WebM Codec SDK
vpx_encoder.h
Go to the documentation of this file.
1/*
2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10#ifndef VPX_VPX_VPX_ENCODER_H_
11#define VPX_VPX_VPX_ENCODER_H_
12
28#ifdef __cplusplus
29extern "C" {
30#endif
31
32#include "./vpx_codec.h" // IWYU pragma: export
33#include "./vpx_ext_ratectrl.h"
34
38#define VPX_TS_MAX_PERIODICITY 16
39
41#define VPX_TS_MAX_LAYERS 5
42
44#define VPX_MAX_LAYERS 12 // 3 temporal + 4 spatial layers are allowed.
45
47#define VPX_SS_MAX_LAYERS 5
48
50#define VPX_SS_DEFAULT_LAYERS 1
51
65#define VPX_ENCODER_ABI_VERSION \
66 (18 + VPX_CODEC_ABI_VERSION + \
67 VPX_EXT_RATECTRL_ABI_VERSION)
78#define VPX_CODEC_CAP_PSNR 0x10000
85#define VPX_CODEC_CAP_OUTPUT_PARTITION 0x20000
86
94#define VPX_CODEC_USE_PSNR 0x10000
96#define VPX_CODEC_USE_OUTPUT_PARTITION 0x20000
97#define VPX_CODEC_USE_HIGHBITDEPTH 0x40000
103typedef struct vpx_fixed_buf {
104 void *buf;
105 size_t sz;
113typedef int64_t vpx_codec_pts_t;
114
122typedef uint32_t vpx_codec_frame_flags_t;
123#define VPX_FRAME_IS_KEY 0x1u
126#define VPX_FRAME_IS_DROPPABLE 0x2u
128#define VPX_FRAME_IS_INVISIBLE 0x4u
130#define VPX_FRAME_IS_FRAGMENT 0x8u
131
138typedef uint32_t vpx_codec_er_flags_t;
140#define VPX_ERROR_RESILIENT_DEFAULT 0x1u
146#define VPX_ERROR_RESILIENT_PARTITIONS 0x2u
147
161
167typedef struct vpx_codec_cx_pkt {
169 union {
170 struct {
171 void *buf;
172 size_t sz;
176 unsigned long duration;
184 unsigned int width[VPX_SS_MAX_LAYERS];
185 unsigned int height[VPX_SS_MAX_LAYERS];
192 struct vpx_psnr_pkt {
193 unsigned int samples[4];
194 uint64_t sse[4];
195 double psnr[4];
199 /* This packet size is fixed to allow codecs to extend this
200 * interface without having to manage storage for raw packets,
201 * i.e., if it's smaller than 128 bytes, you can store in the
202 * packet list directly.
203 */
204 char pad[128 - sizeof(enum vpx_codec_cx_pkt_kind)];
214 void *user_data);
215
219 void *user_priv;
221
226typedef struct vpx_rational {
227 int num;
228 int den;
232typedef enum vpx_enc_pass {
237
244};
245
257 VPX_KF_DISABLED = 0
259
268#define VPX_EFLAG_FORCE_KF (1 << 0)
276typedef struct vpx_codec_enc_cfg {
277 /*
278 * generic settings (g)
279 */
280
285 unsigned int g_usage;
286
293 unsigned int g_threads;
294
303 unsigned int g_profile;
312 unsigned int g_w;
313
321 unsigned int g_h;
322
330
337 unsigned int g_input_bit_depth;
338
352
360
367
380 unsigned int g_lag_in_frames;
381
382 /*
383 * rate control settings (rc)
384 */
385
400
408 unsigned int rc_resize_allowed;
409
415 unsigned int rc_scaled_width;
416
422 unsigned int rc_scaled_height;
423
431
439
449
456
463
470 unsigned int rc_target_bitrate;
471
472 /*
473 * quantizer settings
474 */
475
483 unsigned int rc_min_quantizer;
484
492 unsigned int rc_max_quantizer;
493
494 /*
495 * bitrate tolerance
496 */
497
511 unsigned int rc_undershoot_pct;
512
526 unsigned int rc_overshoot_pct;
527
528 /*
529 * decoder buffer model parameters
530 */
531
541 unsigned int rc_buf_sz;
542
550 unsigned int rc_buf_initial_sz;
551
559 unsigned int rc_buf_optimal_sz;
560
561 /*
562 * 2 pass rate control parameters
563 */
564
574
581
588
595
596 /*
597 * keyframing settings (kf)
598 */
599
607
615 unsigned int kf_min_dist;
616
624 unsigned int kf_max_dist;
625
626 /*
627 * Spatial scalability settings (ss)
628 */
629
634 unsigned int ss_number_layers;
635
642
649
654 unsigned int ts_number_layers;
655
662
669
677 unsigned int ts_periodicity;
678
687
695
704
710
719
728
737
746
755
764
774
783
792
801
810
819
829
839
856typedef struct vpx_svc_parameters {
865
892 vpx_codec_iface_t *iface,
893 const vpx_codec_enc_cfg_t *cfg,
894 vpx_codec_flags_t flags, int ver);
895
900#define vpx_codec_enc_init(ctx, iface, cfg, flags) \
901 vpx_codec_enc_init_ver(ctx, iface, cfg, flags, VPX_ENCODER_ABI_VERSION)
902
925 int num_enc, vpx_codec_flags_t flags, vpx_rational_t *dsf, int ver);
926
931#define vpx_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf) \
932 vpx_codec_enc_init_multi_ver(ctx, iface, cfg, num_enc, flags, dsf, \
933 VPX_ENCODER_ABI_VERSION)
934
956 unsigned int usage);
957
973 const vpx_codec_enc_cfg_t *cfg);
974
996
1002typedef unsigned long vpx_enc_deadline_t;
1004#define VPX_DL_REALTIME 1ul
1006#define VPX_DL_GOOD_QUALITY 1000000ul
1008#define VPX_DL_BEST_QUALITY 0ul
1048 vpx_codec_pts_t pts, unsigned long duration,
1050 vpx_enc_deadline_t deadline);
1051
1102 const vpx_fixed_buf_t *buf,
1103 unsigned int pad_before,
1104 unsigned int pad_after);
1105
1130 vpx_codec_iter_t *iter);
1131
1145
1147#ifdef __cplusplus
1148}
1149#endif
1150#endif // VPX_VPX_VPX_ENCODER_H_
long vpx_codec_flags_t
Initialization-time Feature Enabling.
Definition: vpx_codec.h:170
const void * vpx_codec_iter_t
Iterator.
Definition: vpx_codec.h:190
enum vpx_bit_depth vpx_bit_depth_t
Bit depth for codecThis enumeration determines the bit depth of the codec.
vpx_codec_err_t
Algorithm return codes.
Definition: vpx_codec.h:93
const struct vpx_codec_iface vpx_codec_iface_t
Codec interface structure.
Definition: vpx_codec.h:177
vpx_codec_err_t vpx_codec_enc_init_ver(vpx_codec_ctx_t *ctx, vpx_codec_iface_t *iface, const vpx_codec_enc_cfg_t *cfg, vpx_codec_flags_t flags, int ver)
Initialize an encoder instance.
vpx_codec_err_t vpx_codec_enc_init_multi_ver(vpx_codec_ctx_t *ctx, vpx_codec_iface_t *iface, vpx_codec_enc_cfg_t *cfg, int num_enc, vpx_codec_flags_t flags, vpx_rational_t *dsf, int ver)
Initialize multi-encoder instance.
struct vpx_codec_enc_cfg vpx_codec_enc_cfg_t
Encoder configuration structure.
#define VPX_TS_MAX_LAYERS
Definition: vpx_encoder.h:41
vpx_codec_cx_pkt_kind
Encoder output packet variants.
Definition: vpx_encoder.h:154
void(* vpx_codec_enc_output_cx_pkt_cb_fn_t)(vpx_codec_cx_pkt_t *pkt, void *user_data)
Encoder return output buffer callback.
Definition: vpx_encoder.h:213
vpx_enc_pass
Multi-pass Encoding Pass.
Definition: vpx_encoder.h:232
const vpx_image_t * vpx_codec_get_preview_frame(vpx_codec_ctx_t *ctx)
Get Preview Frame.
struct vpx_svc_parameters vpx_svc_extra_cfg_t
vp9 svc extra configure parameters
struct vpx_fixed_buf vpx_fixed_buf_t
Generic fixed size buffer structure.
const vpx_codec_cx_pkt_t * vpx_codec_get_cx_data(vpx_codec_ctx_t *ctx, vpx_codec_iter_t *iter)
Encoded data iterator.
#define VPX_TS_MAX_PERIODICITY
Definition: vpx_encoder.h:38
unsigned long vpx_enc_deadline_t
Encode Deadline.
Definition: vpx_encoder.h:1002
struct vpx_codec_enc_output_cx_cb_pair vpx_codec_priv_output_cx_pkt_cb_pair_t
Callback function pointer / user data pair storage.
#define VPX_MAX_LAYERS
Definition: vpx_encoder.h:44
uint32_t vpx_codec_er_flags_t
Error Resilient flags.
Definition: vpx_encoder.h:138
int64_t vpx_codec_pts_t
Time Stamp Type.
Definition: vpx_encoder.h:113
#define VPX_SS_MAX_LAYERS
Definition: vpx_encoder.h:47
long vpx_enc_frame_flags_t
Encoded Frame Flags.
Definition: vpx_encoder.h:267
vpx_codec_err_t vpx_codec_enc_config_default(vpx_codec_iface_t *iface, vpx_codec_enc_cfg_t *cfg, unsigned int usage)
Get a default configuration.
vpx_kf_mode
Keyframe placement mode.
Definition: vpx_encoder.h:254
struct vpx_codec_cx_pkt vpx_codec_cx_pkt_t
Encoder output packet.
uint32_t vpx_codec_frame_flags_t
Compressed Frame Flags.
Definition: vpx_encoder.h:122
vpx_codec_err_t vpx_codec_enc_config_set(vpx_codec_ctx_t *ctx, const vpx_codec_enc_cfg_t *cfg)
Set or change configuration.
vpx_fixed_buf_t * vpx_codec_get_global_headers(vpx_codec_ctx_t *ctx)
Get global stream headers.
struct vpx_rational vpx_rational_t
Rational Number.
vpx_codec_err_t vpx_codec_set_cx_data_buf(vpx_codec_ctx_t *ctx, const vpx_fixed_buf_t *buf, unsigned int pad_before, unsigned int pad_after)
Set compressed data output buffer.
vpx_codec_err_t vpx_codec_encode(vpx_codec_ctx_t *ctx, const vpx_image_t *img, vpx_codec_pts_t pts, unsigned long duration, vpx_enc_frame_flags_t flags, vpx_enc_deadline_t deadline)
Encode a frame.
vpx_rc_mode
Rate control mode.
Definition: vpx_encoder.h:239
@ VPX_CODEC_PSNR_PKT
Definition: vpx_encoder.h:158
@ VPX_CODEC_CX_FRAME_PKT
Definition: vpx_encoder.h:155
@ VPX_CODEC_STATS_PKT
Definition: vpx_encoder.h:156
@ VPX_CODEC_CUSTOM_PKT
Definition: vpx_encoder.h:159
@ VPX_CODEC_FPMB_STATS_PKT
Definition: vpx_encoder.h:157
@ VPX_RC_LAST_PASS
Definition: vpx_encoder.h:235
@ VPX_RC_ONE_PASS
Definition: vpx_encoder.h:233
@ VPX_RC_FIRST_PASS
Definition: vpx_encoder.h:234
@ VPX_KF_DISABLED
Definition: vpx_encoder.h:257
@ VPX_KF_FIXED
Definition: vpx_encoder.h:255
@ VPX_KF_AUTO
Definition: vpx_encoder.h:256
@ VPX_Q
Definition: vpx_encoder.h:243
@ VPX_CQ
Definition: vpx_encoder.h:242
@ VPX_CBR
Definition: vpx_encoder.h:241
@ VPX_VBR
Definition: vpx_encoder.h:240
Codec context structure.
Definition: vpx_codec.h:200
Encoder output packet.
Definition: vpx_encoder.h:167
char pad[128 - sizeof(enum vpx_codec_cx_pkt_kind)]
Definition: vpx_encoder.h:204
vpx_codec_frame_flags_t flags
Definition: vpx_encoder.h:177
vpx_fixed_buf_t twopass_stats
Definition: vpx_encoder.h:190
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:168
unsigned int samples[4]
Definition: vpx_encoder.h:193
double psnr[4]
Definition: vpx_encoder.h:195
unsigned int width[5]
Width and height of frames in this packet. VP8 will only use the first one.
Definition: vpx_encoder.h:184
struct vpx_codec_cx_pkt::@1::@2 frame
uint8_t spatial_layer_encoded[5]
Flag to indicate if spatial layer frame in this packet is encoded or dropped. VP8 will always be set ...
Definition: vpx_encoder.h:188
vpx_fixed_buf_t firstpass_mb_stats
Definition: vpx_encoder.h:191
size_t sz
Definition: vpx_encoder.h:172
void * buf
Definition: vpx_encoder.h:171
vpx_codec_pts_t pts
time stamp to show frame (in timebase units)
Definition: vpx_encoder.h:174
vpx_fixed_buf_t raw
Definition: vpx_encoder.h:197
unsigned int height[5]
Definition: vpx_encoder.h:185
int partition_id
the partition id defines the decoding order of the partitions. Only applicable when "output partition...
Definition: vpx_encoder.h:181
uint64_t sse[4]
Definition: vpx_encoder.h:194
unsigned long duration
duration to show frame (in timebase units)
Definition: vpx_encoder.h:176
union vpx_codec_cx_pkt::@1 data
Encoder configuration structure.
Definition: vpx_encoder.h:276
vpx_rational_t rd_mult_arf_qp_fac
Rate-distortion multiplier for alt-ref frames. The multiplier is a crucial parameter in the calculati...
Definition: vpx_encoder.h:838
unsigned int rc_resize_allowed
Enable/disable spatial resampling, if supported by the codec.
Definition: vpx_encoder.h:408
int temporal_layering_mode
Temporal layering mode indicating which temporal layering scheme to use.
Definition: vpx_encoder.h:703
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: vpx_encoder.h:615
vpx_rational_t sr_default_decay_limit
Second reference default decay limit.
Definition: vpx_encoder.h:736
unsigned int rc_min_quantizer
Minimum (Best Quality) Quantizer.
Definition: vpx_encoder.h:483
unsigned int ts_number_layers
Number of temporal coding layers.
Definition: vpx_encoder.h:654
vpx_rational_t kf_max_total_boost_factor
Keyframe maximum total boost adjustment factor.
Definition: vpx_encoder.h:791
vpx_fixed_buf_t rc_twopass_stats_in
Two-pass stats buffer.
Definition: vpx_encoder.h:455
unsigned int rc_resize_down_thresh
Spatial resampling down watermark.
Definition: vpx_encoder.h:438
unsigned int ss_number_layers
Number of spatial coding layers.
Definition: vpx_encoder.h:634
unsigned int rc_2pass_vbr_bias_pct
Two-pass mode CBR/VBR bias.
Definition: vpx_encoder.h:573
unsigned int rc_2pass_vbr_minsection_pct
Two-pass mode per-GOP minimum bitrate.
Definition: vpx_encoder.h:580
int ss_enable_auto_alt_ref[5]
Enable auto alt reference flags for each spatial layer.
Definition: vpx_encoder.h:641
unsigned int rc_scaled_width
Internal coded frame width.
Definition: vpx_encoder.h:415
vpx_rational_t rd_mult_key_qp_fac
Rate-distortion multiplier for key frames. The multiplier is a crucial parameter in the calculation o...
Definition: vpx_encoder.h:848
unsigned int g_profile
Bitstream profile to use.
Definition: vpx_encoder.h:303
unsigned int layer_target_bitrate[12]
Target bitrate for each spatial/temporal layer.
Definition: vpx_encoder.h:694
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:321
enum vpx_kf_mode kf_mode
Keyframe placement mode.
Definition: vpx_encoder.h:606
unsigned int ts_layer_id[16]
Template defining the membership of frames to temporal layers.
Definition: vpx_encoder.h:686
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: vpx_encoder.h:359
unsigned int ts_periodicity
Length of the sequence defining frame temporal layer membership.
Definition: vpx_encoder.h:677
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition: vpx_encoder.h:526
vpx_fixed_buf_t rc_firstpass_mb_stats_in
first pass mb stats buffer.
Definition: vpx_encoder.h:462
unsigned int rc_scaled_height
Internal coded frame height.
Definition: vpx_encoder.h:422
unsigned int g_usage
Deprecated: Algorithm specific "usage" value.
Definition: vpx_encoder.h:285
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:312
unsigned int ss_target_bitrate[5]
Target bitrate for each spatial layer.
Definition: vpx_encoder.h:648
unsigned int rc_buf_sz
Decoder Buffer Size.
Definition: vpx_encoder.h:541
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: vpx_encoder.h:399
struct vpx_rational g_timebase
Stream timebase units.
Definition: vpx_encoder.h:351
enum vpx_enc_pass g_pass
Multi-pass Encoding Mode.
Definition: vpx_encoder.h:366
vpx_rational_t kf_frame_max_boost_first_factor
Keyframe maximum boost adjustment factor, for the first keyframe in a chunk.
Definition: vpx_encoder.h:773
unsigned int rc_resize_up_thresh
Spatial resampling up watermark.
Definition: vpx_encoder.h:430
unsigned int rc_max_quantizer
Maximum (Worst Quality) Quantizer.
Definition: vpx_encoder.h:492
unsigned int rc_2pass_vbr_corpus_complexity
Two-pass corpus vbr mode complexity control Used only in VP9: A value representing the corpus midpoin...
Definition: vpx_encoder.h:594
vpx_rational_t zm_factor
Zero motion power factor.
Definition: vpx_encoder.h:818
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: vpx_encoder.h:380
enum vpx_rc_mode rc_end_usage
Rate control algorithm to use.
Definition: vpx_encoder.h:448
vpx_rational_t active_wq_factor
Active worst quality factor.
Definition: vpx_encoder.h:718
vpx_rational_t gf_max_total_boost_factor
Golden frame maximum total boost adjustment factor.
Definition: vpx_encoder.h:800
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition: vpx_encoder.h:550
vpx_bit_depth_t g_bit_depth
Bit-depth of the codec.
Definition: vpx_encoder.h:329
vpx_rational_t gf_frame_max_boost_factor
Golden frame maximum boost adjustment factor.
Definition: vpx_encoder.h:809
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition: vpx_encoder.h:559
int use_vizier_rc_params
A flag indicating whether to use external rate control parameters. By default is 0....
Definition: vpx_encoder.h:709
unsigned int rc_2pass_vbr_maxsection_pct
Two-pass mode per-GOP maximum bitrate.
Definition: vpx_encoder.h:587
unsigned int rc_target_bitrate
Target data rate.
Definition: vpx_encoder.h:470
unsigned int ts_target_bitrate[5]
Target bitrate for each temporal layer.
Definition: vpx_encoder.h:661
vpx_rational_t err_per_mb_factor
Error per macroblock adjustment factor.
Definition: vpx_encoder.h:727
unsigned int g_input_bit_depth
Bit-depth of the input frames.
Definition: vpx_encoder.h:337
vpx_rational_t kf_frame_min_boost_factor
Keyframe minimum boost adjustment factor.
Definition: vpx_encoder.h:763
vpx_rational_t sr_diff_factor
Second reference difference factor.
Definition: vpx_encoder.h:745
vpx_rational_t kf_err_per_mb_factor
Keyframe error per macroblock adjustment factor.
Definition: vpx_encoder.h:754
unsigned int rc_undershoot_pct
Rate control adaptation undershoot control.
Definition: vpx_encoder.h:511
unsigned int ts_rate_decimator[5]
Frame rate decimation factor for each temporal layer.
Definition: vpx_encoder.h:668
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: vpx_encoder.h:624
vpx_rational_t rd_mult_inter_qp_fac
Rate-distortion multiplier for inter frames. The multiplier is a crucial parameter in the calculation...
Definition: vpx_encoder.h:828
vpx_rational_t kf_frame_max_boost_subs_factor
Keyframe maximum boost adjustment factor, for subsequent keyframes.
Definition: vpx_encoder.h:782
unsigned int g_threads
Maximum number of threads to use.
Definition: vpx_encoder.h:293
Callback function pointer / user data pair storage.
Definition: vpx_encoder.h:217
void * user_priv
Definition: vpx_encoder.h:219
vpx_codec_enc_output_cx_pkt_cb_fn_t output_cx_pkt
Definition: vpx_encoder.h:218
Generic fixed size buffer structure.
Definition: vpx_encoder.h:103
size_t sz
Definition: vpx_encoder.h:105
void * buf
Definition: vpx_encoder.h:104
Image Descriptor.
Definition: vpx_image.h:76
Rational Number.
Definition: vpx_encoder.h:226
int den
Definition: vpx_encoder.h:228
int num
Definition: vpx_encoder.h:227
vp9 svc extra configure parameters
Definition: vpx_encoder.h:856
int min_quantizers[12]
Definition: vpx_encoder.h:858
int scaling_factor_num[12]
Definition: vpx_encoder.h:859
int speed_per_layer[12]
Definition: vpx_encoder.h:861
int max_quantizers[12]
Definition: vpx_encoder.h:857
int scaling_factor_den[12]
Definition: vpx_encoder.h:860
int loopfilter_ctrl[12]
Definition: vpx_encoder.h:863
int temporal_layering_mode
Definition: vpx_encoder.h:862
Describes the codec algorithm interface to applications.
Defines structs and callbacks needed for external rate control.