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"
33#include "./vpx_ext_ratectrl.h"
34#include "./vpx_tpl.h"
35
39#define VPX_TS_MAX_PERIODICITY 16
40
42#define VPX_TS_MAX_LAYERS 5
43
45#define VPX_MAX_LAYERS 12 // 3 temporal + 4 spatial layers are allowed.
46
48#define VPX_SS_MAX_LAYERS 5
49
51#define VPX_SS_DEFAULT_LAYERS 1
52
61#define VPX_ENCODER_ABI_VERSION \
62 (16 + VPX_CODEC_ABI_VERSION + VPX_EXT_RATECTRL_ABI_VERSION + \
63 VPX_TPL_ABI_VERSION)
74#define VPX_CODEC_CAP_PSNR 0x10000
81#define VPX_CODEC_CAP_OUTPUT_PARTITION 0x20000
82
90#define VPX_CODEC_USE_PSNR 0x10000
92#define VPX_CODEC_USE_OUTPUT_PARTITION 0x20000
93#define VPX_CODEC_USE_HIGHBITDEPTH 0x40000
99typedef struct vpx_fixed_buf {
100 void *buf;
101 size_t sz;
109typedef int64_t vpx_codec_pts_t;
110
118typedef uint32_t vpx_codec_frame_flags_t;
119#define VPX_FRAME_IS_KEY 0x1u
122#define VPX_FRAME_IS_DROPPABLE 0x2u
124#define VPX_FRAME_IS_INVISIBLE 0x4u
126#define VPX_FRAME_IS_FRAGMENT 0x8u
127
134typedef uint32_t vpx_codec_er_flags_t;
136#define VPX_ERROR_RESILIENT_DEFAULT 0x1u
142#define VPX_ERROR_RESILIENT_PARTITIONS 0x2u
143
157
163typedef struct vpx_codec_cx_pkt {
165 union {
166 struct {
167 void *buf;
168 size_t sz;
172 unsigned long duration;
180 unsigned int width[VPX_SS_MAX_LAYERS];
181 unsigned int height[VPX_SS_MAX_LAYERS];
188 struct vpx_psnr_pkt {
189 unsigned int samples[4];
190 uint64_t sse[4];
191 double psnr[4];
195 /* This packet size is fixed to allow codecs to extend this
196 * interface without having to manage storage for raw packets,
197 * i.e., if it's smaller than 128 bytes, you can store in the
198 * packet list directly.
199 */
200 char pad[128 - sizeof(enum vpx_codec_cx_pkt_kind)];
210 void *user_data);
211
215 void *user_priv;
217
222typedef struct vpx_rational {
223 int num;
224 int den;
228typedef enum vpx_enc_pass {
233
240};
241
253 VPX_KF_DISABLED = 0
255
264#define VPX_EFLAG_FORCE_KF (1 << 0)
272typedef struct vpx_codec_enc_cfg {
273 /*
274 * generic settings (g)
275 */
276
281 unsigned int g_usage;
282
289 unsigned int g_threads;
290
299 unsigned int g_profile;
308 unsigned int g_w;
309
317 unsigned int g_h;
318
326
333 unsigned int g_input_bit_depth;
334
348
356
363
376 unsigned int g_lag_in_frames;
377
378 /*
379 * rate control settings (rc)
380 */
381
396
404 unsigned int rc_resize_allowed;
405
411 unsigned int rc_scaled_width;
412
418 unsigned int rc_scaled_height;
419
427
435
445
452
459
464 unsigned int rc_target_bitrate;
465
466 /*
467 * quantizer settings
468 */
469
477 unsigned int rc_min_quantizer;
478
486 unsigned int rc_max_quantizer;
487
488 /*
489 * bitrate tolerance
490 */
491
505 unsigned int rc_undershoot_pct;
506
520 unsigned int rc_overshoot_pct;
521
522 /*
523 * decoder buffer model parameters
524 */
525
535 unsigned int rc_buf_sz;
536
544 unsigned int rc_buf_initial_sz;
545
553 unsigned int rc_buf_optimal_sz;
554
555 /*
556 * 2 pass rate control parameters
557 */
558
568
575
582
589
590 /*
591 * keyframing settings (kf)
592 */
593
601
609 unsigned int kf_min_dist;
610
618 unsigned int kf_max_dist;
619
620 /*
621 * Spatial scalability settings (ss)
622 */
623
628 unsigned int ss_number_layers;
629
636
643
648 unsigned int ts_number_layers;
649
656
663
671 unsigned int ts_periodicity;
672
681
689
698
704
713
722
731
740
749
758
768
777
786
795
804
813
823
833
850typedef struct vpx_svc_parameters {
859
886 vpx_codec_iface_t *iface,
887 const vpx_codec_enc_cfg_t *cfg,
888 vpx_codec_flags_t flags, int ver);
889
894#define vpx_codec_enc_init(ctx, iface, cfg, flags) \
895 vpx_codec_enc_init_ver(ctx, iface, cfg, flags, VPX_ENCODER_ABI_VERSION)
896
919 int num_enc, vpx_codec_flags_t flags, vpx_rational_t *dsf, int ver);
920
925#define vpx_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf) \
926 vpx_codec_enc_init_multi_ver(ctx, iface, cfg, num_enc, flags, dsf, \
927 VPX_ENCODER_ABI_VERSION)
928
950 unsigned int usage);
951
967 const vpx_codec_enc_cfg_t *cfg);
968
990
992#define VPX_DL_REALTIME (1)
994#define VPX_DL_GOOD_QUALITY (1000000)
996#define VPX_DL_BEST_QUALITY (0)
1034 vpx_codec_pts_t pts, unsigned long duration,
1036 unsigned long deadline);
1037
1082 const vpx_fixed_buf_t *buf,
1083 unsigned int pad_before,
1084 unsigned int pad_after);
1085
1110 vpx_codec_iter_t *iter);
1111
1125
1127#ifdef __cplusplus
1128}
1129#endif
1130#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:42
vpx_codec_cx_pkt_kind
Encoder output packet variants.
Definition: vpx_encoder.h:150
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:209
vpx_enc_pass
Multi-pass Encoding Pass.
Definition: vpx_encoder.h:228
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:39
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:45
uint32_t vpx_codec_er_flags_t
Error Resilient flags.
Definition: vpx_encoder.h:134
int64_t vpx_codec_pts_t
Time Stamp Type.
Definition: vpx_encoder.h:109
#define VPX_SS_MAX_LAYERS
Definition: vpx_encoder.h:48
long vpx_enc_frame_flags_t
Encoded Frame Flags.
Definition: vpx_encoder.h:263
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:250
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:118
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_rc_mode
Rate control mode.
Definition: vpx_encoder.h:235
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, unsigned long deadline)
Encode a frame.
@ VPX_CODEC_PSNR_PKT
Definition: vpx_encoder.h:154
@ VPX_CODEC_CX_FRAME_PKT
Definition: vpx_encoder.h:151
@ VPX_CODEC_STATS_PKT
Definition: vpx_encoder.h:152
@ VPX_CODEC_CUSTOM_PKT
Definition: vpx_encoder.h:155
@ VPX_CODEC_FPMB_STATS_PKT
Definition: vpx_encoder.h:153
@ VPX_RC_LAST_PASS
Definition: vpx_encoder.h:231
@ VPX_RC_ONE_PASS
Definition: vpx_encoder.h:229
@ VPX_RC_FIRST_PASS
Definition: vpx_encoder.h:230
@ VPX_KF_DISABLED
Definition: vpx_encoder.h:253
@ VPX_KF_FIXED
Definition: vpx_encoder.h:251
@ VPX_KF_AUTO
Definition: vpx_encoder.h:252
@ VPX_Q
Definition: vpx_encoder.h:239
@ VPX_CQ
Definition: vpx_encoder.h:238
@ VPX_CBR
Definition: vpx_encoder.h:237
@ VPX_VBR
Definition: vpx_encoder.h:236
Codec context structure.
Definition: vpx_codec.h:200
Encoder output packet.
Definition: vpx_encoder.h:163
char pad[128 - sizeof(enum vpx_codec_cx_pkt_kind)]
Definition: vpx_encoder.h:200
vpx_codec_frame_flags_t flags
Definition: vpx_encoder.h:173
vpx_fixed_buf_t twopass_stats
Definition: vpx_encoder.h:186
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:164
double psnr[4]
Definition: vpx_encoder.h:191
unsigned int width[5]
Width and height of frames in this packet. VP8 will only use the first one.
Definition: vpx_encoder.h:180
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:184
vpx_fixed_buf_t firstpass_mb_stats
Definition: vpx_encoder.h:187
size_t sz
Definition: vpx_encoder.h:168
void * buf
Definition: vpx_encoder.h:167
vpx_codec_pts_t pts
time stamp to show frame (in timebase units)
Definition: vpx_encoder.h:170
vpx_fixed_buf_t raw
Definition: vpx_encoder.h:193
unsigned int height[5]
Definition: vpx_encoder.h:181
int partition_id
the partition id defines the decoding order of the partitions. Only applicable when "output partition...
Definition: vpx_encoder.h:177
unsigned long duration
duration to show frame (in timebase units)
Definition: vpx_encoder.h:172
union vpx_codec_cx_pkt::@1 data
Encoder configuration structure.
Definition: vpx_encoder.h:272
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:832
unsigned int rc_resize_allowed
Enable/disable spatial resampling, if supported by the codec.
Definition: vpx_encoder.h:404
int temporal_layering_mode
Temporal layering mode indicating which temporal layering scheme to use.
Definition: vpx_encoder.h:697
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: vpx_encoder.h:609
vpx_rational_t sr_default_decay_limit
Second reference default decay limit.
Definition: vpx_encoder.h:730
unsigned int rc_min_quantizer
Minimum (Best Quality) Quantizer.
Definition: vpx_encoder.h:477
unsigned int ts_number_layers
Number of temporal coding layers.
Definition: vpx_encoder.h:648
vpx_rational_t kf_max_total_boost_factor
Keyframe maximum total boost adjustment factor.
Definition: vpx_encoder.h:785
vpx_fixed_buf_t rc_twopass_stats_in
Two-pass stats buffer.
Definition: vpx_encoder.h:451
unsigned int rc_resize_down_thresh
Spatial resampling down watermark.
Definition: vpx_encoder.h:434
unsigned int ss_number_layers
Number of spatial coding layers.
Definition: vpx_encoder.h:628
unsigned int rc_2pass_vbr_bias_pct
Two-pass mode CBR/VBR bias.
Definition: vpx_encoder.h:567
unsigned int rc_2pass_vbr_minsection_pct
Two-pass mode per-GOP minimum bitrate.
Definition: vpx_encoder.h:574
int ss_enable_auto_alt_ref[5]
Enable auto alt reference flags for each spatial layer.
Definition: vpx_encoder.h:635
unsigned int rc_scaled_width
Internal coded frame width.
Definition: vpx_encoder.h:411
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:842
unsigned int g_profile
Bitstream profile to use.
Definition: vpx_encoder.h:299
unsigned int layer_target_bitrate[12]
Target bitrate for each spatial/temporal layer.
Definition: vpx_encoder.h:688
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:317
enum vpx_kf_mode kf_mode
Keyframe placement mode.
Definition: vpx_encoder.h:600
unsigned int ts_layer_id[16]
Template defining the membership of frames to temporal layers.
Definition: vpx_encoder.h:680
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: vpx_encoder.h:355
unsigned int ts_periodicity
Length of the sequence defining frame temporal layer membership.
Definition: vpx_encoder.h:671
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition: vpx_encoder.h:520
vpx_fixed_buf_t rc_firstpass_mb_stats_in
first pass mb stats buffer.
Definition: vpx_encoder.h:458
unsigned int rc_scaled_height
Internal coded frame height.
Definition: vpx_encoder.h:418
unsigned int g_usage
Deprecated: Algorithm specific "usage" value.
Definition: vpx_encoder.h:281
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:308
unsigned int ss_target_bitrate[5]
Target bitrate for each spatial layer.
Definition: vpx_encoder.h:642
unsigned int rc_buf_sz
Decoder Buffer Size.
Definition: vpx_encoder.h:535
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: vpx_encoder.h:395
struct vpx_rational g_timebase
Stream timebase units.
Definition: vpx_encoder.h:347
enum vpx_enc_pass g_pass
Multi-pass Encoding Mode.
Definition: vpx_encoder.h:362
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:767
unsigned int rc_resize_up_thresh
Spatial resampling up watermark.
Definition: vpx_encoder.h:426
unsigned int rc_max_quantizer
Maximum (Worst Quality) Quantizer.
Definition: vpx_encoder.h:486
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:588
vpx_rational_t zm_factor
Zero motion power factor.
Definition: vpx_encoder.h:812
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: vpx_encoder.h:376
enum vpx_rc_mode rc_end_usage
Rate control algorithm to use.
Definition: vpx_encoder.h:444
vpx_rational_t active_wq_factor
Active worst quality factor.
Definition: vpx_encoder.h:712
vpx_rational_t gf_max_total_boost_factor
Golden frame maximum total boost adjustment factor.
Definition: vpx_encoder.h:794
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition: vpx_encoder.h:544
vpx_bit_depth_t g_bit_depth
Bit-depth of the codec.
Definition: vpx_encoder.h:325
vpx_rational_t gf_frame_max_boost_factor
Golden frame maximum boost adjustment factor.
Definition: vpx_encoder.h:803
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition: vpx_encoder.h:553
int use_vizier_rc_params
A flag indicating whether to use external rate control parameters. By default is 0....
Definition: vpx_encoder.h:703
unsigned int rc_2pass_vbr_maxsection_pct
Two-pass mode per-GOP maximum bitrate.
Definition: vpx_encoder.h:581
unsigned int rc_target_bitrate
Target data rate.
Definition: vpx_encoder.h:464
unsigned int ts_target_bitrate[5]
Target bitrate for each temporal layer.
Definition: vpx_encoder.h:655
vpx_rational_t err_per_mb_factor
Error per macroblock adjustment factor.
Definition: vpx_encoder.h:721
unsigned int g_input_bit_depth
Bit-depth of the input frames.
Definition: vpx_encoder.h:333
vpx_rational_t kf_frame_min_boost_factor
Keyframe minimum boost adjustment factor.
Definition: vpx_encoder.h:757
vpx_rational_t sr_diff_factor
Second reference difference factor.
Definition: vpx_encoder.h:739
vpx_rational_t kf_err_per_mb_factor
Keyframe error per macroblock adjustment factor.
Definition: vpx_encoder.h:748
unsigned int rc_undershoot_pct
Rate control adaptation undershoot control.
Definition: vpx_encoder.h:505
unsigned int ts_rate_decimator[5]
Frame rate decimation factor for each temporal layer.
Definition: vpx_encoder.h:662
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: vpx_encoder.h:618
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:822
vpx_rational_t kf_frame_max_boost_subs_factor
Keyframe maximum boost adjustment factor, for subsequent keyframes.
Definition: vpx_encoder.h:776
unsigned int g_threads
Maximum number of threads to use.
Definition: vpx_encoder.h:289
Callback function pointer / user data pair storage.
Definition: vpx_encoder.h:213
void * user_priv
Definition: vpx_encoder.h:215
vpx_codec_enc_output_cx_pkt_cb_fn_t output_cx_pkt
Definition: vpx_encoder.h:214
Generic fixed size buffer structure.
Definition: vpx_encoder.h:99
size_t sz
Definition: vpx_encoder.h:101
void * buf
Definition: vpx_encoder.h:100
Image Descriptor.
Definition: vpx_image.h:72
Rational Number.
Definition: vpx_encoder.h:222
int den
Definition: vpx_encoder.h:224
int num
Definition: vpx_encoder.h:223
vp9 svc extra configure parameters
Definition: vpx_encoder.h:850
int min_quantizers[12]
Definition: vpx_encoder.h:852
int scaling_factor_num[12]
Definition: vpx_encoder.h:853
int speed_per_layer[12]
Definition: vpx_encoder.h:855
int max_quantizers[12]
Definition: vpx_encoder.h:851
int scaling_factor_den[12]
Definition: vpx_encoder.h:854
int loopfilter_ctrl[12]
Definition: vpx_encoder.h:857
int temporal_layering_mode
Definition: vpx_encoder.h:856
Describes the codec algorithm interface to applications.
Describes the TPL stats descriptor and associated operations.