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_ENCODER_H_
11 #define VPX_VPX_ENCODER_H_
12 
28 #ifdef __cplusplus
29 extern "C" {
30 #endif
31 
32 #include "./vpx_codec.h"
33 
37 #define VPX_TS_MAX_PERIODICITY 16
38 
40 #define VPX_TS_MAX_LAYERS 5
41 
43 #define MAX_PERIODICITY VPX_TS_MAX_PERIODICITY
44 
46 #define VPX_MAX_LAYERS 12 // 3 temporal + 4 spatial layers are allowed.
47 
49 #define MAX_LAYERS VPX_MAX_LAYERS // 3 temporal + 4 spatial layers allowed.
50 
52 #define VPX_SS_MAX_LAYERS 5
53 
55 #define VPX_SS_DEFAULT_LAYERS 1
56 
65 #define VPX_ENCODER_ABI_VERSION \
66  (5 + VPX_CODEC_ABI_VERSION)
77 #define VPX_CODEC_CAP_PSNR 0x10000
84 #define VPX_CODEC_CAP_OUTPUT_PARTITION 0x20000
85 
88 #define VPX_CODEC_CAP_HIGHBITDEPTH 0x40000
89 
97 #define VPX_CODEC_USE_PSNR 0x10000
99 #define VPX_CODEC_USE_OUTPUT_PARTITION 0x20000
100 #define VPX_CODEC_USE_HIGHBITDEPTH 0x40000
106 typedef struct vpx_fixed_buf {
107  void *buf;
108  size_t sz;
109 } vpx_fixed_buf_t;
116 typedef int64_t vpx_codec_pts_t;
117 
125 typedef uint32_t vpx_codec_frame_flags_t;
126 #define VPX_FRAME_IS_KEY 0x1
129 #define VPX_FRAME_IS_DROPPABLE 0x2
130 
131 #define VPX_FRAME_IS_INVISIBLE 0x4
132 
133 #define VPX_FRAME_IS_FRAGMENT 0x8
134 
141 typedef uint32_t vpx_codec_er_flags_t;
143 #define VPX_ERROR_RESILIENT_DEFAULT 0x1
144 
148 #define VPX_ERROR_RESILIENT_PARTITIONS 0x2
149 
161 // Spatial SVC is still experimental and may be removed before the next ABI
162 // bump.
163 #if VPX_ENCODER_ABI_VERSION > (5 + VPX_CODEC_ABI_VERSION)
164  VPX_CODEC_SPATIAL_SVC_LAYER_SIZES,
165  VPX_CODEC_SPATIAL_SVC_LAYER_PSNR,
166 #endif
168 };
169 
175 typedef struct vpx_codec_cx_pkt {
177  union {
178  struct {
179  void *buf;
180  size_t sz;
184  unsigned long duration;
190  } frame;
193  struct vpx_psnr_pkt {
194  unsigned int samples[4];
195  uint64_t sse[4];
196  double psnr[4];
197  } psnr;
199 // Spatial SVC is still experimental and may be removed before the next
200 // ABI bump.
201 #if VPX_ENCODER_ABI_VERSION > (5 + VPX_CODEC_ABI_VERSION)
202  size_t layer_sizes[VPX_SS_MAX_LAYERS];
203  struct vpx_psnr_pkt layer_psnr[VPX_SS_MAX_LAYERS];
204 #endif
205 
206  /* This packet size is fixed to allow codecs to extend this
207  * interface without having to manage storage for raw packets,
208  * i.e., if it's smaller than 128 bytes, you can store in the
209  * packet list directly.
210  */
211  char pad[128 - sizeof(enum vpx_codec_cx_pkt_kind)];
212  } data;
220 // putting the definitions here for now. (agrange: find if there
221 // is a better place for this)
223  void *user_data);
224 
228  void *user_priv;
230 
235 typedef struct vpx_rational {
236  int num;
237  int den;
238 } vpx_rational_t;
245 };
246 
253 };
254 
267 };
268 
277 #define VPX_EFLAG_FORCE_KF (1 << 0)
285 typedef struct vpx_codec_enc_cfg {
286  /*
287  * generic settings (g)
288  */
289 
297  unsigned int g_usage;
298 
305  unsigned int g_threads;
306 
315  unsigned int g_profile;
324  unsigned int g_w;
325 
333  unsigned int g_h;
334 
342 
349  unsigned int g_input_bit_depth;
350 
363  struct vpx_rational g_timebase;
364 
372 
378  enum vpx_enc_pass g_pass;
379 
392  unsigned int g_lag_in_frames;
393 
394  /*
395  * rate control settings (rc)
396  */
397 
414  unsigned int rc_dropframe_thresh;
415 
423  unsigned int rc_resize_allowed;
424 
430  unsigned int rc_scaled_width;
431 
437  unsigned int rc_scaled_height;
438 
445  unsigned int rc_resize_up_thresh;
446 
453  unsigned int rc_resize_down_thresh;
454 
463  enum vpx_rc_mode rc_end_usage;
464 
471 
478 
483  unsigned int rc_target_bitrate;
484 
485  /*
486  * quantizer settings
487  */
488 
497  unsigned int rc_min_quantizer;
498 
507  unsigned int rc_max_quantizer;
508 
509  /*
510  * bitrate tolerance
511  */
512 
523  unsigned int rc_undershoot_pct;
524 
535  unsigned int rc_overshoot_pct;
536 
537  /*
538  * decoder buffer model parameters
539  */
540 
550  unsigned int rc_buf_sz;
551 
559  unsigned int rc_buf_initial_sz;
560 
568  unsigned int rc_buf_optimal_sz;
569 
570  /*
571  * 2 pass rate control parameters
572  */
573 
582  unsigned int rc_2pass_vbr_bias_pct;
583 
590 
597 
598  /*
599  * keyframing settings (kf)
600  */
601 
608  enum vpx_kf_mode kf_mode;
609 
617  unsigned int kf_min_dist;
618 
626  unsigned int kf_max_dist;
627 
628  /*
629  * Spatial scalability settings (ss)
630  */
631 
636  unsigned int ss_number_layers;
637 
643  int ss_enable_auto_alt_ref[VPX_SS_MAX_LAYERS];
644 
650  unsigned int ss_target_bitrate[VPX_SS_MAX_LAYERS];
651 
656  unsigned int ts_number_layers;
657 
663  unsigned int ts_target_bitrate[VPX_TS_MAX_LAYERS];
664 
670  unsigned int ts_rate_decimator[VPX_TS_MAX_LAYERS];
671 
679  unsigned int ts_periodicity;
680 
688  unsigned int ts_layer_id[VPX_TS_MAX_PERIODICITY];
689 
696  unsigned int layer_target_bitrate[VPX_MAX_LAYERS];
697 
713 typedef struct vpx_svc_parameters {
721 
745  vpx_codec_iface_t *iface,
746  const vpx_codec_enc_cfg_t *cfg,
747  vpx_codec_flags_t flags, int ver);
748 
753 #define vpx_codec_enc_init(ctx, iface, cfg, flags) \
754  vpx_codec_enc_init_ver(ctx, iface, cfg, flags, VPX_ENCODER_ABI_VERSION)
755 
778  int num_enc, vpx_codec_flags_t flags, vpx_rational_t *dsf, int ver);
779 
784 #define vpx_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf) \
785  vpx_codec_enc_init_multi_ver(ctx, iface, cfg, num_enc, flags, dsf, \
786  VPX_ENCODER_ABI_VERSION)
787 
808  vpx_codec_enc_cfg_t *cfg,
809  unsigned int reserved);
810 
826  const vpx_codec_enc_cfg_t *cfg);
827 
840 
842 #define VPX_DL_REALTIME (1)
843 
844 #define VPX_DL_GOOD_QUALITY (1000000)
845 
846 #define VPX_DL_BEST_QUALITY (0)
847 
884  vpx_codec_pts_t pts, unsigned long duration,
885  vpx_enc_frame_flags_t flags,
886  unsigned long deadline);
887 
932  const vpx_fixed_buf_t *buf,
933  unsigned int pad_before,
934  unsigned int pad_after);
935 
960  vpx_codec_iter_t *iter);
961 
975 
977 #ifdef __cplusplus
978 }
979 #endif
980 #endif // VPX_VPX_ENCODER_H_
Rational Number.
Definition: vpx_encoder.h:235
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition: vpx_encoder.h:559
vpx_fixed_buf_t twopass_stats
Definition: vpx_encoder.h:191
int min_quantizers[12]
Definition: vpx_encoder.h:715
unsigned int ts_number_layers
Number of temporal coding layers.
Definition: vpx_encoder.h:656
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.
#define VPX_MAX_LAYERS
Definition: vpx_encoder.h:46
Definition: vpx_encoder.h:252
Image Descriptor.
Definition: vpx_image.h:88
int temporal_layering_mode
Definition: vpx_encoder.h:719
vpx_codec_pts_t pts
time stamp to show frame (in timebase units)
Definition: vpx_encoder.h:182
vpx_kf_mode
Keyframe placement mode.
Definition: vpx_encoder.h:263
unsigned int rc_scaled_width
Internal coded frame width.
Definition: vpx_encoder.h:430
#define VPX_TS_MAX_LAYERS
Definition: vpx_encoder.h:40
Definition: vpx_encoder.h:250
unsigned int g_usage
Algorithm specific "usage" value.
Definition: vpx_encoder.h:297
struct vpx_codec_enc_cfg vpx_codec_enc_cfg_t
Encoder configuration structure.
struct vpx_fixed_buf vpx_fixed_buf_t
Generic fixed size buffer structure.
unsigned int rc_buf_sz
Decoder Buffer Size.
Definition: vpx_encoder.h:550
unsigned int g_input_bit_depth
Bit-depth of the input frames.
Definition: vpx_encoder.h:349
vpx_codec_enc_output_cx_pkt_cb_fn_t output_cx_pkt
Definition: vpx_encoder.h:227
int den
Definition: vpx_encoder.h:237
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.
unsigned int rc_max_quantizer
Maximum (Worst Quality) Quantizer.
Definition: vpx_encoder.h:507
struct vpx_svc_parameters vpx_svc_extra_cfg_t
vp9 svc extra configure parameters
unsigned int rc_min_quantizer
Minimum (Best Quality) Quantizer.
Definition: vpx_encoder.h:497
Definition: vpx_encoder.h:158
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: vpx_encoder.h:626
int partition_id
the partition id defines the decoding order of the partitions. Only applicable when "output partition...
Definition: vpx_encoder.h:189
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: vpx_encoder.h:392
vpx_enc_pass
Multi-pass Encoding Pass.
Definition: vpx_encoder.h:241
Encoder configuration structure.
Definition: vpx_encoder.h:285
Definition: vpx_encoder.h:167
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:222
Definition: vpx_encoder.h:160
Definition: vpx_encoder.h:265
void * user_priv
Definition: vpx_encoder.h:228
Encoder output packet.
Definition: vpx_encoder.h:175
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition: vpx_encoder.h:535
void * buf
Definition: vpx_encoder.h:107
unsigned int rc_2pass_vbr_bias_pct
Two-pass mode CBR/VBR bias.
Definition: vpx_encoder.h:582
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition: vpx_encoder.h:568
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.
Generic fixed size buffer structure.
Definition: vpx_encoder.h:106
uint32_t vpx_codec_frame_flags_t
Compressed Frame Flags.
Definition: vpx_encoder.h:125
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: vpx_encoder.h:617
Definition: vpx_encoder.h:243
unsigned int g_profile
Bitstream profile to use.
Definition: vpx_encoder.h:315
Definition: vpx_encoder.h:244
struct vpx_codec_cx_pkt::@1::@2 frame
#define VPX_SS_MAX_LAYERS
Definition: vpx_encoder.h:52
int scaling_factor_num[12]
Definition: vpx_encoder.h:716
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:324
enum vpx_bit_depth vpx_bit_depth_t
Bit depth for codecThis enumeration determines the bit depth of the codec.
unsigned int rc_undershoot_pct
Rate control adaptation undershoot control.
Definition: vpx_encoder.h:523
uint64_t sse[4]
Definition: vpx_encoder.h:195
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:333
size_t sz
Definition: vpx_encoder.h:180
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:176
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: vpx_encoder.h:414
vpx_fixed_buf_t raw
Definition: vpx_encoder.h:198
Callback function pointer / user data pair storage.
Definition: vpx_encoder.h:226
long vpx_enc_frame_flags_t
Encoded Frame Flags.
Definition: vpx_encoder.h:276
unsigned int rc_target_bitrate
Target data rate.
Definition: vpx_encoder.h:483
int num
Definition: vpx_encoder.h:236
const struct vpx_codec_iface vpx_codec_iface_t
Codec interface structure.
Definition: vpx_codec.h:169
char pad[128-sizeof(enum vpx_codec_cx_pkt_kind)]
Definition: vpx_encoder.h:211
long vpx_codec_flags_t
Initialization-time Feature Enabling.
Definition: vpx_codec.h:162
vpx_codec_cx_pkt_kind
Encoder output packet variants.
Definition: vpx_encoder.h:156
Definition: vpx_encoder.h:249
double psnr[4]
Definition: vpx_encoder.h:196
Definition: vpx_encoder.h:264
unsigned int g_threads
Maximum number of threads to use.
Definition: vpx_encoder.h:305
unsigned int samples[4]
Definition: vpx_encoder.h:194
vpx_fixed_buf_t firstpass_mb_stats
Definition: vpx_encoder.h:192
vpx_rc_mode
Rate control mode.
Definition: vpx_encoder.h:248
const vpx_image_t * vpx_codec_get_preview_frame(vpx_codec_ctx_t *ctx)
Get Preview Frame.
unsigned int ss_number_layers
Number of spatial coding layers.
Definition: vpx_encoder.h:636
unsigned long duration
duration to show frame (in timebase units)
Definition: vpx_encoder.h:184
vpx_bit_depth_t g_bit_depth
Bit-depth of the codec.
Definition: vpx_encoder.h:341
uint32_t vpx_codec_er_flags_t
Error Resilient flags.
Definition: vpx_encoder.h:141
unsigned int rc_resize_allowed
Enable/disable spatial resampling, if supported by the codec.
Definition: vpx_encoder.h:423
vpx_codec_err_t
Algorithm return codes.
Definition: vpx_codec.h:89
const vpx_codec_cx_pkt_t * vpx_codec_get_cx_data(vpx_codec_ctx_t *ctx, vpx_codec_iter_t *iter)
Encoded data iterator.
union vpx_codec_cx_pkt::@1 data
struct vpx_codec_cx_pkt vpx_codec_cx_pkt_t
Encoder output packet.
struct vpx_rational vpx_rational_t
Rational Number.
int temporal_layering_mode
Temporal layering mode indicating which temporal layering scheme to use.
Definition: vpx_encoder.h:705
struct vpx_codec_enc_output_cx_cb_pair vpx_codec_priv_output_cx_pkt_cb_pair_t
Callback function pointer / user data pair storage.
vpx_fixed_buf_t rc_twopass_stats_in
Two-pass stats buffer.
Definition: vpx_encoder.h:470
Definition: vpx_encoder.h:266
int speed_per_layer[12]
Definition: vpx_encoder.h:718
unsigned int rc_scaled_height
Internal coded frame height.
Definition: vpx_encoder.h:437
Definition: vpx_encoder.h:159
vpx_fixed_buf_t rc_firstpass_mb_stats_in
first pass mb stats buffer.
Definition: vpx_encoder.h:477
int64_t vpx_codec_pts_t
Time Stamp Type.
Definition: vpx_encoder.h:116
vpx_codec_err_t vpx_codec_enc_config_default(vpx_codec_iface_t *iface, vpx_codec_enc_cfg_t *cfg, unsigned int reserved)
Get a default configuration.
#define VPX_TS_MAX_PERIODICITY
Definition: vpx_encoder.h:37
vpx_fixed_buf_t * vpx_codec_get_global_headers(vpx_codec_ctx_t *ctx)
Get global stream headers.
Definition: vpx_encoder.h:251
Describes the codec algorithm interface to applications.
unsigned int ts_periodicity
Length of the sequence defining frame temporal layer membership.
Definition: vpx_encoder.h:679
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.
size_t sz
Definition: vpx_encoder.h:108
unsigned int rc_resize_down_thresh
Spatial resampling down watermark.
Definition: vpx_encoder.h:453
int scaling_factor_den[12]
Definition: vpx_encoder.h:717
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.
const void * vpx_codec_iter_t
Iterator.
Definition: vpx_codec.h:182
Definition: vpx_encoder.h:157
int max_quantizers[12]
Definition: vpx_encoder.h:714
vpx_codec_frame_flags_t flags
Definition: vpx_encoder.h:185
void * buf
Definition: vpx_encoder.h:179
unsigned int rc_2pass_vbr_maxsection_pct
Two-pass mode per-GOP maximum bitrate.
Definition: vpx_encoder.h:596
vp9 svc extra configure parameters
Definition: vpx_encoder.h:713
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: vpx_encoder.h:371
unsigned int rc_2pass_vbr_minsection_pct
Two-pass mode per-GOP minimum bitrate.
Definition: vpx_encoder.h:589
unsigned int rc_resize_up_thresh
Spatial resampling up watermark.
Definition: vpx_encoder.h:445
Definition: vpx_encoder.h:242
Codec context structure.
Definition: vpx_codec.h:192