second try of improved lambda/qp guessing for motion estimation RD in first pass
[libav.git] / libavcodec / mpegvideo.h
1 /*
2 * Generic DCT based hybrid video encoder
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
4 * Copyright (c) 2002-2004 Michael Niedermayer
5 *
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
10 *
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19 */
20
21 /**
22 * @file mpegvideo.h
23 * mpegvideo header.
24 */
25
26 #ifndef AVCODEC_MPEGVIDEO_H
27 #define AVCODEC_MPEGVIDEO_H
28
29 #include "dsputil.h"
30 #include "bitstream.h"
31
32 #define FRAME_SKIPPED 100 ///< return value for header parsers if frame is not coded
33
34 enum OutputFormat {
35 FMT_MPEG1,
36 FMT_H261,
37 FMT_H263,
38 FMT_MJPEG,
39 FMT_H264,
40 };
41
42 #define EDGE_WIDTH 16
43
44 #define MPEG_BUF_SIZE (16 * 1024)
45
46 #define QMAT_SHIFT_MMX 16
47 #define QMAT_SHIFT 22
48
49 #define MAX_FCODE 7
50 #define MAX_MV 2048
51
52 #define MAX_THREADS 8
53
54 #define MAX_PICTURE_COUNT 32
55
56 #define ME_MAP_SIZE 64
57 #define ME_MAP_SHIFT 3
58 #define ME_MAP_MV_BITS 11
59
60 /* run length table */
61 #define MAX_RUN 64
62 #define MAX_LEVEL 64
63
64 #define I_TYPE FF_I_TYPE ///< Intra
65 #define P_TYPE FF_P_TYPE ///< Predicted
66 #define B_TYPE FF_B_TYPE ///< Bi-dir predicted
67 #define S_TYPE FF_S_TYPE ///< S(GMC)-VOP MPEG4
68 #define SI_TYPE FF_SI_TYPE ///< Switching Intra
69 #define SP_TYPE FF_SP_TYPE ///< Switching Predicted
70
71 #define MAX_MB_BYTES (30*16*16*3/8 + 120)
72
73 typedef struct Predictor{
74 double coeff;
75 double count;
76 double decay;
77 } Predictor;
78
79 typedef struct RateControlEntry{
80 int pict_type;
81 float qscale;
82 int mv_bits;
83 int i_tex_bits;
84 int p_tex_bits;
85 int misc_bits;
86 uint64_t expected_bits;
87 int new_pict_type;
88 float new_qscale;
89 int mc_mb_var_sum;
90 int mb_var_sum;
91 int i_count;
92 int f_code;
93 int b_code;
94 }RateControlEntry;
95
96 /**
97 * rate control context.
98 */
99 typedef struct RateControlContext{
100 FILE *stats_file;
101 int num_entries; ///< number of RateControlEntries
102 RateControlEntry *entry;
103 double buffer_index; ///< amount of bits in the video/audio buffer
104 Predictor pred[5];
105 double short_term_qsum; ///< sum of recent qscales
106 double short_term_qcount; ///< count of recent qscales
107 double pass1_rc_eq_output_sum;///< sum of the output of the rc equation, this is used for normalization
108 double pass1_wanted_bits; ///< bits which should have been outputed by the pass1 code (including complexity init)
109 double last_qscale;
110 double last_qscale_for[5]; ///< last qscale for a specific pict type, used for max_diff & ipb factor stuff
111 int last_mc_mb_var_sum;
112 int last_mb_var_sum;
113 uint64_t i_cplx_sum[5];
114 uint64_t p_cplx_sum[5];
115 uint64_t mv_bits_sum[5];
116 uint64_t qscale_sum[5];
117 int frame_count[5];
118 int last_non_b_pict_type;
119 }RateControlContext;
120
121 /**
122 * Scantable.
123 */
124 typedef struct ScanTable{
125 const uint8_t *scantable;
126 uint8_t permutated[64];
127 uint8_t raster_end[64];
128 #ifdef ARCH_POWERPC
129 /** Used by dct_quantise_alitvec to find last-non-zero */
130 uint8_t __align8 inverse[64];
131 #endif
132 } ScanTable;
133
134 /**
135 * Picture.
136 */
137 typedef struct Picture{
138 FF_COMMON_FRAME
139
140 /**
141 * halfpel luma planes.
142 */
143 uint8_t *interpolated[3];
144 int16_t (*motion_val_base[2])[2];
145 uint32_t *mb_type_base;
146 #define MB_TYPE_INTRA MB_TYPE_INTRA4x4 //default mb_type if theres just one type
147 #define IS_INTRA4x4(a) ((a)&MB_TYPE_INTRA4x4)
148 #define IS_INTRA16x16(a) ((a)&MB_TYPE_INTRA16x16)
149 #define IS_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
150 #define IS_INTRA(a) ((a)&7)
151 #define IS_INTER(a) ((a)&(MB_TYPE_16x16|MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8))
152 #define IS_SKIP(a) ((a)&MB_TYPE_SKIP)
153 #define IS_INTRA_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
154 #define IS_INTERLACED(a) ((a)&MB_TYPE_INTERLACED)
155 #define IS_DIRECT(a) ((a)&MB_TYPE_DIRECT2)
156 #define IS_GMC(a) ((a)&MB_TYPE_GMC)
157 #define IS_16X16(a) ((a)&MB_TYPE_16x16)
158 #define IS_16X8(a) ((a)&MB_TYPE_16x8)
159 #define IS_8X16(a) ((a)&MB_TYPE_8x16)
160 #define IS_8X8(a) ((a)&MB_TYPE_8x8)
161 #define IS_SUB_8X8(a) ((a)&MB_TYPE_16x16) //note reused
162 #define IS_SUB_8X4(a) ((a)&MB_TYPE_16x8) //note reused
163 #define IS_SUB_4X8(a) ((a)&MB_TYPE_8x16) //note reused
164 #define IS_SUB_4X4(a) ((a)&MB_TYPE_8x8) //note reused
165 #define IS_ACPRED(a) ((a)&MB_TYPE_ACPRED)
166 #define IS_QUANT(a) ((a)&MB_TYPE_QUANT)
167 #define IS_DIR(a, part, list) ((a) & (MB_TYPE_P0L0<<((part)+2*(list))))
168 #define USES_LIST(a, list) ((a) & ((MB_TYPE_P0L0|MB_TYPE_P1L0)<<(2*(list)))) ///< does this mb use listX, note doesnt work if subMBs
169 #define HAS_CBP(a) ((a)&MB_TYPE_CBP)
170
171 int field_poc[2]; ///< h264 top/bottom POC
172 int poc; ///< h264 frame POC
173 int frame_num; ///< h264 frame_num
174 int pic_id; ///< h264 pic_num or long_term_pic_idx
175 int long_ref; ///< 1->long term reference 0->short term reference
176 int ref_poc[2][16]; ///< h264 POCs of the frames used as reference
177 int ref_count[2]; ///< number of entries in ref_poc
178
179 int mb_var_sum; ///< sum of MB variance for current frame
180 int mc_mb_var_sum; ///< motion compensated MB variance for current frame
181 uint16_t *mb_var; ///< Table for MB variances
182 uint16_t *mc_mb_var; ///< Table for motion compensated MB variances
183 uint8_t *mb_mean; ///< Table for MB luminance
184 int32_t *mb_cmp_score; ///< Table for MB cmp scores, for mb decision FIXME remove
185 int b_frame_score; /* */
186 } Picture;
187
188 typedef struct ParseContext{
189 uint8_t *buffer;
190 int index;
191 int last_index;
192 int buffer_size;
193 uint32_t state; ///< contains the last few bytes in MSB order
194 int frame_start_found;
195 int overread; ///< the number of bytes which where irreversibly read from the next frame
196 int overread_index; ///< the index into ParseContext.buffer of the overreaded bytes
197 } ParseContext;
198
199 struct MpegEncContext;
200
201 /**
202 * Motion estimation context.
203 */
204 typedef struct MotionEstContext{
205 AVCodecContext *avctx;
206 int skip; ///< set if ME is skipped for the current MB
207 int co_located_mv[4][2]; ///< mv from last p frame for direct mode ME
208 int direct_basis_mv[4][2];
209 uint8_t *scratchpad; ///< data area for the me algo, so that the ME doesnt need to malloc/free
210 uint8_t *best_mb;
211 uint8_t *temp_mb[2];
212 uint8_t *temp;
213 int best_bits;
214 uint32_t *map; ///< map to avoid duplicate evaluations
215 uint32_t *score_map; ///< map to store the scores
216 int map_generation;
217 int pre_penalty_factor;
218 int penalty_factor;
219 int sub_penalty_factor;
220 int mb_penalty_factor;
221 int flags;
222 int sub_flags;
223 int mb_flags;
224 int pre_pass; ///< = 1 for the pre pass
225 int dia_size;
226 int xmin;
227 int xmax;
228 int ymin;
229 int ymax;
230 int pred_x;
231 int pred_y;
232 uint8_t *src[4][4];
233 uint8_t *ref[4][4];
234 int stride;
235 int uvstride;
236 /* temp variables for picture complexity calculation */
237 int mc_mb_var_sum_temp;
238 int mb_var_sum_temp;
239 int scene_change_score;
240 /* cmp, chroma_cmp;*/
241 op_pixels_func (*hpel_put)[4];
242 op_pixels_func (*hpel_avg)[4];
243 qpel_mc_func (*qpel_put)[16];
244 qpel_mc_func (*qpel_avg)[16];
245 uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
246 uint8_t *current_mv_penalty;
247 int (*sub_motion_search)(struct MpegEncContext * s,
248 int *mx_ptr, int *my_ptr, int dmin,
249 int src_index, int ref_index,
250 int size, int h);
251 }MotionEstContext;
252
253 /**
254 * MpegEncContext.
255 */
256 typedef struct MpegEncContext {
257 struct AVCodecContext *avctx;
258 /* the following parameters must be initialized before encoding */
259 int width, height;///< picture size. must be a multiple of 16
260 int gop_size;
261 int intra_only; ///< if true, only intra pictures are generated
262 int bit_rate; ///< wanted bit rate
263 enum OutputFormat out_format; ///< output format
264 int h263_pred; ///< use mpeg4/h263 ac/dc predictions
265
266 /* the following codec id fields are deprecated in favor of codec_id */
267 int h263_plus; ///< h263 plus headers
268 int h263_msmpeg4; ///< generate MSMPEG4 compatible stream (deprecated, use msmpeg4_version instead)
269 int h263_flv; ///< use flv h263 header
270
271 enum CodecID codec_id; /* see CODEC_ID_xxx */
272 int fixed_qscale; ///< fixed qscale if non zero
273 int encoding; ///< true if we are encoding (vs decoding)
274 int flags; ///< AVCodecContext.flags (HQ, MV4, ...)
275 int flags2; ///< AVCodecContext.flags2
276 int max_b_frames; ///< max number of b-frames for encoding
277 int luma_elim_threshold;
278 int chroma_elim_threshold;
279 int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
280 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
281 /* the following fields are managed internally by the encoder */
282
283 /** bit output */
284 PutBitContext pb;
285
286 /* sequence parameters */
287 int context_initialized;
288 int input_picture_number; ///< used to set pic->display_picture_number, shouldnt be used for/by anything else
289 int coded_picture_number; ///< used to set pic->coded_picture_number, shouldnt be used for/by anything else
290 int picture_number; //FIXME remove, unclear definition
291 int picture_in_gop_number; ///< 0-> first pic in gop, ...
292 int b_frames_since_non_b; ///< used for encoding, relative to not yet reordered input
293 int64_t user_specified_pts;///< last non zero pts from AVFrame which was passed into avcodec_encode_video()
294 int mb_width, mb_height; ///< number of MBs horizontally & vertically
295 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
296 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
297 int b4_stride; ///< 4*mb_width+1 used for some 4x4 block arrays to allow simple addressing
298 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
299 int mb_num; ///< number of MBs of a picture
300 int linesize; ///< line size, in bytes, may be different from width
301 int uvlinesize; ///< line size, for chroma in bytes, may be different from width
302 Picture *picture; ///< main picture buffer
303 Picture **input_picture; ///< next pictures on display order for encoding
304 Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
305
306 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
307 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
308 struct MpegEncContext *thread_context[MAX_THREADS];
309
310 /**
311 * copy of the previous picture structure.
312 * note, linesize & data, might not match the previous picture (for field pictures)
313 */
314 Picture last_picture;
315
316 /**
317 * copy of the next picture structure.
318 * note, linesize & data, might not match the next picture (for field pictures)
319 */
320 Picture next_picture;
321
322 /**
323 * copy of the source picture structure for encoding.
324 * note, linesize & data, might not match the source picture (for field pictures)
325 */
326 Picture new_picture;
327
328 /**
329 * copy of the current picture structure.
330 * note, linesize & data, might not match the current picture (for field pictures)
331 */
332 Picture current_picture; ///< buffer to store the decompressed current picture
333
334 Picture *last_picture_ptr; ///< pointer to the previous picture.
335 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
336 Picture *current_picture_ptr; ///< pointer to the current picture
337 uint8_t *visualization_buffer[3]; //< temporary buffer vor MV visualization
338 int last_dc[3]; ///< last DC values for MPEG1
339 int16_t *dc_val_base;
340 int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
341 int16_t dc_cache[4*5];
342 int y_dc_scale, c_dc_scale;
343 const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
344 const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
345 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
346 uint8_t *coded_block_base;
347 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
348 int16_t (*ac_val_base)[16];
349 int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
350 int ac_pred;
351 uint8_t *prev_pict_types; ///< previous picture types in bitstream order, used for mb skip
352 #define PREV_PICT_TYPES_BUFFER_SIZE 256
353 int mb_skipped; ///< MUST BE SET only during DECODING
354 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
355 and used for b-frame encoding & decoding (contains skip table of next P Frame) */
356 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
357 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
358 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
359 uint8_t *allocated_edge_emu_buffer;
360 uint8_t *edge_emu_buffer; ///< points into the middle of allocated_edge_emu_buffer
361 uint8_t *rd_scratchpad; ///< scratchpad for rate distortion mb decision
362 uint8_t *obmc_scratchpad;
363 uint8_t *b_scratchpad; ///< scratchpad used for writing into write only buffers
364
365 int qscale; ///< QP
366 int chroma_qscale; ///< chroma QP
367 int lambda; ///< lagrange multipler used in rate distortion
368 int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
369 int *lambda_table;
370 int adaptive_quant; ///< use adaptive quantization
371 int dquant; ///< qscale difference to prev qscale
372 int pict_type; ///< I_TYPE, P_TYPE, B_TYPE, ...
373 int last_pict_type; //FIXME removes
374 int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
375 int dropable;
376 int frame_rate_index;
377 int last_lambda_for[5]; ///< last lambda for a specific pict type
378
379 /* motion compensation */
380 int unrestricted_mv; ///< mv can point outside of the coded picture
381 int h263_long_vectors; ///< use horrible h263v1 long vector mode
382 int decode; ///< if 0 then decoding will be skipped (for encoding b frames for example)
383
384 DSPContext dsp; ///< pointers for accelerated dsp functions
385 int f_code; ///< forward MV resolution
386 int b_code; ///< backward MV resolution for B Frames (mpeg4)
387 int16_t (*p_mv_table_base)[2];
388 int16_t (*b_forw_mv_table_base)[2];
389 int16_t (*b_back_mv_table_base)[2];
390 int16_t (*b_bidir_forw_mv_table_base)[2];
391 int16_t (*b_bidir_back_mv_table_base)[2];
392 int16_t (*b_direct_mv_table_base)[2];
393 int16_t (*p_field_mv_table_base[2][2])[2];
394 int16_t (*b_field_mv_table_base[2][2][2])[2];
395 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
396 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
397 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
398 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
399 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
400 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
401 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
402 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
403 uint8_t (*p_field_select_table[2]);
404 uint8_t (*b_field_select_table[2][2]);
405 int me_method; ///< ME algorithm
406 int mv_dir;
407 #define MV_DIR_BACKWARD 1
408 #define MV_DIR_FORWARD 2
409 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
410 int mv_type;
411 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
412 #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
413 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
414 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
415 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
416 /**motion vectors for a macroblock
417 first coordinate : 0 = forward 1 = backward
418 second " : depend on type
419 third " : 0 = x, 1 = y
420 */
421 int mv[2][4][2];
422 int field_select[2][2];
423 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
424 uint8_t *fcode_tab; ///< smallest fcode needed for each MV
425
426 MotionEstContext me;
427
428 int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
429 for b-frames rounding mode is allways 0 */
430
431 int hurry_up; /**< when set to 1 during decoding, b frames will be skipped
432 when set to 2 idct/dequant will be skipped too */
433
434 /* macroblock layer */
435 int mb_x, mb_y;
436 int mb_skip_run;
437 int mb_intra;
438 uint16_t *mb_type; ///< Table for candidate MB types for encoding
439 #define CANDIDATE_MB_TYPE_INTRA 0x01
440 #define CANDIDATE_MB_TYPE_INTER 0x02
441 #define CANDIDATE_MB_TYPE_INTER4V 0x04
442 #define CANDIDATE_MB_TYPE_SKIPPED 0x08
443 //#define MB_TYPE_GMC 0x10
444
445 #define CANDIDATE_MB_TYPE_DIRECT 0x10
446 #define CANDIDATE_MB_TYPE_FORWARD 0x20
447 #define CANDIDATE_MB_TYPE_BACKWARD 0x40
448 #define CANDIDATE_MB_TYPE_BIDIR 0x80
449
450 #define CANDIDATE_MB_TYPE_INTER_I 0x100
451 #define CANDIDATE_MB_TYPE_FORWARD_I 0x200
452 #define CANDIDATE_MB_TYPE_BACKWARD_I 0x400
453 #define CANDIDATE_MB_TYPE_BIDIR_I 0x800
454
455 int block_index[6]; ///< index to current MB in block based arrays with edges
456 int block_wrap[6];
457 uint8_t *dest[3];
458
459 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
460
461 /** matrix transmitted in the bitstream */
462 uint16_t intra_matrix[64];
463 uint16_t chroma_intra_matrix[64];
464 uint16_t inter_matrix[64];
465 uint16_t chroma_inter_matrix[64];
466 #define QUANT_BIAS_SHIFT 8
467 int intra_quant_bias; ///< bias for the quantizer
468 int inter_quant_bias; ///< bias for the quantizer
469 int min_qcoeff; ///< minimum encodable coefficient
470 int max_qcoeff; ///< maximum encodable coefficient
471 int ac_esc_length; ///< num of bits needed to encode the longest esc
472 uint8_t *intra_ac_vlc_length;
473 uint8_t *intra_ac_vlc_last_length;
474 uint8_t *inter_ac_vlc_length;
475 uint8_t *inter_ac_vlc_last_length;
476 uint8_t *luma_dc_vlc_length;
477 uint8_t *chroma_dc_vlc_length;
478 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
479
480 int coded_score[6];
481
482 /** precomputed matrix (combine qscale and DCT renorm) */
483 int (*q_intra_matrix)[64];
484 int (*q_inter_matrix)[64];
485 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
486 uint16_t (*q_intra_matrix16)[2][64];
487 uint16_t (*q_inter_matrix16)[2][64];
488 int block_last_index[12]; ///< last non zero coefficient in block
489 /* scantables */
490 ScanTable __align8 intra_scantable;
491 ScanTable intra_h_scantable;
492 ScanTable intra_v_scantable;
493 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
494
495 /* noise reduction */
496 int (*dct_error_sum)[64];
497 int dct_count[2];
498 uint16_t (*dct_offset)[64];
499
500 void *opaque; ///< private data for the user
501
502 /* bit rate control */
503 int64_t wanted_bits;
504 int64_t total_bits;
505 int frame_bits; ///< bits used for the current frame
506 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
507
508 /* statistics, used for 2-pass encoding */
509 int mv_bits;
510 int header_bits;
511 int i_tex_bits;
512 int p_tex_bits;
513 int i_count;
514 int f_count;
515 int b_count;
516 int skip_count;
517 int misc_bits; ///< cbp, mb_type
518 int last_bits; ///< temp var used for calculating the above vars
519
520 /* error concealment / resync */
521 int error_count;
522 uint8_t *error_status_table; ///< table of the error status of each MB
523 #define VP_START 1 ///< current MB is the first after a resync marker
524 #define AC_ERROR 2
525 #define DC_ERROR 4
526 #define MV_ERROR 8
527 #define AC_END 16
528 #define DC_END 32
529 #define MV_END 64
530 //FIXME some prefix?
531
532 int resync_mb_x; ///< x position of last resync marker
533 int resync_mb_y; ///< y position of last resync marker
534 GetBitContext last_resync_gb; ///< used to search for the next resync marker
535 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
536 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
537 int error_resilience;
538
539 ParseContext parse_context;
540
541 /* H.263 specific */
542 int gob_index;
543 int obmc; ///< overlapped block motion compensation
544
545 /* H.263+ specific */
546 int umvplus; ///< == H263+ && unrestricted_mv
547 int h263_aic; ///< Advanded INTRA Coding (AIC)
548 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
549 int h263_slice_structured;
550 int alt_inter_vlc; ///< alternative inter vlc
551 int modified_quant;
552 int loop_filter;
553 int custom_pcf;
554
555 /* mpeg4 specific */
556 int time_increment_bits; ///< number of bits to represent the fractional part of time
557 int last_time_base;
558 int time_base; ///< time in seconds of last I,P,S Frame
559 int64_t time; ///< time of current frame
560 int64_t last_non_b_time;
561 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
562 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
563 uint16_t pp_field_time;
564 uint16_t pb_field_time; ///< like above, just for interlaced
565 int shape;
566 int vol_sprite_usage;
567 int sprite_width;
568 int sprite_height;
569 int sprite_left;
570 int sprite_top;
571 int sprite_brightness_change;
572 int num_sprite_warping_points;
573 int real_sprite_warping_points;
574 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
575 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
576 int sprite_shift[2]; ///< sprite shift [isChroma]
577 int mcsel;
578 int quant_precision;
579 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
580 int scalability;
581 int hierachy_type;
582 int enhancement_type;
583 int new_pred;
584 int reduced_res_vop;
585 int aspect_ratio_info; //FIXME remove
586 int sprite_warping_accuracy;
587 int low_latency_sprite;
588 int data_partitioning; ///< data partitioning flag from header
589 int partitioned_frame; ///< is current frame partitioned
590 int rvlc; ///< reversible vlc
591 int resync_marker; ///< could this stream contain resync markers
592 int low_delay; ///< no reordering needed / has no b-frames
593 int vo_type;
594 int vol_control_parameters; ///< does the stream contain the low_delay flag, used to workaround buggy encoders
595 int intra_dc_threshold; ///< QP above whch the ac VLC should be used for intra dc
596 PutBitContext tex_pb; ///< used for data partitioned VOPs
597 PutBitContext pb2; ///< used for data partitioned VOPs
598 int mpeg_quant;
599 int t_frame; ///< time distance of first I -> B, used for interlaced b frames
600 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
601
602 /* divx specific, used to workaround (many) bugs in divx5 */
603 int divx_version;
604 int divx_build;
605 int divx_packed;
606 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
607 int bitstream_buffer_size;
608 int allocated_bitstream_buffer_size;
609
610 int xvid_build;
611
612 /* lavc specific stuff, used to workaround bugs in libavcodec */
613 int lavc_build;
614
615 /* RV10 specific */
616 int rv10_version; ///< RV10 version: 0 or 3
617 int rv10_first_dc_coded[3];
618
619 /* MJPEG specific */
620 struct MJpegContext *mjpeg_ctx;
621 int mjpeg_vsample[3]; ///< vertical sampling factors, default = {2, 1, 1}
622 int mjpeg_hsample[3]; ///< horizontal sampling factors, default = {2, 1, 1}
623 int mjpeg_write_tables; ///< do we want to have quantisation- and huffmantables in the jpeg file ?
624 int mjpeg_data_only_frames; ///< frames only with SOI, SOS and EOI markers
625
626 /* MSMPEG4 specific */
627 int mv_table_index;
628 int rl_table_index;
629 int rl_chroma_table_index;
630 int dc_table_index;
631 int use_skip_mb_code;
632 int slice_height; ///< in macroblocks
633 int first_slice_line; ///< used in mpeg4 too to handle resync markers
634 int flipflop_rounding;
635 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
636 int per_mb_rl_table;
637 int esc3_level_length;
638 int esc3_run_length;
639 /** [mb_intra][isChroma][level][run][last] */
640 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
641 int inter_intra_pred;
642 int mspel;
643
644 /* decompression specific */
645 GetBitContext gb;
646
647 /* Mpeg1 specific */
648 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
649 int last_mv_dir; ///< last mv_dir, used for b frame encoding
650 int broken_link; ///< no_output_of_prior_pics_flag
651 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
652
653 /* MPEG2 specific - I wish I had not to support this mess. */
654 int progressive_sequence;
655 int mpeg_f_code[2][2];
656 int picture_structure;
657 /* picture type */
658 #define PICT_TOP_FIELD 1
659 #define PICT_BOTTOM_FIELD 2
660 #define PICT_FRAME 3
661
662 int intra_dc_precision;
663 int frame_pred_frame_dct;
664 int top_field_first;
665 int concealment_motion_vectors;
666 int q_scale_type;
667 int intra_vlc_format;
668 int alternate_scan;
669 int repeat_first_field;
670 int chroma_420_type;
671 int chroma_format;
672 #define CHROMA_420 1
673 #define CHROMA_422 2
674 #define CHROMA_444 3
675 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
676 int chroma_y_shift;
677
678 int progressive_frame;
679 int full_pel[2];
680 int interlaced_dct;
681 int first_slice;
682 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
683
684 /* RTP specific */
685 int rtp_mode;
686
687 uint8_t *ptr_lastgob;
688 int swap_uv;//vcr2 codec is mpeg2 varint with UV swaped
689 short * pblocks[12];
690
691 DCTELEM (*block)[64]; ///< points to one of the following blocks
692 DCTELEM (*blocks)[6][64]; // for HQ mode we need to keep the best block
693 int (*decode_mb)(struct MpegEncContext *s, DCTELEM block[6][64]); // used by some codecs to avoid a switch()
694 #define SLICE_OK 0
695 #define SLICE_ERROR -1
696 #define SLICE_END -2 ///<end marker found
697 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
698
699 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
700 DCTELEM *block/*align 16*/, int n, int qscale);
701 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
702 DCTELEM *block/*align 16*/, int n, int qscale);
703 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
704 DCTELEM *block/*align 16*/, int n, int qscale);
705 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
706 DCTELEM *block/*align 16*/, int n, int qscale);
707 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
708 DCTELEM *block/*align 16*/, int n, int qscale);
709 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
710 DCTELEM *block/*align 16*/, int n, int qscale);
711 void (*dct_unquantize_h261_intra)(struct MpegEncContext *s,
712 DCTELEM *block/*align 16*/, int n, int qscale);
713 void (*dct_unquantize_h261_inter)(struct MpegEncContext *s,
714 DCTELEM *block/*align 16*/, int n, int qscale);
715 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
716 DCTELEM *block/*align 16*/, int n, int qscale);
717 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
718 DCTELEM *block/*align 16*/, int n, int qscale);
719 int (*dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
720 int (*fast_dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
721 void (*denoise_dct)(struct MpegEncContext *s, DCTELEM *block);
722 } MpegEncContext;
723
724
725 int DCT_common_init(MpegEncContext *s);
726 void MPV_decode_defaults(MpegEncContext *s);
727 int MPV_common_init(MpegEncContext *s);
728 void MPV_common_end(MpegEncContext *s);
729 void MPV_decode_mb(MpegEncContext *s, DCTELEM block[12][64]);
730 int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx);
731 void MPV_frame_end(MpegEncContext *s);
732 int MPV_encode_init(AVCodecContext *avctx);
733 int MPV_encode_end(AVCodecContext *avctx);
734 int MPV_encode_picture(AVCodecContext *avctx, unsigned char *buf, int buf_size, void *data);
735 #ifdef HAVE_MMX
736 void MPV_common_init_mmx(MpegEncContext *s);
737 #endif
738 #ifdef ARCH_ALPHA
739 void MPV_common_init_axp(MpegEncContext *s);
740 #endif
741 #ifdef HAVE_MLIB
742 void MPV_common_init_mlib(MpegEncContext *s);
743 #endif
744 #ifdef HAVE_MMI
745 void MPV_common_init_mmi(MpegEncContext *s);
746 #endif
747 #ifdef ARCH_ARMV4L
748 void MPV_common_init_armv4l(MpegEncContext *s);
749 #endif
750 #ifdef ARCH_POWERPC
751 void MPV_common_init_ppc(MpegEncContext *s);
752 #endif
753 extern void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
754 void ff_copy_bits(PutBitContext *pb, uint8_t *src, int length);
755 void ff_clean_intra_table_entries(MpegEncContext *s);
756 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
757 void ff_draw_horiz_band(MpegEncContext *s, int y, int h);
758 void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize, int block_w, int block_h,
759 int src_x, int src_y, int w, int h);
760 #define END_NOT_FOUND -100
761 int ff_combine_frame(ParseContext *pc, int next, uint8_t **buf, int *buf_size);
762 void ff_parse_close(AVCodecParserContext *s);
763 void ff_mpeg_flush(AVCodecContext *avctx);
764 void ff_print_debug_info(MpegEncContext *s, AVFrame *pict);
765 void ff_write_quant_matrix(PutBitContext *pb, int16_t *matrix);
766 int ff_find_unused_picture(MpegEncContext *s, int shared);
767 void ff_denoise_dct(MpegEncContext *s, DCTELEM *block);
768 void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
769
770 void ff_er_frame_start(MpegEncContext *s);
771 void ff_er_frame_end(MpegEncContext *s);
772 void ff_er_add_slice(MpegEncContext *s, int startx, int starty, int endx, int endy, int status);
773
774
775 extern enum PixelFormat ff_yuv420p_list[2];
776
777 void ff_init_block_index(MpegEncContext *s);
778
779 static inline void ff_update_block_index(MpegEncContext *s){
780 const int block_size= 8>>s->avctx->lowres;
781
782 s->block_index[0]+=2;
783 s->block_index[1]+=2;
784 s->block_index[2]+=2;
785 s->block_index[3]+=2;
786 s->block_index[4]++;
787 s->block_index[5]++;
788 s->dest[0]+= 2*block_size;
789 s->dest[1]+= block_size;
790 s->dest[2]+= block_size;
791 }
792
793 static inline int get_bits_diff(MpegEncContext *s){
794 const int bits= put_bits_count(&s->pb);
795 const int last= s->last_bits;
796
797 s->last_bits = bits;
798
799 return bits - last;
800 }
801
802 /* motion_est.c */
803 void ff_estimate_p_frame_motion(MpegEncContext * s,
804 int mb_x, int mb_y);
805 void ff_estimate_b_frame_motion(MpegEncContext * s,
806 int mb_x, int mb_y);
807 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
808 void ff_fix_long_p_mvs(MpegEncContext * s);
809 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
810 int16_t (*mv_table)[2], int f_code, int type, int truncate);
811 void ff_init_me(MpegEncContext *s);
812 int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
813 inline int ff_epzs_motion_search(MpegEncContext * s, int *mx_ptr, int *my_ptr,
814 int P[10][2], int src_index, int ref_index, int16_t (*last_mv)[2],
815 int ref_mv_scale, int size, int h);
816 int inline ff_get_mb_score(MpegEncContext * s, int mx, int my, int src_index,
817 int ref_index, int size, int h, int add_rate);
818
819 /* mpeg12.c */
820 extern const int16_t ff_mpeg1_default_intra_matrix[64];
821 extern const int16_t ff_mpeg1_default_non_intra_matrix[64];
822 extern const uint8_t ff_mpeg1_dc_scale_table[128];
823
824 void mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
825 void mpeg1_encode_mb(MpegEncContext *s,
826 DCTELEM block[6][64],
827 int motion_x, int motion_y);
828 void ff_mpeg1_encode_init(MpegEncContext *s);
829 void ff_mpeg1_encode_slice_header(MpegEncContext *s);
830 void ff_mpeg1_clean_buffers(MpegEncContext *s);
831 int ff_mpeg1_find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size);
832
833
834 /** RLTable. */
835 typedef struct RLTable {
836 int n; ///< number of entries of table_vlc minus 1
837 int last; ///< number of values for last = 0
838 const uint16_t (*table_vlc)[2];
839 const int8_t *table_run;
840 const int8_t *table_level;
841 uint8_t *index_run[2]; ///< encoding only
842 int8_t *max_level[2]; ///< encoding & decoding
843 int8_t *max_run[2]; ///< encoding & decoding
844 VLC vlc; ///< decoding only deprected FIXME remove
845 RL_VLC_ELEM *rl_vlc[32]; ///< decoding only
846 } RLTable;
847
848 void init_rl(RLTable *rl, int use_static);
849 void init_vlc_rl(RLTable *rl, int use_static);
850
851 static inline int get_rl_index(const RLTable *rl, int last, int run, int level)
852 {
853 int index;
854 index = rl->index_run[last][run];
855 if (index >= rl->n)
856 return rl->n;
857 if (level > rl->max_level[last][run])
858 return rl->n;
859 return index + level - 1;
860 }
861
862 extern const uint8_t ff_mpeg4_y_dc_scale_table[32];
863 extern const uint8_t ff_mpeg4_c_dc_scale_table[32];
864 extern const uint8_t ff_aic_dc_scale_table[32];
865 extern const int16_t ff_mpeg4_default_intra_matrix[64];
866 extern const int16_t ff_mpeg4_default_non_intra_matrix[64];
867 extern const uint8_t ff_h263_chroma_qscale_table[32];
868 extern const uint8_t ff_h263_loop_filter_strength[32];
869
870 /* h261.c */
871 void ff_h261_loop_filter(MpegEncContext *s);
872 void ff_h261_reorder_mb_index(MpegEncContext* s);
873 void ff_h261_encode_mb(MpegEncContext *s,
874 DCTELEM block[6][64],
875 int motion_x, int motion_y);
876 void ff_h261_encode_picture_header(MpegEncContext * s, int picture_number);
877 void ff_h261_encode_init(MpegEncContext *s);
878
879
880 /* h263.c, h263dec.c */
881 int ff_h263_decode_init(AVCodecContext *avctx);
882 int ff_h263_decode_frame(AVCodecContext *avctx,
883 void *data, int *data_size,
884 uint8_t *buf, int buf_size);
885 int ff_h263_decode_end(AVCodecContext *avctx);
886 void h263_encode_mb(MpegEncContext *s,
887 DCTELEM block[6][64],
888 int motion_x, int motion_y);
889 void mpeg4_encode_mb(MpegEncContext *s,
890 DCTELEM block[6][64],
891 int motion_x, int motion_y);
892 void h263_encode_picture_header(MpegEncContext *s, int picture_number);
893 void ff_flv_encode_picture_header(MpegEncContext *s, int picture_number);
894 void h263_encode_gob_header(MpegEncContext * s, int mb_line);
895 int16_t *h263_pred_motion(MpegEncContext * s, int block, int dir,
896 int *px, int *py);
897 void mpeg4_pred_ac(MpegEncContext * s, DCTELEM *block, int n,
898 int dir);
899 void ff_set_mpeg4_time(MpegEncContext * s, int picture_number);
900 void mpeg4_encode_picture_header(MpegEncContext *s, int picture_number);
901 void h263_encode_init(MpegEncContext *s);
902 void h263_decode_init_vlc(MpegEncContext *s);
903 int h263_decode_picture_header(MpegEncContext *s);
904 int ff_h263_decode_gob_header(MpegEncContext *s);
905 int ff_mpeg4_decode_picture_header(MpegEncContext * s, GetBitContext *gb);
906 void ff_h263_update_motion_val(MpegEncContext * s);
907 void ff_h263_loop_filter(MpegEncContext * s);
908 void ff_set_qscale(MpegEncContext * s, int qscale);
909 int ff_h263_decode_mba(MpegEncContext *s);
910 void ff_h263_encode_mba(MpegEncContext *s);
911
912 int intel_h263_decode_picture_header(MpegEncContext *s);
913 int flv_h263_decode_picture_header(MpegEncContext *s);
914 int ff_h263_decode_mb(MpegEncContext *s,
915 DCTELEM block[6][64]);
916 int ff_mpeg4_decode_mb(MpegEncContext *s,
917 DCTELEM block[6][64]);
918 int h263_get_picture_format(int width, int height);
919 void ff_mpeg4_encode_video_packet_header(MpegEncContext *s);
920 void ff_mpeg4_clean_buffers(MpegEncContext *s);
921 void ff_mpeg4_stuffing(PutBitContext * pbc);
922 void ff_mpeg4_init_partitions(MpegEncContext *s);
923 void ff_mpeg4_merge_partitions(MpegEncContext *s);
924 void ff_clean_mpeg4_qscales(MpegEncContext *s);
925 void ff_clean_h263_qscales(MpegEncContext *s);
926 int ff_mpeg4_decode_partitions(MpegEncContext *s);
927 int ff_mpeg4_get_video_packet_prefix_length(MpegEncContext *s);
928 int ff_h263_resync(MpegEncContext *s);
929 int ff_h263_get_gob_height(MpegEncContext *s);
930 int ff_mpeg4_set_direct_mv(MpegEncContext *s, int mx, int my);
931 int ff_h263_round_chroma(int x);
932 void ff_h263_encode_motion(MpegEncContext * s, int val, int f_code);
933 int ff_mpeg4_find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size);
934
935
936 /* rv10.c */
937 void rv10_encode_picture_header(MpegEncContext *s, int picture_number);
938 int rv_decode_dc(MpegEncContext *s, int n);
939 void rv20_encode_picture_header(MpegEncContext *s, int picture_number);
940
941
942 /* msmpeg4.c */
943 void msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number);
944 void msmpeg4_encode_ext_header(MpegEncContext * s);
945 void msmpeg4_encode_mb(MpegEncContext * s,
946 DCTELEM block[6][64],
947 int motion_x, int motion_y);
948 int msmpeg4_decode_picture_header(MpegEncContext * s);
949 int msmpeg4_decode_ext_header(MpegEncContext * s, int buf_size);
950 int ff_msmpeg4_decode_init(MpegEncContext *s);
951 void ff_msmpeg4_encode_init(MpegEncContext *s);
952 int ff_wmv2_decode_picture_header(MpegEncContext * s);
953 int ff_wmv2_decode_secondary_picture_header(MpegEncContext * s);
954 void ff_wmv2_add_mb(MpegEncContext *s, DCTELEM block[6][64], uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr);
955 void ff_mspel_motion(MpegEncContext *s,
956 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
957 uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
958 int motion_x, int motion_y, int h);
959 int ff_wmv2_encode_picture_header(MpegEncContext * s, int picture_number);
960 void ff_wmv2_encode_mb(MpegEncContext * s,
961 DCTELEM block[6][64],
962 int motion_x, int motion_y);
963
964 /* mjpeg.c */
965 int mjpeg_init(MpegEncContext *s);
966 void mjpeg_close(MpegEncContext *s);
967 void mjpeg_encode_mb(MpegEncContext *s,
968 DCTELEM block[6][64]);
969 void mjpeg_picture_header(MpegEncContext *s);
970 void mjpeg_picture_trailer(MpegEncContext *s);
971 void ff_mjpeg_stuffing(PutBitContext * pbc);
972
973
974 /* rate control */
975 int ff_rate_control_init(MpegEncContext *s);
976 float ff_rate_estimate_qscale(MpegEncContext *s, int dry_run);
977 void ff_write_pass1_stats(MpegEncContext *s);
978 void ff_rate_control_uninit(MpegEncContext *s);
979 double ff_eval(char *s, double *const_value, const char **const_name,
980 double (**func1)(void *, double), const char **func1_name,
981 double (**func2)(void *, double, double), char **func2_name,
982 void *opaque);
983 int ff_vbv_update(MpegEncContext *s, int frame_size);
984 void ff_get_2pass_fcode(MpegEncContext *s);
985
986 #endif /* AVCODEC_MPEGVIDEO_H */