYPE_INT, {.i64 = FF_ME_EPZS }, FF_ME_ZERO, FF_ME_XONE, FF_MPV_OPT_FLAGS, "motion_est" }, \
662 { "zero", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_ZERO }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
663 { "epzs", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_EPZS }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
664 { "xone", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_XONE }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
665 { "force_duplicated_matrix", "Always write luma and chroma matrix for mjpeg, useful for rtp streaming.", FF_MPV_OFFSET(force_duplicated_matrix), AV_OPT_TYPE_BOOL, {.i64 = 0 }, 0, 1, FF_MPV_OPT_FLAGS }, \
666 {"b_strategy", "Strategy to choose between I/P/B-frames", FF_MPV_OFFSET(b_frame_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 2, FF_MPV_OPT_FLAGS }, \
667 {"b_sensitivity", "Adjust sensitivity of b_frame_strategy 1", FF_MPV_OFFSET(b_sensitivity), AV_OPT_TYPE_INT, {.i64 = 40 }, 1, INT_MAX, FF_MPV_OPT_FLAGS }, \
668 {"brd_scale", "Downscale frames for dynamic B-frame decision", FF_MPV_OFFSET(brd_scale), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 3, FF_MPV_OPT_FLAGS }, \
669 {"skip_threshold", "Frame skip threshold", FF_MPV_OFFSET(frame_skip_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
670 {"skip_factor", "Frame skip factor", FF_MPV_OFFSET(frame_skip_factor), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
671 {"skip_exp", "Frame skip exponent", FF_MPV_OFFSET(frame_skip_exp), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
672 {"skip_cmp", "Frame skip compare function", FF_MPV_OFFSET(frame_skip_cmp), AV_OPT_TYPE_INT, {.i64 = FF_CMP_DCTMAX }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
673 {"sc_threshold", "Scene change threshold", FF_MPV_OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
674 {"noise_reduction", "Noise reduction", FF_MPV_OFFSET(noise_reduction), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
675 {"mpeg_quant", "Use MPEG quantizers instead of H.263", FF_MPV_OFFSET(mpeg_quant), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 1, FF_MPV_OPT_FLAGS }, \
676 {"ps", "RTP payload size in bytes", FF_MPV_OFFSET(rtp_payload_size), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
677 {"mepc", "Motion estimation bitrate penalty compensation (1.0 = 256)", FF_MPV_OFFSET(me_penalty_compensation), AV_OPT_TYPE_INT, {.i64 = 256 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
678 {"mepre", "pre motion estimation", FF_MPV_OFFSET(me_pre), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
679 {"intra_penalty", "Penalty for intra blocks in block decision", FF_MPV_OFFSET(intra_penalty), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX/2, FF_MPV_OPT_FLAGS }, \
680 {"a53cc", "Use A53 Closed Captions (if available)", FF_MPV_OFFSET(a53_cc), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, FF_MPV_OPT_FLAGS }, \
681 FF_MPV_RC_STRATEGY_OPTS
682
683 extern const AVOption ff_mpv_generic_options[];
684
685 /**
686 * Set the given MpegEncContext to common defaults (same for encoding
687 * and decoding). The changed fields will not depend upon the prior
688 * state of the MpegEncContext.
689 */
690 void ff_mpv_common_defaults(MpegEncContext *s);
691
692 void ff_dct_encode_init_x86(MpegEncContext *s);
693
694 int ff_mpv_common_init(MpegEncContext *s);
695 void ff_mpv_common_init_arm(MpegEncContext *s);
696 void ff_mpv_common_init_axp(MpegEncContext *s);
697 void ff_mpv_common_init_neon(MpegEncContext *s);
698 void ff_mpv_common_init_ppc(MpegEncContext *s);
699 void ff_mpv_common_init_x86(MpegEncContext *s);
700 void ff_mpv_common_init_mips(MpegEncContext *s);
701
702 int ff_mpv_common_frame_size_change(MpegEncContext *s);
703 void ff_mpv_common_end(MpegEncContext *s);
704
705 void ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx);
706 void ff_mpv_reconstruct_mb(MpegEncContext *s, int16_t block[12][64]);
707 void ff_mpv_report_decode_progress(MpegEncContext *s);
708
709 int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx);
710 void ff_mpv_frame_end(MpegEncContext *s);
711
712 int ff_mpv_encode_init(AVCodecContext *avctx);
713 void ff_mpv_encode_init_x86(MpegEncContext *s);
714
715 int ff_mpv_encode_end(AVCodecContext *avctx);
716 int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
717 const AVFrame *frame, int *got_packet);
718 int ff_mpv_reallocate_putbitbuffer(MpegEncContext *s, size_t threshold, size_t size_increase);
719
720 void ff_clean_intra_table_entries(MpegEncContext *s);
721 void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h);
722 void ff_mpeg_flush(AVCodecContext *avctx);
723
724 void ff_print_debug_info(MpegEncContext *s, Picture *p, AVFrame *pict);
725
726 int ff_mpv_export_qp_table(MpegEncContext *s, AVFrame *f, Picture *p, int qp_type);
727
728 void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
729
730 int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
731 int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
732 void ff_set_qscale(MpegEncContext * s, int qscale);
733
734 void ff_mpv_idct_init(MpegEncContext *s);
735 int ff_dct_encode_init(MpegEncContext *s);
736 void ff_convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16)[2][64],
737 const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra);
738 int ff_dct_quantize_c(MpegEncContext *s, int16_t *block, int n, int qscale, int *overflow);
739 void ff_block_permute(int16_t *block, uint8_t *permutation,
740 const uint8_t *scantable, int last);
741 void ff_init_block_index(MpegEncContext *s);
742
743 void ff_mpv_motion(MpegEncContext *s,
744 uint8_t *dest_y, uint8_t *dest_cb,
745 uint8_t *dest_cr, int dir,
746 uint8_t **ref_picture,
747 op_pixels_func (*pix_op)[4],
748 qpel_mc_func (*qpix_op)[16]);
749
ff_update_block_index(MpegEncContext * s)750 static inline void ff_update_block_index(MpegEncContext *s){
751 const int bytes_per_pixel = 1 + (s->avctx->bits_per_raw_sample > 8);
752 const int block_size= (8*bytes_per_pixel) >> s->avctx->lowres;
753
754 s->block_index[0]+=2;
755 s->block_index[1]+=2;
756 s->block_index[2]+=2;
757 s->block_index[3]+=2;
758 s->block_index[4]++;
759 s->block_index[5]++;
760 s->dest[0]+= 2*block_size;
761 s->dest[1]+= (2 >> s->chroma_x_shift) * block_size;
762 s->dest[2]+= (2 >> s->chroma_x_shift) * block_size;
763 }
764
get_bits_diff(MpegEncContext * s)765 static inline int get_bits_diff(MpegEncContext *s){
766 const int bits= put_bits_count(&s->pb);
767 const int last= s->last_bits;
768
769 s->last_bits = bits;
770
771 return bits - last;
772 }
773
mpeg_get_qscale(MpegEncContext * s)774 static inline int mpeg_get_qscale(MpegEncContext *s)
775 {
776 int qscale = get_bits(&s->gb, 5);
777 if (s->q_scale_type)
778 return ff_mpeg2_non_linear_qscale[qscale];
779 else
780 return qscale << 1;
781 }
782
783 #endif /* AVCODEC_MPEGVIDEO_H */
784