2 * Generic DCT based hybrid video encoder
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
4 * Copyright (c) 2002-2004 Michael Niedermayer
6 * This file is part of FFmpeg.
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.
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.
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
28 #ifndef AVCODEC_MPEGVIDEO_H
29 #define AVCODEC_MPEGVIDEO_H
35 #include "error_resilience.h"
38 #include "h264chroma.h"
44 #include "motion_est.h"
45 #include "mpegpicture.h"
46 #include "mpegvideodsp.h"
47 #include "mpegvideoencdsp.h"
48 #include "pixblockdsp.h"
50 #include "ratecontrol.h"
52 #include "mpegutils.h"
53 #include "mpeg12data.h"
58 #include "libavutil/opt.h"
59 #include "libavutil/timecode.h"
61 #define MAX_THREADS 32
63 #define MAX_B_FRAMES 16
66 #define SEQ_END_CODE 0x000001b7
67 #define SEQ_START_CODE 0x000001b3
68 #define GOP_START_CODE 0x000001b8
69 #define PICTURE_START_CODE 0x00000100
70 #define SLICE_MIN_START_CODE 0x00000101
71 #define SLICE_MAX_START_CODE 0x000001af
72 #define EXT_START_CODE 0x000001b5
73 #define USER_START_CODE 0x000001b2
78 typedef struct MpegEncContext {
81 int y_dc_scale, c_dc_scale;
83 int block_last_index[12]; ///< last non zero coefficient in block
84 int h263_aic; ///< Advanced INTRA Coding (AIC)
87 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce the cache usage
88 ScanTable intra_scantable;
89 ScanTable intra_h_scantable;
90 ScanTable intra_v_scantable;
92 /* WARNING: changes above this line require updates to hardcoded
93 * offsets used in ASM. */
95 struct AVCodecContext *avctx;
96 /* the following parameters must be initialized before encoding */
97 int width, height;///< picture size. must be a multiple of 16
99 int intra_only; ///< if true, only intra pictures are generated
100 int64_t bit_rate; ///< wanted bit rate
101 enum OutputFormat out_format; ///< output format
102 int h263_pred; ///< use MPEG-4/H.263 ac/dc predictions
103 int pb_frame; ///< PB-frame mode (0 = none, 1 = base, 2 = improved)
105 /* the following codec id fields are deprecated in favor of codec_id */
106 int h263_plus; ///< H.263+ headers
107 int h263_flv; ///< use flv H.263 header
109 enum AVCodecID codec_id; /* see AV_CODEC_ID_xxx */
110 int fixed_qscale; ///< fixed qscale if non zero
111 int encoding; ///< true if we are encoding (vs decoding)
112 int max_b_frames; ///< max number of B-frames for encoding
113 int luma_elim_threshold;
114 int chroma_elim_threshold;
115 int strict_std_compliance; ///< strictly follow the std (MPEG-4, ...)
116 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
117 int codec_tag; ///< internal codec_tag upper case converted from avctx codec_tag
118 /* the following fields are managed internally by the encoder */
120 /* sequence parameters */
121 int context_initialized;
122 int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
123 int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
124 int picture_number; //FIXME remove, unclear definition
125 int picture_in_gop_number; ///< 0-> first pic in gop, ...
126 int mb_width, mb_height; ///< number of MBs horizontally & vertically
127 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
128 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
129 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
130 int mb_num; ///< number of MBs of a picture
131 ptrdiff_t linesize; ///< line size, in bytes, may be different from width
132 ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
133 Picture *picture; ///< main picture buffer
134 Picture **input_picture; ///< next pictures on display order for encoding
135 Picture **reordered_input_picture; ///< pointer to the next pictures in coded order for encoding
137 int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_encode_video2()
139 * pts difference between the first and second input frame, used for
140 * calculating dts of the first frame when there's a delay */
143 * reordered pts to be used as dts for the next output frame when there's
145 int64_t reordered_pts;
150 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
151 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
152 struct MpegEncContext *thread_context[MAX_THREADS];
153 int slice_context_count; ///< number of used thread_contexts
156 * copy of the previous picture structure.
157 * note, linesize & data, might not match the previous picture (for field pictures)
159 Picture last_picture;
162 * copy of the next picture structure.
163 * note, linesize & data, might not match the next picture (for field pictures)
165 Picture next_picture;
168 * copy of the source picture structure for encoding.
169 * note, linesize & data, might not match the source picture (for field pictures)
174 * copy of the current picture structure.
175 * note, linesize & data, might not match the current picture (for field pictures)
177 Picture current_picture; ///< buffer to store the decompressed current picture
179 Picture *last_picture_ptr; ///< pointer to the previous picture.
180 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
181 Picture *current_picture_ptr; ///< pointer to the current picture
182 int last_dc[3]; ///< last DC values for MPEG-1
183 int16_t *dc_val_base;
184 int16_t *dc_val[3]; ///< used for MPEG-4 DC prediction, all 3 arrays must be continuous
185 const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
186 const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
187 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (H.263)
188 uint8_t *coded_block_base;
189 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
190 int16_t (*ac_val_base)[16];
191 int16_t (*ac_val[3])[16]; ///< used for MPEG-4 AC prediction, all 3 arrays must be continuous
192 int mb_skipped; ///< MUST BE SET only during DECODING
193 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
194 and used for B-frame encoding & decoding (contains skip table of next P-frame) */
195 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
196 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
197 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
199 ScratchpadContext sc;
202 int chroma_qscale; ///< chroma QP
203 unsigned int lambda; ///< Lagrange multiplier used in rate distortion
204 unsigned int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
206 int adaptive_quant; ///< use adaptive quantization
207 int dquant; ///< qscale difference to prev qscale
208 int closed_gop; ///< MPEG1/2 GOP is closed
209 int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ...
211 int last_pict_type; //FIXME removes
212 int last_non_b_pict_type; ///< used for MPEG-4 gmc B-frames & ratecontrol
214 int frame_rate_index;
215 AVRational mpeg2_frame_rate_ext;
216 int last_lambda_for[5]; ///< last lambda for a specific pict type
217 int skipdct; ///< skip dct and code zero residual
219 /* motion compensation */
220 int unrestricted_mv; ///< mv can point outside of the coded picture
221 int h263_long_vectors; ///< use horrible H.263v1 long vector mode
223 BlockDSPContext bdsp;
225 H264ChromaContext h264chroma;
229 MpegVideoDSPContext mdsp;
230 MpegvideoEncDSPContext mpvencdsp;
231 PixblockDSPContext pdsp;
233 VideoDSPContext vdsp;
234 H263DSPContext h263dsp;
235 int f_code; ///< forward MV resolution
236 int b_code; ///< backward MV resolution for B-frames (MPEG-4)
237 int16_t (*p_mv_table_base)[2];
238 int16_t (*b_forw_mv_table_base)[2];
239 int16_t (*b_back_mv_table_base)[2];
240 int16_t (*b_bidir_forw_mv_table_base)[2];
241 int16_t (*b_bidir_back_mv_table_base)[2];
242 int16_t (*b_direct_mv_table_base)[2];
243 int16_t (*p_field_mv_table_base[2][2])[2];
244 int16_t (*b_field_mv_table_base[2][2][2])[2];
245 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) P-frame encoding
246 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode B-frame encoding
247 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode B-frame encoding
248 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode B-frame encoding
249 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode B-frame encoding
250 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode B-frame encoding
251 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced P-frame encoding
252 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced B-frame encoding
253 uint8_t (*p_field_select_table[2]);
254 uint8_t (*b_field_select_table[2][2]);
255 int motion_est; ///< ME algorithm
256 int me_penalty_compensation;
257 int me_pre; ///< prepass for motion estimation
259 #define MV_DIR_FORWARD 1
260 #define MV_DIR_BACKWARD 2
261 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (MPEG-4)
263 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
264 #define MV_TYPE_8X8 1 ///< 4 vectors (H.263, MPEG-4 4MV)
265 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
266 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
267 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
268 /**motion vectors for a macroblock
269 first coordinate : 0 = forward 1 = backward
270 second " : depend on type
271 third " : 0 = x, 1 = y
274 int field_select[2][2];
275 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG-1 & B-frame MPEG-4
276 uint8_t *fcode_tab; ///< smallest fcode needed for each MV
277 int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
281 int no_rounding; /**< apply no rounding to motion compensation (MPEG-4, msmpeg4, ...)
282 for B-frames rounding mode is always 0 */
284 /* macroblock layer */
288 uint16_t *mb_type; ///< Table for candidate MB types for encoding (defines in mpegutils.h)
290 int block_index[6]; ///< index to current MB in block based arrays with edges
294 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
296 /** matrix transmitted in the bitstream */
297 uint16_t intra_matrix[64];
298 uint16_t chroma_intra_matrix[64];
299 uint16_t inter_matrix[64];
300 uint16_t chroma_inter_matrix[64];
301 int force_duplicated_matrix; ///< Force duplication of mjpeg matrices, useful for rtp streaming
303 int intra_quant_bias; ///< bias for the quantizer
304 int inter_quant_bias; ///< bias for the quantizer
305 int min_qcoeff; ///< minimum encodable coefficient
306 int max_qcoeff; ///< maximum encodable coefficient
307 int ac_esc_length; ///< num of bits needed to encode the longest esc
308 uint8_t *intra_ac_vlc_length;
309 uint8_t *intra_ac_vlc_last_length;
310 uint8_t *intra_chroma_ac_vlc_length;
311 uint8_t *intra_chroma_ac_vlc_last_length;
312 uint8_t *inter_ac_vlc_length;
313 uint8_t *inter_ac_vlc_last_length;
314 uint8_t *luma_dc_vlc_length;
315 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
319 /** precomputed matrix (combine qscale and DCT renorm) */
320 int (*q_intra_matrix)[64];
321 int (*q_chroma_intra_matrix)[64];
322 int (*q_inter_matrix)[64];
323 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
324 uint16_t (*q_intra_matrix16)[2][64];
325 uint16_t (*q_chroma_intra_matrix16)[2][64];
326 uint16_t (*q_inter_matrix16)[2][64];
328 /* noise reduction */
329 int (*dct_error_sum)[64];
331 uint16_t (*dct_offset)[64];
333 /* bit rate control */
335 int frame_bits; ///< bits used for the current frame
336 int stuffing_bits; ///< bits used for stuffing
337 int next_lambda; ///< next lambda used for retrying to encode a frame
338 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
340 /* statistics, used for 2-pass encoding */
349 int misc_bits; ///< cbp, mb_type
350 int last_bits; ///< temp var used for calculating the above vars
352 /* error concealment / resync */
353 int resync_mb_x; ///< x position of last resync marker
354 int resync_mb_y; ///< y position of last resync marker
355 GetBitContext last_resync_gb; ///< used to search for the next resync marker
356 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
357 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed B-frames
359 ParseContext parse_context;
363 int obmc; ///< overlapped block motion compensation
364 int mb_info; ///< interval for outputting info about mb offsets as side data
365 int prev_mb_info, last_mb_info;
366 uint8_t *mb_info_ptr;
371 /* H.263+ specific */
372 int umvplus; ///< == H.263+ && unrestricted_mv
373 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
374 int h263_slice_structured;
375 int alt_inter_vlc; ///< alternative inter vlc
380 /* MPEG-4 specific */
381 ///< number of bits to represent the fractional part of time (encoder only)
382 int time_increment_bits;
384 int time_base; ///< time in seconds of last I,P,S Frame
385 int64_t time; ///< time of current frame
386 int64_t last_non_b_time;
387 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
388 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
389 uint16_t pp_field_time;
390 uint16_t pb_field_time; ///< like above, just for interlaced
391 int real_sprite_warping_points;
392 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
393 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
396 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
397 int aspect_ratio_info; //FIXME remove
398 int sprite_warping_accuracy;
399 int data_partitioning; ///< data partitioning flag from header
400 int partitioned_frame; ///< is current frame partitioned
401 int low_delay; ///< no reordering needed / has no B-frames
403 PutBitContext tex_pb; ///< used for data partitioned VOPs
404 PutBitContext pb2; ///< used for data partitioned VOPs
406 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG-4
408 /* divx specific, used to workaround (many) bugs in divx5 */
410 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
411 int bitstream_buffer_size;
412 unsigned int allocated_bitstream_buffer_size;
415 int rv10_version; ///< RV10 version: 0 or 3
416 int rv10_first_dc_coded[3];
419 struct MJpegContext *mjpeg_ctx;
424 /* MSMPEG4 specific */
427 int rl_chroma_table_index;
429 int use_skip_mb_code;
430 int slice_height; ///< in macroblocks
431 int first_slice_line; ///< used in MPEG-4 too to handle resync markers
432 int flipflop_rounding;
433 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
435 int esc3_level_length;
437 /** [mb_intra][isChroma][level][run][last] */
438 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
439 int inter_intra_pred;
442 /* decompression specific */
445 /* MPEG-1 specific */
446 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & MPEG-1 specific
447 int last_mv_dir; ///< last mv_dir, used for B-frame encoding
448 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
450 /* MPEG-2-specific - I wished not to have to support this mess. */
451 int progressive_sequence;
452 int mpeg_f_code[2][2];
454 // picture structure defines are loaded from mpegutils.h
455 int picture_structure;
457 int64_t timecode_frame_start; ///< GOP timecode frame start number, in non drop frame format
458 int intra_dc_precision;
459 int frame_pred_frame_dct;
461 int concealment_motion_vectors;
464 int intra_vlc_format;
467 int repeat_first_field;
473 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
476 int progressive_frame;
479 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
480 int drop_frame_timecode; ///< timecode is in drop frame format.
481 int scan_offset; ///< reserve space for SVCD scan offset user data.
485 int rtp_payload_size;
487 char *tc_opt_str; ///< timecode option string
488 AVTimecode tc; ///< timecode context
490 uint8_t *ptr_lastgob;
491 int swap_uv; //vcr2 codec is an MPEG-2 variant with U and V swapped
492 int pack_pblocks; //xvmc needs to keep blocks without gaps.
493 int16_t (*pblocks[12])[64];
495 int16_t (*block)[64]; ///< points to one of the following blocks
496 int16_t (*blocks)[12][64]; // for HQ mode we need to keep the best block
497 int (*decode_mb)(struct MpegEncContext *s, int16_t block[6][64]); // used by some codecs to avoid a switch()
499 #define SLICE_ERROR -1
500 #define SLICE_END -2 ///<end marker found
501 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
503 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
504 int16_t *block/*align 16*/, int n, int qscale);
505 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
506 int16_t *block/*align 16*/, int n, int qscale);
507 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
508 int16_t *block/*align 16*/, int n, int qscale);
509 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
510 int16_t *block/*align 16*/, int n, int qscale);
511 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
512 int16_t *block/*align 16*/, int n, int qscale);
513 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
514 int16_t *block/*align 16*/, int n, int qscale);
515 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (MPEG-4 can use both)
516 int16_t *block/*align 16*/, int n, int qscale);
517 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (MPEG-4 can use both)
518 int16_t *block/*align 16*/, int n, int qscale);
519 int (*dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
520 int (*fast_dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
521 void (*denoise_dct)(struct MpegEncContext *s, int16_t *block);
523 int mpv_flags; ///< flags set by private options
524 int quantizer_noise_shaping;
527 * ratecontrol qmin qmax limiting method
528 * 0-> clipping, 1-> use a nice continuous function to limit qscale within qmin/qmax.
533 float rc_initial_cplx;
534 float rc_buffer_aggressivity;
535 float border_masking;
541 /* temp buffers for rate control */
542 float *cplx_tab, *bits_tab;
544 /* flag to indicate a reinitialization is required, e.g. after
545 * a frame size change */
552 /* temporary frames used by b_frame_strategy = 2 */
553 AVFrame *tmp_frames[MAX_B_FRAMES + 2];
554 int b_frame_strategy;
557 /* frame skip options for encoding */
558 int frame_skip_threshold;
559 int frame_skip_factor;
563 int scenechange_threshold;
567 /* mpegvideo_enc common options */
568 #define FF_MPV_FLAG_SKIP_RD 0x0001
569 #define FF_MPV_FLAG_STRICT_GOP 0x0002
570 #define FF_MPV_FLAG_QP_RD 0x0004
571 #define FF_MPV_FLAG_CBP_RD 0x0008
572 #define FF_MPV_FLAG_NAQ 0x0010
573 #define FF_MPV_FLAG_MV0 0x0020
576 MPV_RC_STRATEGY_FFMPEG,
577 MPV_RC_STRATEGY_XVID,
581 #define FF_MPV_OPT_CMP_FUNC \
582 { "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" }, \
583 { "sse", "Sum of squared errors", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_SSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
584 { "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" }, \
585 { "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" }, \
586 { "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" }, \
587 { "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" }, \
588 { "rd", "Rate distortion optimal, slow", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_RD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
589 { "zero", "Zero", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_ZERO }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
590 { "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" }, \
591 { "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" }, \
592 { "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" }, \
593 { "dct264", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCT264 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
594 { "dctmax", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCTMAX }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
595 { "chroma", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_CHROMA }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
596 { "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" }
598 #ifndef FF_MPV_OFFSET
599 #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x)
601 #define FF_MPV_OPT_FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
602 #define FF_MPV_COMMON_OPTS \
603 FF_MPV_OPT_CMP_FUNC, \
604 { "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" },\
605 { "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" },\
606 { "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" },\
607 { "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" },\
608 { "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" },\
609 { "naq", "normalize adaptive quantization", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_NAQ }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
610 { "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" },\
611 { "luma_elim_threshold", "single coefficient elimination threshold for luminance (negative values also consider dc coefficient)",\
612 FF_MPV_OFFSET(luma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
613 { "chroma_elim_threshold", "single coefficient elimination threshold for chrominance (negative values also consider dc coefficient)",\
614 FF_MPV_OFFSET(chroma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
615 { "quantizer_noise_shaping", NULL, FF_MPV_OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
616 { "error_rate", "Simulate errors in the bitstream to test error concealment.", \
617 FF_MPV_OFFSET(error_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
618 {"qsquish", "how to keep quantizer between qmin and qmax (0 = clip, 1 = use differentiable function)", \
619 FF_MPV_OFFSET(rc_qsquish), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, 0, 99, FF_MPV_OPT_FLAGS}, \
620 {"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}, \
621 {"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}, \
622 {"rc_eq", "Set rate control equation. When computing the expression, besides the standard functions " \
623 "defined in the section 'Expression Evaluation', the following functions are available: " \
624 "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv " \
625 "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.", \
626 FF_MPV_OFFSET(rc_eq), AV_OPT_TYPE_STRING, .flags = FF_MPV_OPT_FLAGS }, \
627 {"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}, \
628 {"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}, \
629 {"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}, \
630 {"lmin", "minimum Lagrange factor (VBR)", FF_MPV_OFFSET(lmin), AV_OPT_TYPE_INT, {.i64 = 2*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
631 {"lmax", "maximum Lagrange factor (VBR)", FF_MPV_OFFSET(lmax), AV_OPT_TYPE_INT, {.i64 = 31*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
632 {"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 }, \
633 {"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 }, \
634 {"rc_strategy", "ratecontrol method", FF_MPV_OFFSET(rc_strategy), AV_OPT_TYPE_INT, {.i64 = MPV_RC_STRATEGY_FFMPEG }, 0, NB_MPV_RC_STRATEGY-1, FF_MPV_OPT_FLAGS, "rc_strategy" }, \
635 { "ffmpeg", "default native rate control", 0, AV_OPT_TYPE_CONST, { .i64 = MPV_RC_STRATEGY_FFMPEG }, 0, 0, FF_MPV_OPT_FLAGS, "rc_strategy" }, \
636 { "xvid", "libxvid (2 pass only)", 0, AV_OPT_TYPE_CONST, { .i64 = MPV_RC_STRATEGY_XVID }, 0, 0, FF_MPV_OPT_FLAGS, "rc_strategy" }, \
637 {"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" }, \
638 { "zero", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_ZERO }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
639 { "epzs", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_EPZS }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
640 { "xone", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_XONE }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
641 { "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 }, \
642 {"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 }, \
643 {"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 }, \
644 {"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 }, \
645 {"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 }, \
646 {"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 }, \
647 {"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 }, \
648 {"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" }, \
649 {"sc_threshold", "Scene change threshold", FF_MPV_OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
650 {"noise_reduction", "Noise reduction", FF_MPV_OFFSET(noise_reduction), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
651 {"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 }, \
652 {"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 }, \
653 {"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 }, \
654 {"mepre", "pre motion estimation", FF_MPV_OFFSET(me_pre), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
656 extern const AVOption ff_mpv_generic_options[];
659 * Set the given MpegEncContext to common defaults (same for encoding
660 * and decoding). The changed fields will not depend upon the prior
661 * state of the MpegEncContext.
663 void ff_mpv_common_defaults(MpegEncContext *s);
665 void ff_dct_encode_init_x86(MpegEncContext *s);
667 int ff_mpv_common_init(MpegEncContext *s);
668 void ff_mpv_common_init_arm(MpegEncContext *s);
669 void ff_mpv_common_init_axp(MpegEncContext *s);
670 void ff_mpv_common_init_neon(MpegEncContext *s);
671 void ff_mpv_common_init_ppc(MpegEncContext *s);
672 void ff_mpv_common_init_x86(MpegEncContext *s);
673 void ff_mpv_common_init_mips(MpegEncContext *s);
675 int ff_mpv_common_frame_size_change(MpegEncContext *s);
676 void ff_mpv_common_end(MpegEncContext *s);
678 void ff_mpv_decode_defaults(MpegEncContext *s);
679 void ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx);
680 void ff_mpv_reconstruct_mb(MpegEncContext *s, int16_t block[12][64]);
681 void ff_mpv_report_decode_progress(MpegEncContext *s);
683 int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx);
684 void ff_mpv_frame_end(MpegEncContext *s);
686 int ff_mpv_encode_init(AVCodecContext *avctx);
687 void ff_mpv_encode_init_x86(MpegEncContext *s);
689 int ff_mpv_encode_end(AVCodecContext *avctx);
690 int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
691 const AVFrame *frame, int *got_packet);
692 int ff_mpv_reallocate_putbitbuffer(MpegEncContext *s, size_t threshold, size_t size_increase);
694 void ff_clean_intra_table_entries(MpegEncContext *s);
695 void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h);
696 void ff_mpeg_flush(AVCodecContext *avctx);
698 void ff_print_debug_info(MpegEncContext *s, Picture *p, AVFrame *pict);
699 void ff_print_debug_info2(AVCodecContext *avctx, AVFrame *pict, uint8_t *mbskip_table,
700 uint32_t *mbtype_table, int8_t *qscale_table, int16_t (*motion_val[2])[2],
702 int mb_width, int mb_height, int mb_stride, int quarter_sample);
704 int ff_mpv_export_qp_table(MpegEncContext *s, AVFrame *f, Picture *p, int qp_type);
706 void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
708 int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
709 int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
710 void ff_set_qscale(MpegEncContext * s, int qscale);
712 void ff_mpv_idct_init(MpegEncContext *s);
713 int ff_dct_encode_init(MpegEncContext *s);
714 void ff_convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16)[2][64],
715 const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra);
716 int ff_dct_quantize_c(MpegEncContext *s, int16_t *block, int n, int qscale, int *overflow);
717 void ff_block_permute(int16_t *block, uint8_t *permutation,
718 const uint8_t *scantable, int last);
719 void ff_init_block_index(MpegEncContext *s);
721 void ff_mpv_motion(MpegEncContext *s,
722 uint8_t *dest_y, uint8_t *dest_cb,
723 uint8_t *dest_cr, int dir,
724 uint8_t **ref_picture,
725 op_pixels_func (*pix_op)[4],
726 qpel_mc_func (*qpix_op)[16]);
728 static inline void ff_update_block_index(MpegEncContext *s){
729 const int block_size= 8 >> s->avctx->lowres;
731 s->block_index[0]+=2;
732 s->block_index[1]+=2;
733 s->block_index[2]+=2;
734 s->block_index[3]+=2;
737 s->dest[0]+= 2*block_size;
738 s->dest[1]+= block_size;
739 s->dest[2]+= block_size;
742 static inline int get_bits_diff(MpegEncContext *s){
743 const int bits= put_bits_count(&s->pb);
744 const int last= s->last_bits;
751 #endif /* AVCODEC_MPEGVIDEO_H */