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 
93 #define VPX_CODEC_USE_PSNR 0x10000
95 #define VPX_CODEC_USE_OUTPUT_PARTITION 0x20000
96 #define VPX_CODEC_USE_HIGHBITDEPTH 0x40000
102 typedef struct vpx_fixed_buf {
103  void *buf;
104  size_t sz;
105 } vpx_fixed_buf_t;
112 typedef int64_t vpx_codec_pts_t;
113 
121 typedef uint32_t vpx_codec_frame_flags_t;
122 #define VPX_FRAME_IS_KEY 0x1
125 #define VPX_FRAME_IS_DROPPABLE 0x2
126 
127 #define VPX_FRAME_IS_INVISIBLE 0x4
128 
129 #define VPX_FRAME_IS_FRAGMENT 0x8
130 
137 typedef uint32_t vpx_codec_er_flags_t;
139 #define VPX_ERROR_RESILIENT_DEFAULT 0x1
140 
144 #define VPX_ERROR_RESILIENT_PARTITIONS 0x2
145 
157 // Spatial SVC is still experimental and may be removed before the next ABI
158 // bump.
159 #if VPX_ENCODER_ABI_VERSION > (5 + VPX_CODEC_ABI_VERSION)
160  VPX_CODEC_SPATIAL_SVC_LAYER_SIZES,
161  VPX_CODEC_SPATIAL_SVC_LAYER_PSNR,
162 #endif
164 };
165 
171 typedef struct vpx_codec_cx_pkt {
173  union {
174  struct {
175  void *buf;
176  size_t sz;
180  unsigned long duration;
186  } frame;
189  struct vpx_psnr_pkt {
190  unsigned int samples[4];
191  uint64_t sse[4];
192  double psnr[4];
193  } psnr;
195 // Spatial SVC is still experimental and may be removed before the next
196 // ABI bump.
197 #if VPX_ENCODER_ABI_VERSION > (5 + VPX_CODEC_ABI_VERSION)
198  size_t layer_sizes[VPX_SS_MAX_LAYERS];
199  struct vpx_psnr_pkt layer_psnr[VPX_SS_MAX_LAYERS];
200 #endif
201 
202  /* This packet size is fixed to allow codecs to extend this
203  * interface without having to manage storage for raw packets,
204  * i.e., if it's smaller than 128 bytes, you can store in the
205  * packet list directly.
206  */
207  char pad[128 - sizeof(enum vpx_codec_cx_pkt_kind)];
208  } data;
216 // putting the definitions here for now. (agrange: find if there
217 // is a better place for this)
219  void *user_data);
220 
224  void *user_priv;
226 
231 typedef struct vpx_rational {
232  int num;
233  int den;
234 } vpx_rational_t;
241 };
242 
249 };
250 
263 };
264 
273 #define VPX_EFLAG_FORCE_KF (1 << 0)
281 typedef struct vpx_codec_enc_cfg {
282  /*
283  * generic settings (g)
284  */
285 
293  unsigned int g_usage;
294 
301  unsigned int g_threads;
302 
311  unsigned int g_profile;
320  unsigned int g_w;
321 
329  unsigned int g_h;
330 
338 
345  unsigned int g_input_bit_depth;
346 
359  struct vpx_rational g_timebase;
360 
368 
374  enum vpx_enc_pass g_pass;
375 
388  unsigned int g_lag_in_frames;
389 
390  /*
391  * rate control settings (rc)
392  */
393 
410  unsigned int rc_dropframe_thresh;
411 
419  unsigned int rc_resize_allowed;
420 
426  unsigned int rc_scaled_width;
427 
433  unsigned int rc_scaled_height;
434 
441  unsigned int rc_resize_up_thresh;
442 
449  unsigned int rc_resize_down_thresh;
450 
459  enum vpx_rc_mode rc_end_usage;
460 
467 
474 
479  unsigned int rc_target_bitrate;
480 
481  /*
482  * quantizer settings
483  */
484 
493  unsigned int rc_min_quantizer;
494 
503  unsigned int rc_max_quantizer;
504 
505  /*
506  * bitrate tolerance
507  */
508 
519  unsigned int rc_undershoot_pct;
520 
531  unsigned int rc_overshoot_pct;
532 
533  /*
534  * decoder buffer model parameters
535  */
536 
546  unsigned int rc_buf_sz;
547 
555  unsigned int rc_buf_initial_sz;
556 
564  unsigned int rc_buf_optimal_sz;
565 
566  /*
567  * 2 pass rate control parameters
568  */
569 
578  unsigned int rc_2pass_vbr_bias_pct;
579 
586 
593 
594  /*
595  * keyframing settings (kf)
596  */
597 
604  enum vpx_kf_mode kf_mode;
605 
613  unsigned int kf_min_dist;
614 
622  unsigned int kf_max_dist;
623 
624  /*
625  * Spatial scalability settings (ss)
626  */
627 
632  unsigned int ss_number_layers;
633 
639  int ss_enable_auto_alt_ref[VPX_SS_MAX_LAYERS];
640 
646  unsigned int ss_target_bitrate[VPX_SS_MAX_LAYERS];
647 
652  unsigned int ts_number_layers;
653 
659  unsigned int ts_target_bitrate[VPX_TS_MAX_LAYERS];
660 
666  unsigned int ts_rate_decimator[VPX_TS_MAX_LAYERS];
667 
675  unsigned int ts_periodicity;
676 
684  unsigned int ts_layer_id[VPX_TS_MAX_PERIODICITY];
685 
692  unsigned int layer_target_bitrate[VPX_MAX_LAYERS];
693 
709 typedef struct vpx_svc_parameters {
710  int max_quantizers[VPX_MAX_LAYERS];
711  int min_quantizers[VPX_MAX_LAYERS];
712  int scaling_factor_num[VPX_MAX_LAYERS];
713  int scaling_factor_den[VPX_MAX_LAYERS];
714  int speed_per_layer[VPX_MAX_LAYERS];
717 
741  vpx_codec_iface_t *iface,
742  const vpx_codec_enc_cfg_t *cfg,
743  vpx_codec_flags_t flags, int ver);
744 
749 #define vpx_codec_enc_init(ctx, iface, cfg, flags) \
750  vpx_codec_enc_init_ver(ctx, iface, cfg, flags, VPX_ENCODER_ABI_VERSION)
751 
774  int num_enc, vpx_codec_flags_t flags, vpx_rational_t *dsf, int ver);
775 
780 #define vpx_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf) \
781  vpx_codec_enc_init_multi_ver(ctx, iface, cfg, num_enc, flags, dsf, \
782  VPX_ENCODER_ABI_VERSION)
783 
804  vpx_codec_enc_cfg_t *cfg,
805  unsigned int reserved);
806 
822  const vpx_codec_enc_cfg_t *cfg);
823 
836 
838 #define VPX_DL_REALTIME (1)
839 
840 #define VPX_DL_GOOD_QUALITY (1000000)
841 
842 #define VPX_DL_BEST_QUALITY (0)
843 
880  vpx_codec_pts_t pts, unsigned long duration,
882  unsigned long deadline);
883 
928  const vpx_fixed_buf_t *buf,
929  unsigned int pad_before,
930  unsigned int pad_after);
931 
956  vpx_codec_iter_t *iter);
957 
971 
973 #ifdef __cplusplus
974 }
975 #endif
976 #endif // VPX_VPX_ENCODER_H_
Rational Number.
Definition: vpx_encoder.h:231
unsigned int rc_buf_initial_sz
Decoder Buffer Initial Size.
Definition: vpx_encoder.h:555
vpx_fixed_buf_t twopass_stats
Definition: vpx_encoder.h:187
unsigned int ts_number_layers
Number of temporal coding layers.
Definition: vpx_encoder.h:652
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:248
Image Descriptor.
Definition: vpx_image.h:88
int temporal_layering_mode
Definition: vpx_encoder.h:715
vpx_codec_pts_t pts
time stamp to show frame (in timebase units)
Definition: vpx_encoder.h:178
vpx_kf_mode
Keyframe placement mode.
Definition: vpx_encoder.h:259
unsigned int rc_scaled_width
Internal coded frame width.
Definition: vpx_encoder.h:426
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:218
#define VPX_TS_MAX_LAYERS
Definition: vpx_encoder.h:40
Definition: vpx_encoder.h:246
unsigned int g_usage
Algorithm specific "usage" value.
Definition: vpx_encoder.h:293
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:546
unsigned int g_input_bit_depth
Bit-depth of the input frames.
Definition: vpx_encoder.h:345
vpx_codec_enc_output_cx_pkt_cb_fn_t output_cx_pkt
Definition: vpx_encoder.h:223
int den
Definition: vpx_encoder.h:233
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:503
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:493
Definition: vpx_encoder.h:154
unsigned int kf_max_dist
Keyframe maximum interval.
Definition: vpx_encoder.h:622
int partition_id
the partition id defines the decoding order of the partitions. Only applicable when "output partition...
Definition: vpx_encoder.h:185
unsigned int g_lag_in_frames
Allow lagged encoding.
Definition: vpx_encoder.h:388
vpx_enc_pass
Multi-pass Encoding Pass.
Definition: vpx_encoder.h:237
Encoder configuration structure.
Definition: vpx_encoder.h:281
Definition: vpx_encoder.h:163
Definition: vpx_encoder.h:156
Definition: vpx_encoder.h:261
void * user_priv
Definition: vpx_encoder.h:224
Encoder output packet.
Definition: vpx_encoder.h:171
unsigned int rc_overshoot_pct
Rate control adaptation overshoot control.
Definition: vpx_encoder.h:531
void * buf
Definition: vpx_encoder.h:103
unsigned int rc_2pass_vbr_bias_pct
Two-pass mode CBR/VBR bias.
Definition: vpx_encoder.h:578
unsigned int rc_buf_optimal_sz
Decoder Buffer Optimal Size.
Definition: vpx_encoder.h:564
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:102
uint32_t vpx_codec_frame_flags_t
Compressed Frame Flags.
Definition: vpx_encoder.h:121
unsigned int kf_min_dist
Keyframe minimum interval.
Definition: vpx_encoder.h:613
Definition: vpx_encoder.h:239
unsigned int g_profile
Bitstream profile to use.
Definition: vpx_encoder.h:311
Definition: vpx_encoder.h:240
struct vpx_codec_cx_pkt::@1::@2 frame
#define VPX_SS_MAX_LAYERS
Definition: vpx_encoder.h:52
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:320
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:519
uint64_t sse[4]
Definition: vpx_encoder.h:191
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:329
size_t sz
Definition: vpx_encoder.h:176
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:172
unsigned int rc_dropframe_thresh
Temporal resampling configuration, if supported by the codec.
Definition: vpx_encoder.h:410
vpx_fixed_buf_t raw
Definition: vpx_encoder.h:194
Callback function pointer / user data pair storage.
Definition: vpx_encoder.h:222
long vpx_enc_frame_flags_t
Encoded Frame Flags.
Definition: vpx_encoder.h:272
unsigned int rc_target_bitrate
Target data rate.
Definition: vpx_encoder.h:479
int num
Definition: vpx_encoder.h:232
const struct vpx_codec_iface vpx_codec_iface_t
Codec interface structure.
Definition: vpx_codec.h:173
char pad[128-sizeof(enum vpx_codec_cx_pkt_kind)]
Definition: vpx_encoder.h:207
long vpx_codec_flags_t
Initialization-time Feature Enabling.
Definition: vpx_codec.h:166
vpx_codec_cx_pkt_kind
Encoder output packet variants.
Definition: vpx_encoder.h:152
Definition: vpx_encoder.h:245
double psnr[4]
Definition: vpx_encoder.h:192
Definition: vpx_encoder.h:260
unsigned int g_threads
Maximum number of threads to use.
Definition: vpx_encoder.h:301
unsigned int samples[4]
Definition: vpx_encoder.h:190
vpx_fixed_buf_t firstpass_mb_stats
Definition: vpx_encoder.h:188
vpx_rc_mode
Rate control mode.
Definition: vpx_encoder.h:244
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:632
unsigned long duration
duration to show frame (in timebase units)
Definition: vpx_encoder.h:180
vpx_bit_depth_t g_bit_depth
Bit-depth of the codec.
Definition: vpx_encoder.h:337
uint32_t vpx_codec_er_flags_t
Error Resilient flags.
Definition: vpx_encoder.h:137
unsigned int rc_resize_allowed
Enable/disable spatial resampling, if supported by the codec.
Definition: vpx_encoder.h:419
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:701
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:466
Definition: vpx_encoder.h:262
unsigned int rc_scaled_height
Internal coded frame height.
Definition: vpx_encoder.h:433
Definition: vpx_encoder.h:155
vpx_fixed_buf_t rc_firstpass_mb_stats_in
first pass mb stats buffer.
Definition: vpx_encoder.h:473
int64_t vpx_codec_pts_t
Time Stamp Type.
Definition: vpx_encoder.h:112
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:247
Describes the codec algorithm interface to applications.
unsigned int ts_periodicity
Length of the sequence defining frame temporal layer membership.
Definition: vpx_encoder.h:675
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:104
unsigned int rc_resize_down_thresh
Spatial resampling down watermark.
Definition: vpx_encoder.h:449
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:186
Definition: vpx_encoder.h:153
vpx_codec_frame_flags_t flags
Definition: vpx_encoder.h:181
void * buf
Definition: vpx_encoder.h:175
unsigned int rc_2pass_vbr_maxsection_pct
Two-pass mode per-GOP maximum bitrate.
Definition: vpx_encoder.h:592
vp9 svc extra configure parameters
Definition: vpx_encoder.h:709
vpx_codec_er_flags_t g_error_resilient
Enable error resilient modes.
Definition: vpx_encoder.h:367
unsigned int rc_2pass_vbr_minsection_pct
Two-pass mode per-GOP minimum bitrate.
Definition: vpx_encoder.h:585
unsigned int rc_resize_up_thresh
Spatial resampling up watermark.
Definition: vpx_encoder.h:441
Definition: vpx_encoder.h:238
Codec context structure.
Definition: vpx_codec.h:196