AOMedia AV1 Codec
aom_encoder.h
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2016, Alliance for Open Media. All rights reserved.
3  *
4  * This source code is subject to the terms of the BSD 2 Clause License and
5  * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6  * was not distributed with this source code in the LICENSE file, you can
7  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8  * Media Patent License 1.0 was not distributed with this source code in the
9  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10  */
11 #ifndef AOM_AOM_AOM_ENCODER_H_
12 #define AOM_AOM_AOM_ENCODER_H_
13 
29 #ifdef __cplusplus
30 extern "C" {
31 #endif
32 
33 #include "aom/aom_codec.h" // IWYU pragma: export
35 
54 #define AOM_ENCODER_ABI_VERSION \
55  (10 + AOM_CODEC_ABI_VERSION + /*AOM_EXT_PART_ABI_VERSION=*/3)
56 
66 #define AOM_CODEC_CAP_PSNR 0x10000
70 #define AOM_CODEC_CAP_HIGHBITDEPTH 0x40000
71 
79 #define AOM_CODEC_USE_PSNR 0x10000
80 #define AOM_CODEC_USE_HIGHBITDEPTH 0x40000
81 // 0x80000 was used for the experimental feature AOM_CODEC_USE_PRESET during
82 // libaom v3.11.0 development but was removed before the release.
83 
88 typedef struct aom_fixed_buf {
89  void *buf;
90  size_t sz;
99 typedef uint32_t aom_codec_er_flags_t;
101 #define AOM_ERROR_RESILIENT_DEFAULT 0x1
102 
115 };
116 
122 typedef struct aom_codec_cx_pkt {
124  union {
125  struct {
126  void *buf;
127  size_t sz;
131  unsigned long duration;
139  } frame;
142  struct aom_psnr_pkt {
143  unsigned int samples[4];
144  uint64_t sse[4];
145  double psnr[4];
148  unsigned int samples_hbd[4];
151  uint64_t sse_hbd[4];
154  double psnr_hbd[4];
155  } psnr;
157  } data;
164 typedef struct aom_rational {
165  int num;
166  int den;
167 } aom_rational_t;
182 };
183 
190 };
191 
204 };
205 
207 typedef enum {
219  AOM_SUPERRES_AUTO,
221 
227 typedef struct cfg_options {
231  unsigned int init_by_cfg_file;
235  unsigned int super_block_size;
239  unsigned int max_partition_size;
243  unsigned int min_partition_size;
259  unsigned int disable_flip_idtx;
263  unsigned int disable_cdef;
267  unsigned int disable_lr;
271  unsigned int disable_obmc;
275  unsigned int disable_warp_motion;
279  unsigned int disable_global_motion;
283  unsigned int disable_dist_wtd_comp;
287  unsigned int disable_diff_wtd_comp;
295  unsigned int disable_masked_comp;
303  unsigned int disable_palette;
307  unsigned int disable_intrabc;
311  unsigned int disable_cfl;
315  unsigned int disable_smooth_intra;
319  unsigned int disable_filter_intra;
323  unsigned int disable_dual_filter;
335  unsigned int disable_tx_64x64;
351  unsigned int disable_paeth_intra;
355  unsigned int disable_trellis_quant;
359  unsigned int disable_ref_frame_mv;
363  unsigned int reduced_reference_set;
367  unsigned int reduced_tx_type_set;
368 } cfg_options_t;
369 
379 #define AOM_EFLAG_FORCE_KF (1 << 0)
380 
381 #define AOM_EFLAG_CALCULATE_PSNR (1 << 1)
382 
384 #define AOM_EFLAG_FREEZE_INTERNAL_STATE (1 << 2)
385 
392 typedef struct aom_codec_enc_cfg {
393  /*
394  * generic settings (g)
395  */
396 
404  unsigned int g_usage;
405 
412  unsigned int g_threads;
413 
422  unsigned int g_profile;
431  unsigned int g_w;
432 
440  unsigned int g_h;
441 
450  unsigned int g_limit;
451 
458 
465 
473 
484  unsigned int g_input_bit_depth;
485 
499 
506  aom_codec_er_flags_t g_error_resilient;
507 
514 
527  unsigned int g_lag_in_frames;
528 
529  /*
530  * rate control settings (rc)
531  */
532 
549  unsigned int rc_dropframe_thresh;
550 
558  unsigned int rc_resize_mode;
559 
566  unsigned int rc_resize_denominator;
567 
575 
583 
595 
605 
613  unsigned int rc_superres_qthresh;
614 
623 
633 
640 
647 
653  unsigned int rc_target_bitrate;
654 
655  /*
656  * quantizer settings
657  */
658 
667  unsigned int rc_min_quantizer;
668 
677  unsigned int rc_max_quantizer;
678 
679  /*
680  * bitrate tolerance
681  */
682 
690  unsigned int rc_undershoot_pct;
691 
699  unsigned int rc_overshoot_pct;
700 
701  /*
702  * decoder buffer model parameters
703  */
704 
714  unsigned int rc_buf_sz;
715 
723  unsigned int rc_buf_initial_sz;
724 
732  unsigned int rc_buf_optimal_sz;
733 
734  /*
735  * 2 pass rate control parameters
736  */
737 
746  unsigned int rc_2pass_vbr_bias_pct;
747 
754 
761 
762  /*
763  * keyframing settings (kf)
764  */
765 
770 
778 
786  unsigned int kf_min_dist;
787 
795  unsigned int kf_max_dist;
796 
802  unsigned int sframe_dist;
803 
819  unsigned int sframe_mode;
820 
827  unsigned int large_scale_tile;
828 
834  unsigned int monochrome;
835 
846 
853  unsigned int save_as_annexb;
854 
862 
870 
875 #define MAX_TILE_WIDTHS 64 // maximum tile width array length
876 
883 
888 #define MAX_TILE_HEIGHTS 64 // maximum tile height array length
889 
896 
907  unsigned int use_fixed_qp_offsets;
908 
914 
948  aom_codec_iface_t *iface,
949  const aom_codec_enc_cfg_t *cfg,
950  aom_codec_flags_t flags, int ver);
951 
956 #define aom_codec_enc_init(ctx, iface, cfg, flags) \
957  aom_codec_enc_init_ver(ctx, iface, cfg, flags, AOM_ENCODER_ABI_VERSION)
958 
981  aom_codec_enc_cfg_t *cfg,
982  unsigned int usage);
983 
999  const aom_codec_enc_cfg_t *cfg);
1000 
1025 
1027 #define AOM_USAGE_GOOD_QUALITY 0u
1028 
1029 #define AOM_USAGE_REALTIME 1u
1030 
1031 #define AOM_USAGE_ALL_INTRA 2u
1032 
1072  aom_codec_pts_t pts, unsigned long duration,
1073  aom_enc_frame_flags_t flags);
1074 
1119  const aom_fixed_buf_t *buf,
1120  unsigned int pad_before,
1121  unsigned int pad_after);
1122 
1147  aom_codec_iter_t *iter);
1148 
1162 
1164 #ifdef __cplusplus
1165 }
1166 #endif
1167 #endif // AOM_AOM_AOM_ENCODER_H_
Definition: aom_encoder.h:202
void * buf
Definition: aom_encoder.h:89
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: aom_encoder.h:795
aom_codec_err_t aom_codec_enc_init_ver(aom_codec_ctx_t *ctx, aom_codec_iface_t *iface, const aom_codec_enc_cfg_t *cfg, aom_codec_flags_t flags, int ver)
Initialize an encoder instance.
unsigned int min_partition_size
min partition size 8, 16, 32, 64, 128
Definition: aom_encoder.h:243
unsigned int rc_superres_qthresh
Frame super-resolution q threshold.
Definition: aom_encoder.h:613
Definition: aom_encoder.h:181
unsigned int g_w
Width of the frame.
Definition: aom_encoder.h:431
aom_superres_mode rc_superres_mode
Frame super-resolution scaling mode.
Definition: aom_encoder.h:582
unsigned int init_by_cfg_file
Indicate init by cfg file 0 or 1.
Definition: aom_encoder.h:231
uint64_t sse[4]
Definition: aom_encoder.h:144
unsigned int disable_smooth_inter_intra
disable smooth inter/intra
Definition: aom_encoder.h:339
unsigned int rc_target_bitrate
Target data rate.
Definition: aom_encoder.h:653
uint32_t aom_codec_er_flags_t
Error Resilient flags.
Definition: aom_encoder.h:99
int64_t aom_codec_pts_t
Time Stamp Type.
Definition: aom_codec.h:252
Definition: aom_encoder.h:209
unsigned int disable_paeth_intra
disable paeth intra
Definition: aom_encoder.h:351
unsigned int disable_cfl
disable chroma from luma
Definition: aom_encoder.h:311
unsigned int rc_superres_denominator
Frame super-resolution denominator.
Definition: aom_encoder.h:594
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition: aom_encoder.h:732
Encoder configuration structure.
Definition: aom_encoder.h:392
unsigned int samples[4]
Definition: aom_encoder.h:143
enum aom_kf_mode kf_mode
Keyframe placement mode.
Definition: aom_encoder.h:777
unsigned int disable_intra_angle_delta
disable intra angle delta
Definition: aom_encoder.h:327
aom_fixed_buf_t rc_twopass_stats_in
Two-pass stats buffer.
Definition: aom_encoder.h:639
aom_rc_mode
Rate control mode.
Definition: aom_encoder.h:185
struct aom_codec_cx_pkt aom_codec_cx_pkt_t
Encoder output packet.
unsigned int disable_tx_64x64
disable 64x64 transform
Definition: aom_encoder.h:335
aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface, aom_codec_enc_cfg_t *cfg, unsigned int usage)
Get the default configuration for a usage.
aom_fixed_buf_t rc_firstpass_mb_stats_in
first pass mb stats buffer.
Definition: aom_encoder.h:646
Definition: aom_encoder.h:187
aom_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: aom_encoder.h:506
unsigned int disable_masked_comp
disable masked compound
Definition: aom_encoder.h:295
Rational Number.
Definition: aom_encoder.h:164
unsigned int super_block_size
Superblock size 0, 64 or 128.
Definition: aom_encoder.h:235
Codec context structure.
Definition: aom_codec.h:315
unsigned int rc_2pass_vbr_minsection_pct
Two-pass mode per-GOP minimum bitrate.
Definition: aom_encoder.h:753
unsigned int disable_1to4_partition_type
disable 1:4/4:1 partition type
Definition: aom_encoder.h:255
Definition: aom_encoder.h:213
const aom_image_t * aom_codec_get_preview_frame(aom_codec_ctx_t *ctx)
Get Preview Frame.
aom_codec_pts_t pts
time stamp to show frame (in timebase units)
Definition: aom_encoder.h:129
Image Descriptor.
Definition: aom_image.h:211
double psnr[4]
Definition: aom_encoder.h:145
unsigned int rc_undershoot_pct
Rate control adaptation undershoot control.
Definition: aom_encoder.h:690
aom_fixed_buf_t raw
Definition: aom_encoder.h:156
aom_kf_mode
Keyframe placement mode.
Definition: aom_encoder.h:200
aom_codec_err_t aom_codec_set_cx_data_buf(aom_codec_ctx_t *ctx, const aom_fixed_buf_t *buf, unsigned int pad_before, unsigned int pad_after)
Set compressed data output buffer.
enum aom_rc_mode rc_end_usage
Rate control algorithm to use.
Definition: aom_encoder.h:632
Generic fixed size buffer structure.
Definition: aom_encoder.h:88
unsigned int g_profile
Bitstream profile to use.
Definition: aom_encoder.h:422
const aom_codec_cx_pkt_t * aom_codec_get_cx_data(aom_codec_ctx_t *ctx, aom_codec_iter_t *iter)
Encoded data iterator.
aom_codec_err_t aom_codec_encode(aom_codec_ctx_t *ctx, const aom_image_t *img, aom_codec_pts_t pts, unsigned long duration, aom_enc_frame_flags_t flags)
Encode a frame.
unsigned int max_partition_size
max partition size 8, 16, 32, 64, 128
Definition: aom_encoder.h:239
unsigned int disable_diff_wtd_comp
disable diff weighted compound
Definition: aom_encoder.h:287
const struct aom_codec_iface aom_codec_iface_t
Codec interface structure.
Definition: aom_codec.h:271
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition: aom_encoder.h:723
aom_codec_cx_pkt_kind
Encoder output packet variants.
Definition: aom_encoder.h:109
unsigned int samples_hbd[4]
Number of samples, total/y/u/v when input bit-depth < stream bit-depth.
Definition: aom_encoder.h:148
struct aom_rational g_timebase
Stream timebase units.
Definition: aom_encoder.h:498
unsigned int g_limit
Max number of frames to encode.
Definition: aom_encoder.h:450
Definition: aom_encoder.h:113
Definition: aom_encoder.h:111
Encoder Config Options.
Definition: aom_encoder.h:227
size_t sz
Definition: aom_encoder.h:127
unsigned int disable_ref_frame_mv
disable ref frame MV
Definition: aom_encoder.h:359
enum aom_codec_cx_pkt_kind kind
Definition: aom_encoder.h:123
Describes the codec algorithm interface to applications.
unsigned int disable_global_motion
disable global motion
Definition: aom_encoder.h:279
unsigned int rc_2pass_vbr_maxsection_pct
Two-pass mode per-GOP maximum bitrate.
Definition: aom_encoder.h:760
unsigned int disable_inter_intra_wedge
disable inter/intra wedge comp
Definition: aom_encoder.h:347
unsigned int g_forced_max_frame_width
Forced maximum width of the frame.
Definition: aom_encoder.h:457
double psnr_hbd[4]
PSNR, total/y/u/v when input bit-depth < stream bit-depth.
Definition: aom_encoder.h:154
int fwd_kf_enabled
Option to enable forward reference key frame.
Definition: aom_encoder.h:769
unsigned long duration
duration to show frame (in timebase units)
Definition: aom_encoder.h:131
unsigned int disable_trellis_quant
disable trellis quantization
Definition: aom_encoder.h:355
unsigned int sframe_mode
sframe insertion mode
Definition: aom_encoder.h:819
unsigned int g_forced_max_frame_height
Forced maximum height of the frame.
Definition: aom_encoder.h:464
Definition: aom_encoder.h:177
struct aom_codec_cx_pkt::@1::@2 frame
unsigned int disable_rect_partition_type
disable rectangular partition type
Definition: aom_encoder.h:251
unsigned int disable_obmc
disable OBMC
Definition: aom_encoder.h:271
Definition: aom_encoder.h:110
unsigned int rc_2pass_vbr_bias_pct
Two-pass mode CBR/VBR bias.
Definition: aom_encoder.h:746
unsigned int rc_resize_mode
Mode for spatial resampling, if supported by the codec.
Definition: aom_encoder.h:558
unsigned int use_fixed_qp_offsets
Controls how the encoder applies fixed QP offsets.
Definition: aom_encoder.h:907
struct aom_fixed_buf aom_fixed_buf_t
Generic fixed size buffer structure.
unsigned int disable_warp_motion
disable Warped Motion
Definition: aom_encoder.h:275
void * buf
Definition: aom_encoder.h:126
#define MAX_TILE_WIDTHS
Maximum number of tile widths in tile widths array.
Definition: aom_encoder.h:875
unsigned int rc_max_quantizer
Maximum (Worst Quality) Quantizer.
Definition: aom_encoder.h:677
unsigned int disable_intra_edge_filter
disable intra edge filter
Definition: aom_encoder.h:331
#define MAX_TILE_HEIGHTS
Maximum number of tile heights in tile heights array.
Definition: aom_encoder.h:888
unsigned int rc_buf_sz
Decoder Buffer Size.
Definition: aom_encoder.h:714
Definition: aom_encoder.h:112
Definition: aom_encoder.h:188
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition: aom_encoder.h:699
uint64_t sse_hbd[4]
sum squared error, total/y/u/v when input bit-depth < stream bit-depth.
Definition: aom_encoder.h:151
Definition: aom_encoder.h:186
unsigned int large_scale_tile
Tile coding mode.
Definition: aom_encoder.h:827
aom_enc_pass
Multi-pass Encoding Pass.
Definition: aom_encoder.h:176
uint32_t aom_codec_frame_flags_t
Compressed Frame Flags.
Definition: aom_codec.h:287
unsigned int disable_filter_intra
disable filter intra
Definition: aom_encoder.h:319
unsigned int disable_dual_filter
disable dual filter
Definition: aom_encoder.h:323
aom_fixed_buf_t * aom_codec_get_global_headers(aom_codec_ctx_t *ctx)
Get global stream headers.
size_t vis_frame_size
size of the visible frame in this packet
Definition: aom_encoder.h:138
aom_codec_err_t aom_codec_enc_config_set(aom_codec_ctx_t *ctx, const aom_codec_enc_cfg_t *cfg)
Set or change configuration.
unsigned int rc_resize_kf_denominator
Keyframe resize denominator.
Definition: aom_encoder.h:574
int tile_widths[64]
Array of specified tile widths.
Definition: aom_encoder.h:882
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: aom_encoder.h:786
unsigned int full_still_picture_hdr
full_still_picture_hdr
Definition: aom_encoder.h:845
unsigned int monochrome
Monochrome mode.
Definition: aom_encoder.h:834
unsigned int disable_inter_inter_wedge
disable inter/inter wedge comp
Definition: aom_encoder.h:343
unsigned int disable_lr
disable Loop Restoration Filter
Definition: aom_encoder.h:267
const void * aom_codec_iter_t
Iterator.
Definition: aom_codec.h:305
unsigned int disable_palette
disable Palette
Definition: aom_encoder.h:303
unsigned int rc_resize_denominator
Frame resize denominator.
Definition: aom_encoder.h:566
Definition: aom_encoder.h:216
struct aom_rational aom_rational_t
Rational Number.
struct aom_codec_enc_cfg aom_codec_enc_cfg_t
Encoder configuration structure.
int fixed_qp_offsets[5]
Deprecated and ignored. DO NOT USE.
Definition: aom_encoder.h:913
long aom_codec_flags_t
Initialization-time Feature Enabling.
Definition: aom_codec.h:232
int partition_id
the partition id defines the decoding order of the partitions. Only applicable when "output partition...
Definition: aom_encoder.h:136
Definition: aom_encoder.h:114
enum aom_bit_depth aom_bit_depth_t
Bit depth for codecThis enumeration determines the bit depth of the codec.
unsigned int g_usage
Algorithm specific "usage" value.
Definition: aom_encoder.h:404
unsigned int g_input_bit_depth
Bit-depth of the input source.
Definition: aom_encoder.h:484
cfg_options_t encoder_cfg
Options defined per config file.
Definition: aom_encoder.h:918
aom_codec_err_t
Algorithm return codes.
Definition: aom_codec.h:155
unsigned int reduced_reference_set
use reduced reference frame set
Definition: aom_encoder.h:363
aom_fixed_buf_t twopass_stats
Definition: aom_encoder.h:140
Definition: aom_encoder.h:201
unsigned int disable_flip_idtx
disable flip and identity transform type
Definition: aom_encoder.h:259
Definition: aom_encoder.h:189
int den
Definition: aom_encoder.h:166
unsigned int sframe_dist
sframe interval
Definition: aom_encoder.h:802
unsigned int disable_dist_wtd_comp
disable dist weighted compound
Definition: aom_encoder.h:283
aom_fixed_buf_t firstpass_mb_stats
Definition: aom_encoder.h:141
Encoder output packet.
Definition: aom_encoder.h:122
int tile_width_count
Number of explicit tile widths specified.
Definition: aom_encoder.h:861
Definition: aom_encoder.h:180
unsigned int rc_min_quantizer
Minimum (Best Quality) Quantizer.
Definition: aom_encoder.h:667
unsigned int disable_cdef
disable CDEF filter
Definition: aom_encoder.h:263
unsigned int disable_intrabc
disable Intra Block Copy
Definition: aom_encoder.h:307
int num
Definition: aom_encoder.h:165
unsigned int disable_smooth_intra
disable intra smooth mode
Definition: aom_encoder.h:315
aom_superres_mode
Frame super-resolution mode.
Definition: aom_encoder.h:207
aom_codec_frame_flags_t flags
Definition: aom_encoder.h:132
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: aom_encoder.h:527
unsigned int disable_inter_intra_comp
disable inter/intra compound
Definition: aom_encoder.h:291
unsigned int rc_superres_kf_denominator
Keyframe super-resolution denominator.
Definition: aom_encoder.h:604
long aom_enc_frame_flags_t
Encoded Frame Flags.
Definition: aom_encoder.h:377
unsigned int g_threads
Maximum number of threads to use.
Definition: aom_encoder.h:412
Definition: aom_encoder.h:211
union aom_codec_cx_pkt::@1 data
int tile_height_count
Number of explicit tile heights specified.
Definition: aom_encoder.h:869
aom_bit_depth_t g_bit_depth
Bit-depth of the codec.
Definition: aom_encoder.h:472
unsigned int rc_superres_kf_qthresh
Keyframe super-resolution q threshold.
Definition: aom_encoder.h:622
struct cfg_options cfg_options_t
Encoder Config Options.
size_t sz
Definition: aom_encoder.h:90
unsigned int reduced_tx_type_set
use reduced transform type set
Definition: aom_encoder.h:367
unsigned int disable_one_sided_comp
disable one sided compound
Definition: aom_encoder.h:299
int tile_heights[64]
Array of specified tile heights.
Definition: aom_encoder.h:895
Definition: aom_encoder.h:178
unsigned int save_as_annexb
Bitstream syntax mode.
Definition: aom_encoder.h:853
unsigned int disable_ab_partition_type
disable AB Shape partition type
Definition: aom_encoder.h:247
Definition: aom_encoder.h:203
enum aom_enc_pass g_pass
Multi-pass Encoding Mode.
Definition: aom_encoder.h:513
unsigned int g_h
Height of the frame.
Definition: aom_encoder.h:440
Definition: aom_encoder.h:179
Provides function pointer definitions for the external partition.
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: aom_encoder.h:549