aom_encoder.h 34.7 KB
Newer Older
John Koleszar's avatar
John Koleszar committed
1
/*
2
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
John Koleszar's avatar
John Koleszar committed
3
 *
4 5 6 7 8 9
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
John Koleszar's avatar
John Koleszar committed
10
 */
11 12
#ifndef AOM_AOM_ENCODER_H_
#define AOM_AOM_ENCODER_H_
John Koleszar's avatar
John Koleszar committed
13 14 15 16 17 18 19 20 21

/*!\defgroup encoder Encoder Algorithm Interface
 * \ingroup codec
 * This abstraction allows applications using this encoder to easily support
 * multiple video formats with minimal code duplication. This section describes
 * the interface common to all encoders.
 * @{
 */

22
/*!\file
John Koleszar's avatar
John Koleszar committed
23 24 25 26 27 28 29 30 31 32
 * \brief Describes the encoder algorithm interface to applications.
 *
 * This file describes the interface between an application and a
 * video encoder algorithm.
 *
 */
#ifdef __cplusplus
extern "C" {
#endif

33
#include "./aom_codec.h"
John Koleszar's avatar
John Koleszar committed
34

clang-format's avatar
clang-format committed
35 36 37 38 39 40 41 42
/*!\brief Current ABI version number
 *
 * \internal
 * If this file is altered in any way that changes the ABI, this value
 * must be bumped.  Examples include, but are not limited to, changing
 * types, removing or reassigning enums, adding/removing/rearranging
 * fields to structures
 */
43 44
#define AOM_ENCODER_ABI_VERSION \
  (5 + AOM_CODEC_ABI_VERSION) /**<\hideinitializer*/
John Koleszar's avatar
John Koleszar committed
45

clang-format's avatar
clang-format committed
46 47 48
/*! \brief Encoder capabilities bitfield
 *
 *  Each encoder advertises the capabilities it supports as part of its
49
 *  ::aom_codec_iface_t interface structure. Capabilities are extra
clang-format's avatar
clang-format committed
50 51 52
 *  interfaces or functionality, and are not required to be supported
 *  by an encoder.
 *
53
 *  The available flags are specified by AOM_CODEC_CAP_* defines.
clang-format's avatar
clang-format committed
54
 */
55
#define AOM_CODEC_CAP_PSNR 0x10000 /**< Can issue PSNR packets */
John Koleszar's avatar
John Koleszar committed
56

clang-format's avatar
clang-format committed
57
/*! Can output one partition at a time. Each partition is returned in its
58
 *  own AOM_CODEC_CX_FRAME_PKT, with the FRAME_IS_FRAGMENT flag set for
clang-format's avatar
clang-format committed
59 60 61
 *  every partition but the last. In this mode all frames are always
 *  returned partition by partition.
 */
62
#define AOM_CODEC_CAP_OUTPUT_PARTITION 0x20000
63

64 65
/*! Can support input images at greater than 8 bitdepth.
 */
66
#define AOM_CODEC_CAP_HIGHBITDEPTH 0x40000
John Koleszar's avatar
John Koleszar committed
67

clang-format's avatar
clang-format committed
68 69 70 71 72
/*! \brief Initialization-time Feature Enabling
 *
 *  Certain codec features must be known at initialization time, to allow
 *  for proper memory allocation.
 *
73
 *  The available flags are specified by AOM_CODEC_USE_* defines.
clang-format's avatar
clang-format committed
74
 */
75
#define AOM_CODEC_USE_PSNR 0x10000 /**< Calculate PSNR on each frame */
clang-format's avatar
clang-format committed
76
/*!\brief Make the encoder output one  partition at a time. */
77 78
#define AOM_CODEC_USE_OUTPUT_PARTITION 0x20000
#define AOM_CODEC_USE_HIGHBITDEPTH 0x40000 /**< Use high bitdepth */
John Koleszar's avatar
John Koleszar committed
79

clang-format's avatar
clang-format committed
80 81 82 83
/*!\brief Generic fixed size buffer structure
 *
 * This structure is able to hold a reference to any fixed size buffer.
 */
84
typedef struct aom_fixed_buf {
clang-format's avatar
clang-format committed
85 86
  void *buf;       /**< Pointer to the data */
  size_t sz;       /**< Length of the buffer, in chars */
87
} aom_fixed_buf_t; /**< alias for struct aom_fixed_buf */
John Koleszar's avatar
John Koleszar committed
88

clang-format's avatar
clang-format committed
89 90 91 92 93
/*!\brief Time Stamp Type
 *
 * An integer, which when multiplied by the stream's time base, provides
 * the absolute time of a sample.
 */
94
typedef int64_t aom_codec_pts_t;
John Koleszar's avatar
John Koleszar committed
95

clang-format's avatar
clang-format committed
96 97 98 99 100 101 102
/*!\brief Compressed Frame Flags
 *
 * This type represents a bitfield containing information about a compressed
 * frame that may be useful to an application. The most significant 16 bits
 * can be used by an algorithm to provide additional detail, for example to
 * support frame types that are codec specific (MPEG-1 D-frames for example)
 */
103 104
typedef uint32_t aom_codec_frame_flags_t;
#define AOM_FRAME_IS_KEY 0x1 /**< frame is the start of a GOP */
clang-format's avatar
clang-format committed
105 106
/*!\brief frame can be dropped without affecting the stream (no future frame
 * depends on this one) */
107
#define AOM_FRAME_IS_DROPPABLE 0x2
clang-format's avatar
clang-format committed
108
/*!\brief frame should be decoded but will not be shown */
109
#define AOM_FRAME_IS_INVISIBLE 0x4
clang-format's avatar
clang-format committed
110
/*!\brief this is a fragment of the encoded frame */
111
#define AOM_FRAME_IS_FRAGMENT 0x8
clang-format's avatar
clang-format committed
112 113 114 115 116

/*!\brief Error Resilient flags
 *
 * These flags define which error resilient features to enable in the
 * encoder. The flags are specified through the
117
 * aom_codec_enc_cfg::g_error_resilient variable.
clang-format's avatar
clang-format committed
118
 */
119
typedef uint32_t aom_codec_er_flags_t;
clang-format's avatar
clang-format committed
120
/*!\brief Improve resiliency against losses of whole frames */
121
#define AOM_ERROR_RESILIENT_DEFAULT 0x1
clang-format's avatar
clang-format committed
122 123 124 125
/*!\brief The frame partitions are independently decodable by the bool decoder,
 * meaning that partitions can be decoded even though earlier partitions have
 * been lost. Note that intra prediction is still done over the partition
 * boundary. */
126
#define AOM_ERROR_RESILIENT_PARTITIONS 0x2
clang-format's avatar
clang-format committed
127 128 129 130

/*!\brief Encoder output packet variants
 *
 * This enumeration lists the different kinds of data packets that can be
131
 * returned by calls to aom_codec_get_cx_data(). Algorithms \ref MAY
clang-format's avatar
clang-format committed
132 133
 * extend this list to provide additional functionality.
 */
134 135 136 137 138 139
enum aom_codec_cx_pkt_kind {
  AOM_CODEC_CX_FRAME_PKT,    /**< Compressed video frame */
  AOM_CODEC_STATS_PKT,       /**< Two-pass statistics for this frame */
  AOM_CODEC_FPMB_STATS_PKT,  /**< first pass mb statistics for this frame */
  AOM_CODEC_PSNR_PKT,        /**< PSNR statistics for this frame */
  AOM_CODEC_CUSTOM_PKT = 256 /**< Algorithm extensions  */
clang-format's avatar
clang-format committed
140 141 142 143 144 145 146
};

/*!\brief Encoder output packet
 *
 * This structure contains the different kinds of output data the encoder
 * may produce while compressing a frame.
 */
147 148
typedef struct aom_codec_cx_pkt {
  enum aom_codec_cx_pkt_kind kind; /**< packet variant */
clang-format's avatar
clang-format committed
149 150 151 152 153
  union {
    struct {
      void *buf; /**< compressed data buffer */
      size_t sz; /**< length of compressed data */
      /*!\brief time stamp to show frame (in timebase units) */
154
      aom_codec_pts_t pts;
clang-format's avatar
clang-format committed
155 156
      /*!\brief duration to show frame (in timebase units) */
      unsigned long duration;
157
      aom_codec_frame_flags_t flags; /**< flags for this frame */
clang-format's avatar
clang-format committed
158 159 160 161 162
      /*!\brief the partition id defines the decoding order of the partitions.
       * Only applicable when "output partition" mode is enabled. First
       * partition has id 0.*/
      int partition_id;
    } frame;                            /**< data for compressed frame packet */
163 164 165
    aom_fixed_buf_t twopass_stats;      /**< data for two-pass packet */
    aom_fixed_buf_t firstpass_mb_stats; /**< first pass mb packet */
    struct aom_psnr_pkt {
clang-format's avatar
clang-format committed
166 167 168 169
      unsigned int samples[4]; /**< Number of samples, total/y/u/v */
      uint64_t sse[4];         /**< sum squared error, total/y/u/v */
      double psnr[4];          /**< PSNR, total/y/u/v */
    } psnr;                    /**< data for PSNR packet */
170
    aom_fixed_buf_t raw;       /**< data for arbitrary packets */
clang-format's avatar
clang-format committed
171 172 173 174 175 176

    /* This packet size is fixed to allow codecs to extend this
     * interface without having to manage storage for raw packets,
     * i.e., if it's smaller than 128 bytes, you can store in the
     * packet list directly.
     */
177
    char pad[128 - sizeof(enum aom_codec_cx_pkt_kind)]; /**< fixed sz */
clang-format's avatar
clang-format committed
178
  } data;                                               /**< packet data */
179
} aom_codec_cx_pkt_t; /**< alias for struct aom_codec_cx_pkt */
clang-format's avatar
clang-format committed
180 181 182 183 184

/*!\brief Rational Number
 *
 * This structure holds a fractional value.
 */
185
typedef struct aom_rational {
clang-format's avatar
clang-format committed
186 187
  int num;        /**< fraction numerator */
  int den;        /**< fraction denominator */
188
} aom_rational_t; /**< alias for struct aom_rational */
clang-format's avatar
clang-format committed
189 190

/*!\brief Multi-pass Encoding Pass */
191 192 193 194
enum aom_enc_pass {
  AOM_RC_ONE_PASS,   /**< Single pass mode */
  AOM_RC_FIRST_PASS, /**< First pass of multi-pass mode */
  AOM_RC_LAST_PASS   /**< Final pass of multi-pass mode */
clang-format's avatar
clang-format committed
195 196 197
};

/*!\brief Rate control mode */
198 199 200 201 202
enum aom_rc_mode {
  AOM_VBR, /**< Variable Bit Rate (VBR) mode */
  AOM_CBR, /**< Constant Bit Rate (CBR) mode */
  AOM_CQ,  /**< Constrained Quality (CQ)  mode */
  AOM_Q,   /**< Constant Quality (Q) mode */
clang-format's avatar
clang-format committed
203 204 205 206 207 208
};

/*!\brief Keyframe placement mode.
 *
 * This enumeration determines whether keyframes are placed automatically by
 * the encoder or whether this behavior is disabled. Older releases of this
209
 * SDK were implemented such that AOM_KF_FIXED meant keyframes were disabled.
clang-format's avatar
clang-format committed
210
 * This name is confusing for this behavior, so the new symbols to be used
211
 * are AOM_KF_AUTO and AOM_KF_DISABLED.
clang-format's avatar
clang-format committed
212
 */
213 214 215 216
enum aom_kf_mode {
  AOM_KF_FIXED,       /**< deprecated, implies AOM_KF_DISABLED */
  AOM_KF_AUTO,        /**< Encoder determines optimal placement automatically */
  AOM_KF_DISABLED = 0 /**< Encoder does not place keyframes. */
clang-format's avatar
clang-format committed
217
};
John Koleszar's avatar
John Koleszar committed
218

clang-format's avatar
clang-format committed
219 220
/*!\brief Encoded Frame Flags
 *
221
 * This type indicates a bitfield to be passed to aom_codec_encode(), defining
clang-format's avatar
clang-format committed
222
 * per-frame boolean values. By convention, bits common to all codecs will be
223
 * named AOM_EFLAG_*, and bits specific to an algorithm will be named
clang-format's avatar
clang-format committed
224 225
 * /algo/_eflag_*. The lower order 16 bits are reserved for common use.
 */
226 227
typedef long aom_enc_frame_flags_t;
#define AOM_EFLAG_FORCE_KF (1 << 0) /**< Force this frame to be a keyframe */
John Koleszar's avatar
John Koleszar committed
228

clang-format's avatar
clang-format committed
229 230 231 232 233 234
/*!\brief Encoder configuration structure
 *
 * This structure contains the encoder settings that have common representations
 * across all codecs. This doesn't imply that all codecs support all features,
 * however.
 */
235
typedef struct aom_codec_enc_cfg {
clang-format's avatar
clang-format committed
236 237 238
  /*
   * generic settings (g)
   */
John Koleszar's avatar
John Koleszar committed
239

clang-format's avatar
clang-format committed
240 241 242 243 244 245 246 247
  /*!\brief Algorithm specific "usage" value
   *
   * Algorithms may define multiple values for usage, which may convey the
   * intent of how the application intends to use the stream. If this value
   * is non-zero, consult the documentation for the codec to determine its
   * meaning.
   */
  unsigned int g_usage;
John Koleszar's avatar
John Koleszar committed
248

clang-format's avatar
clang-format committed
249 250 251 252 253 254 255
  /*!\brief Maximum number of threads to use
   *
   * For multi-threaded implementations, use no more than this number of
   * threads. The codec may use fewer threads than allowed. The value
   * 0 is equivalent to the value 1.
   */
  unsigned int g_threads;
John Koleszar's avatar
John Koleszar committed
256

clang-format's avatar
clang-format committed
257 258 259 260 261 262 263 264 265
  /*!\brief Bitstream profile to use
   *
   * Some codecs support a notion of multiple bitstream profiles. Typically
   * this maps to a set of features that are turned on or off. Often the
   * profile to use is determined by the features of the intended decoder.
   * Consult the documentation for the codec to determine the valid values
   * for this parameter, or set to zero for a sane default.
   */
  unsigned int g_profile; /**< profile of bitstream to use */
John Koleszar's avatar
John Koleszar committed
266

clang-format's avatar
clang-format committed
267 268 269 270 271 272 273 274
  /*!\brief Width of the frame
   *
   * This value identifies the presentation resolution of the frame,
   * in pixels. Note that the frames passed as input to the encoder must
   * have this resolution. Frames will be presented by the decoder in this
   * resolution, independent of any spatial resampling the encoder may do.
   */
  unsigned int g_w;
John Koleszar's avatar
John Koleszar committed
275

clang-format's avatar
clang-format committed
276 277 278 279 280 281 282 283
  /*!\brief Height of the frame
   *
   * This value identifies the presentation resolution of the frame,
   * in pixels. Note that the frames passed as input to the encoder must
   * have this resolution. Frames will be presented by the decoder in this
   * resolution, independent of any spatial resampling the encoder may do.
   */
  unsigned int g_h;
John Koleszar's avatar
John Koleszar committed
284

clang-format's avatar
clang-format committed
285 286 287 288
  /*!\brief Bit-depth of the codec
   *
   * This value identifies the bit_depth of the codec,
   * Only certain bit-depths are supported as identified in the
289
   * aom_bit_depth_t enum.
clang-format's avatar
clang-format committed
290
   */
291
  aom_bit_depth_t g_bit_depth;
John Koleszar's avatar
John Koleszar committed
292

clang-format's avatar
clang-format committed
293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312
  /*!\brief Bit-depth of the input frames
   *
   * This value identifies the bit_depth of the input frames in bits.
   * Note that the frames passed as input to the encoder must have
   * this bit-depth.
   */
  unsigned int g_input_bit_depth;

  /*!\brief Stream timebase units
   *
   * Indicates the smallest interval of time, in seconds, used by the stream.
   * For fixed frame rate material, or variable frame rate material where
   * frames are timed at a multiple of a given clock (ex: video capture),
   * the \ref RECOMMENDED method is to set the timebase to the reciprocal
   * of the frame rate (ex: 1001/30000 for 29.970 Hz NTSC). This allows the
   * pts to correspond to the frame number, which can be handy. For
   * re-encoding video from containers with absolute time timestamps, the
   * \ref RECOMMENDED method is to set the timebase to that of the parent
   * container or multimedia framework (ex: 1/1000 for ms, as in FLV).
   */
313
  struct aom_rational g_timebase;
John Koleszar's avatar
John Koleszar committed
314

clang-format's avatar
clang-format committed
315 316 317 318 319 320
  /*!\brief Enable error resilient modes.
   *
   * The error resilient bitfield indicates to the encoder which features
   * it should enable to take measures for streaming over lossy or noisy
   * links.
   */
321
  aom_codec_er_flags_t g_error_resilient;
John Koleszar's avatar
John Koleszar committed
322

clang-format's avatar
clang-format committed
323 324 325
  /*!\brief Multi-pass Encoding Mode
   *
   * This value should be set to the current phase for multi-pass encoding.
326
   * For single pass, set to #AOM_RC_ONE_PASS.
clang-format's avatar
clang-format committed
327
   */
328
  enum aom_enc_pass g_pass;
John Koleszar's avatar
John Koleszar committed
329

clang-format's avatar
clang-format committed
330 331 332 333 334 335 336 337 338 339 340 341 342
  /*!\brief Allow lagged encoding
   *
   * If set, this value allows the encoder to consume a number of input
   * frames before producing output frames. This allows the encoder to
   * base decisions for the current frame on future frames. This does
   * increase the latency of the encoding pipeline, so it is not appropriate
   * in all situations (ex: realtime encoding).
   *
   * Note that this is a maximum value -- the encoder may produce frames
   * sooner than the given limit. Set this value to 0 to disable this
   * feature.
   */
  unsigned int g_lag_in_frames;
John Koleszar's avatar
John Koleszar committed
343

clang-format's avatar
clang-format committed
344 345 346
  /*
   * rate control settings (rc)
   */
John Koleszar's avatar
John Koleszar committed
347

clang-format's avatar
clang-format committed
348
  /*!\brief Temporal resampling configuration, if supported by the codec.
John Koleszar's avatar
John Koleszar committed
349
   *
clang-format's avatar
clang-format committed
350 351 352 353 354
   * Temporal resampling allows the codec to "drop" frames as a strategy to
   * meet its target data rate. This can cause temporal discontinuities in
   * the encoded video, which may appear as stuttering during playback. This
   * trade-off is often acceptable, but for many applications is not. It can
   * be disabled in these cases.
John Koleszar's avatar
John Koleszar committed
355
   *
356
   * Note that not all codecs support this feature. All aom AVx codecs do.
clang-format's avatar
clang-format committed
357
   * For other codecs, consult the documentation for that algorithm.
John Koleszar's avatar
John Koleszar committed
358
   *
clang-format's avatar
clang-format committed
359 360 361 362
   * This threshold is described as a percentage of the target data buffer.
   * When the data buffer falls below this percentage of fullness, a
   * dropped frame is indicated. Set the threshold to zero (0) to disable
   * this feature.
John Koleszar's avatar
John Koleszar committed
363
   */
clang-format's avatar
clang-format committed
364
  unsigned int rc_dropframe_thresh;
John Koleszar's avatar
John Koleszar committed
365

366
  /*!\brief Mode for spatial resampling, if supported by the codec.
John Koleszar's avatar
John Koleszar committed
367
   *
clang-format's avatar
clang-format committed
368
   * Spatial resampling allows the codec to compress a lower resolution
369
   * version of the frame, which is then upscaled by the decoder to the
clang-format's avatar
clang-format committed
370 371
   * correct presentation resolution. This increases visual quality at
   * low data rates, at the expense of CPU time on the encoder/decoder.
John Koleszar's avatar
John Koleszar committed
372
   */
373
  unsigned int rc_resize_mode;
John Koleszar's avatar
John Koleszar committed
374

375
  /*!\brief Frame resize denominator.
John Koleszar's avatar
John Koleszar committed
376
   *
377
   * The denominator for resize to use, assuming 8 as the numerator.
John Koleszar's avatar
John Koleszar committed
378
   *
379
   * Valid denominators are  8 - 16 for now.
clang-format's avatar
clang-format committed
380
   */
381
  unsigned int rc_resize_denominator;
clang-format's avatar
clang-format committed
382

383
  /*!\brief Keyframe resize denominator.
384
   *
385
   * The denominator for resize to use, assuming 8 as the numerator.
386
   *
387
   * Valid denominators are  8 - 16 for now.
388
   */
389
  unsigned int rc_resize_kf_denominator;
390

391 392 393 394 395 396
  /*!\brief Frame super-resolution scaling mode.
   *
   * Similar to spatial resampling, frame super-resolution integrates
   * upscaling after the encode/decode process. Taking control of upscaling and
   * using restoration filters should allow it to outperform normal resizing.
   *
397 398
   * Mode 0 is SUPERRES_NONE, mode 1 is SUPERRES_FIXED, mode 2 is
   * SUPERRES_RANDOM and mode 3 is SUPERRES_QTHRESH.
399 400 401
   */
  unsigned int rc_superres_mode;

402
  /*!\brief Frame super-resolution denominator.
403
   *
404
   * The denominator for superres to use. If fixed it will only change if the
405 406 407
   * cumulative scale change over resizing and superres is greater than 1/2;
   * this forces superres to reduce scaling.
   *
408
   * Valid denominators are 8 to 16.
409
   *
410
   * Used only by SUPERRES_FIXED.
411
   */
412
  unsigned int rc_superres_denominator;
413

414
  /*!\brief Keyframe super-resolution denominator.
415
   *
416
   * The denominator for superres to use. If fixed it will only change if the
417 418 419
   * cumulative scale change over resizing and superres is greater than 1/2;
   * this forces superres to reduce scaling.
   *
420
   * Valid denominators are 8 - 16 for now.
421
   */
422
  unsigned int rc_superres_kf_denominator;
423

424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441
  /*!\brief Frame super-resolution q threshold.
   *
   * The q level threshold after which superres is used.
   * Valid values are 1 to 63.
   *
   * Used only by SUPERRES_QTHRESH
   */
  unsigned int rc_superres_qthresh;

  /*!\brief Keyframe super-resolution q threshold.
   *
   * The q level threshold after which superres is used for key frames.
   * Valid values are 1 to 63.
   *
   * Used only by SUPERRES_QTHRESH
   */
  unsigned int rc_superres_kf_qthresh;

clang-format's avatar
clang-format committed
442
  /*!\brief Rate control algorithm to use.
John Koleszar's avatar
John Koleszar committed
443
   *
clang-format's avatar
clang-format committed
444 445 446 447 448 449
   * Indicates whether the end usage of this stream is to be streamed over
   * a bandwidth constrained link, indicating that Constant Bit Rate (CBR)
   * mode should be used, or whether it will be played back on a high
   * bandwidth link, as from a local disk, where higher variations in
   * bitrate are acceptable.
   */
450
  enum aom_rc_mode rc_end_usage;
clang-format's avatar
clang-format committed
451 452

  /*!\brief Two-pass stats buffer.
John Koleszar's avatar
John Koleszar committed
453
   *
clang-format's avatar
clang-format committed
454 455 456
   * A buffer containing all of the stats packets produced in the first
   * pass, concatenated.
   */
457
  aom_fixed_buf_t rc_twopass_stats_in;
clang-format's avatar
clang-format committed
458 459

  /*!\brief first pass mb stats buffer.
John Koleszar's avatar
John Koleszar committed
460
   *
clang-format's avatar
clang-format committed
461 462
   * A buffer containing all of the first pass mb stats packets produced
   * in the first pass, concatenated.
John Koleszar's avatar
John Koleszar committed
463
   */
464
  aom_fixed_buf_t rc_firstpass_mb_stats_in;
John Koleszar's avatar
John Koleszar committed
465

clang-format's avatar
clang-format committed
466
  /*!\brief Target data rate
John Koleszar's avatar
John Koleszar committed
467
   *
clang-format's avatar
clang-format committed
468 469 470 471 472 473 474 475 476
   * Target bandwidth to use for this stream, in kilobits per second.
   */
  unsigned int rc_target_bitrate;

  /*
   * quantizer settings
   */

  /*!\brief Minimum (Best Quality) Quantizer
John Koleszar's avatar
John Koleszar committed
477
   *
clang-format's avatar
clang-format committed
478 479 480 481
   * The quantizer is the most direct control over the quality of the
   * encoded image. The range of valid values for the quantizer is codec
   * specific. Consult the documentation for the codec to determine the
   * values to use. To determine the range programmatically, call
482
   * aom_codec_enc_config_default() with a usage value of 0.
clang-format's avatar
clang-format committed
483 484 485 486
   */
  unsigned int rc_min_quantizer;

  /*!\brief Maximum (Worst Quality) Quantizer
John Koleszar's avatar
John Koleszar committed
487
   *
clang-format's avatar
clang-format committed
488 489 490 491
   * The quantizer is the most direct control over the quality of the
   * encoded image. The range of valid values for the quantizer is codec
   * specific. Consult the documentation for the codec to determine the
   * values to use. To determine the range programmatically, call
492
   * aom_codec_enc_config_default() with a usage value of 0.
clang-format's avatar
clang-format committed
493 494 495 496 497 498 499 500
   */
  unsigned int rc_max_quantizer;

  /*
   * bitrate tolerance
   */

  /*!\brief Rate control adaptation undershoot control
John Koleszar's avatar
John Koleszar committed
501
   *
clang-format's avatar
clang-format committed
502 503 504 505 506
   * This value, expressed as a percentage of the target bitrate,
   * controls the maximum allowed adaptation speed of the codec.
   * This factor controls the maximum amount of bits that can
   * be subtracted from the target bitrate in order to compensate
   * for prior overshoot.
John Koleszar's avatar
John Koleszar committed
507
   *
clang-format's avatar
clang-format committed
508 509 510 511 512
   * Valid values in the range 0-1000.
   */
  unsigned int rc_undershoot_pct;

  /*!\brief Rate control adaptation overshoot control
John Koleszar's avatar
John Koleszar committed
513
   *
clang-format's avatar
clang-format committed
514 515 516 517 518
   * This value, expressed as a percentage of the target bitrate,
   * controls the maximum allowed adaptation speed of the codec.
   * This factor controls the maximum amount of bits that can
   * be added to the target bitrate in order to compensate for
   * prior undershoot.
John Koleszar's avatar
John Koleszar committed
519
   *
clang-format's avatar
clang-format committed
520
   * Valid values in the range 0-1000.
John Koleszar's avatar
John Koleszar committed
521
   */
clang-format's avatar
clang-format committed
522
  unsigned int rc_overshoot_pct;
John Koleszar's avatar
John Koleszar committed
523

clang-format's avatar
clang-format committed
524 525 526
  /*
   * decoder buffer model parameters
   */
John Koleszar's avatar
John Koleszar committed
527

clang-format's avatar
clang-format committed
528
  /*!\brief Decoder Buffer Size
John Koleszar's avatar
John Koleszar committed
529
   *
clang-format's avatar
clang-format committed
530 531 532 533 534 535 536 537 538 539
   * This value indicates the amount of data that may be buffered by the
   * decoding application. Note that this value is expressed in units of
   * time (milliseconds). For example, a value of 5000 indicates that the
   * client will buffer (at least) 5000ms worth of encoded data. Use the
   * target bitrate (#rc_target_bitrate) to convert to bits/bytes, if
   * necessary.
   */
  unsigned int rc_buf_sz;

  /*!\brief Decoder Buffer Initial Size
John Koleszar's avatar
John Koleszar committed
540
   *
clang-format's avatar
clang-format committed
541 542 543 544 545 546 547 548
   * This value indicates the amount of data that will be buffered by the
   * decoding application prior to beginning playback. This value is
   * expressed in units of time (milliseconds). Use the target bitrate
   * (#rc_target_bitrate) to convert to bits/bytes, if necessary.
   */
  unsigned int rc_buf_initial_sz;

  /*!\brief Decoder Buffer Optimal Size
John Koleszar's avatar
John Koleszar committed
549
   *
clang-format's avatar
clang-format committed
550 551 552 553 554 555 556 557 558 559 560 561
   * This value indicates the amount of data that the encoder should try
   * to maintain in the decoder's buffer. This value is expressed in units
   * of time (milliseconds). Use the target bitrate (#rc_target_bitrate)
   * to convert to bits/bytes, if necessary.
   */
  unsigned int rc_buf_optimal_sz;

  /*
   * 2 pass rate control parameters
   */

  /*!\brief Two-pass mode CBR/VBR bias
John Koleszar's avatar
John Koleszar committed
562
   *
clang-format's avatar
clang-format committed
563 564 565 566 567 568 569 570 571
   * Bias, expressed on a scale of 0 to 100, for determining target size
   * for the current frame. The value 0 indicates the optimal CBR mode
   * value should be used. The value 100 indicates the optimal VBR mode
   * value should be used. Values in between indicate which way the
   * encoder should "lean."
   */
  unsigned int rc_2pass_vbr_bias_pct;

  /*!\brief Two-pass mode per-GOP minimum bitrate
John Koleszar's avatar
John Koleszar committed
572
   *
clang-format's avatar
clang-format committed
573 574 575 576 577 578
   * This value, expressed as a percentage of the target bitrate, indicates
   * the minimum bitrate to be used for a single GOP (aka "section")
   */
  unsigned int rc_2pass_vbr_minsection_pct;

  /*!\brief Two-pass mode per-GOP maximum bitrate
John Koleszar's avatar
John Koleszar committed
579
   *
clang-format's avatar
clang-format committed
580 581
   * This value, expressed as a percentage of the target bitrate, indicates
   * the maximum bitrate to be used for a single GOP (aka "section")
John Koleszar's avatar
John Koleszar committed
582
   */
clang-format's avatar
clang-format committed
583
  unsigned int rc_2pass_vbr_maxsection_pct;
John Koleszar's avatar
John Koleszar committed
584

clang-format's avatar
clang-format committed
585 586 587
  /*
   * keyframing settings (kf)
   */
John Koleszar's avatar
John Koleszar committed
588

clang-format's avatar
clang-format committed
589
  /*!\brief Keyframe placement mode
John Koleszar's avatar
John Koleszar committed
590
   *
clang-format's avatar
clang-format committed
591 592 593 594
   * This value indicates whether the encoder should place keyframes at a
   * fixed interval, or determine the optimal placement automatically
   * (as governed by the #kf_min_dist and #kf_max_dist parameters)
   */
595
  enum aom_kf_mode kf_mode;
clang-format's avatar
clang-format committed
596 597

  /*!\brief Keyframe minimum interval
John Koleszar's avatar
John Koleszar committed
598
   *
clang-format's avatar
clang-format committed
599 600 601 602 603 604 605 606
   * This value, expressed as a number of frames, prevents the encoder from
   * placing a keyframe nearer than kf_min_dist to the previous keyframe. At
   * least kf_min_dist frames non-keyframes will be coded before the next
   * keyframe. Set kf_min_dist equal to kf_max_dist for a fixed interval.
   */
  unsigned int kf_min_dist;

  /*!\brief Keyframe maximum interval
John Koleszar's avatar
John Koleszar committed
607
   *
clang-format's avatar
clang-format committed
608 609 610 611
   * This value, expressed as a number of frames, forces the encoder to code
   * a keyframe if one has not been coded in the last kf_max_dist frames.
   * A value of 0 implies all frames will be keyframes. Set kf_min_dist
   * equal to kf_max_dist for a fixed interval.
John Koleszar's avatar
John Koleszar committed
612
   */
clang-format's avatar
clang-format committed
613
  unsigned int kf_max_dist;
614 615 616 617 618 619 620 621

  /*!\brief Tile coding mode
   *
   * This value indicates the tile coding mode.
   * A value of 0 implies a normal non-large-scale tile coding. A value of 1
   * implies a large-scale tile coding.
   */
  unsigned int large_scale_tile;
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663

  /*!\brief Number of explicit tile widths specified
   *
   * This value indicates the number of tile widths specified
   * A value of 0 implies no tile widths are specified.
   * Tile widths are given in the array tile_widths[]
   */
  int tile_width_count;

  /*!\brief Number of explicit tile heights specified
   *
   * This value indicates the number of tile heights specified
   * A value of 0 implies no tile heights are specified.
   * Tile heights are given in the array tile_heights[]
   */
  int tile_height_count;

/*!\brief Maximum number of tile widths in tile widths array
 *
 * This define gives the maximum number of elements in the tile_widths array.
 */
#define MAX_TILE_WIDTHS 64  // maximum tile width array length

  /*!\brief Array of specified tile widths
   *
   * This array specifies tile widths (and may be empty)
   * The number of widths specified is given by tile_width_count
   */
  int tile_widths[MAX_TILE_WIDTHS];

/*!\brief Maximum number of tile heights in tile heights array.
 *
 * This define gives the maximum number of elements in the tile_heights array.
 */
#define MAX_TILE_HEIGHTS 64  // maximum tile height array length

  /*!\brief Array of specified tile heights
   *
   * This array specifies tile heights (and may be empty)
   * The number of heights specified is given by tile_height_count
   */
  int tile_heights[MAX_TILE_HEIGHTS];
664
} aom_codec_enc_cfg_t; /**< alias for struct aom_codec_enc_cfg */
clang-format's avatar
clang-format committed
665 666 667 668

/*!\brief Initialize an encoder instance
 *
 * Initializes a encoder context using the given interface. Applications
669
 * should call the aom_codec_enc_init convenience macro instead of this
clang-format's avatar
clang-format committed
670 671 672 673 674 675 676 677 678
 * function directly, to ensure that the ABI version number parameter
 * is properly initialized.
 *
 * If the library was configured with --disable-multithread, this call
 * is not thread safe and should be guarded with a lock if being used
 * in a multithreaded context.
 *
 * \param[in]    ctx     Pointer to this instance's context.
 * \param[in]    iface   Pointer to the algorithm interface to use.
Yaowu Xu's avatar
Yaowu Xu committed
679
 * \param[in]    cfg     Configuration to use, if known.
680
 * \param[in]    flags   Bitfield of AOM_CODEC_USE_* flags
clang-format's avatar
clang-format committed
681
 * \param[in]    ver     ABI version number. Must be set to
682 683
 *                       AOM_ENCODER_ABI_VERSION
 * \retval #AOM_CODEC_OK
clang-format's avatar
clang-format committed
684
 *     The decoder algorithm initialized.
685
 * \retval #AOM_CODEC_MEM_ERROR
clang-format's avatar
clang-format committed
686 687
 *     Memory allocation failed.
 */
688 689 690 691
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);
clang-format's avatar
clang-format committed
692

693
/*!\brief Convenience macro for aom_codec_enc_init_ver()
clang-format's avatar
clang-format committed
694 695 696
 *
 * Ensures the ABI version parameter is properly set.
 */
697 698
#define aom_codec_enc_init(ctx, iface, cfg, flags) \
  aom_codec_enc_init_ver(ctx, iface, cfg, flags, AOM_ENCODER_ABI_VERSION)
clang-format's avatar
clang-format committed
699 700 701 702

/*!\brief Initialize multi-encoder instance
 *
 * Initializes multi-encoder context using the given interface.
703
 * Applications should call the aom_codec_enc_init_multi convenience macro
clang-format's avatar
clang-format committed
704 705 706 707 708
 * instead of this function directly, to ensure that the ABI version number
 * parameter is properly initialized.
 *
 * \param[in]    ctx     Pointer to this instance's context.
 * \param[in]    iface   Pointer to the algorithm interface to use.
Yaowu Xu's avatar
Yaowu Xu committed
709
 * \param[in]    cfg     Configuration to use, if known.
clang-format's avatar
clang-format committed
710
 * \param[in]    num_enc Total number of encoders.
711
 * \param[in]    flags   Bitfield of AOM_CODEC_USE_* flags
clang-format's avatar
clang-format committed
712 713
 * \param[in]    dsf     Pointer to down-sampling factors.
 * \param[in]    ver     ABI version number. Must be set to
714 715
 *                       AOM_ENCODER_ABI_VERSION
 * \retval #AOM_CODEC_OK
clang-format's avatar
clang-format committed
716
 *     The decoder algorithm initialized.
717
 * \retval #AOM_CODEC_MEM_ERROR
clang-format's avatar
clang-format committed
718 719
 *     Memory allocation failed.
 */
720 721 722
aom_codec_err_t aom_codec_enc_init_multi_ver(
    aom_codec_ctx_t *ctx, aom_codec_iface_t *iface, aom_codec_enc_cfg_t *cfg,
    int num_enc, aom_codec_flags_t flags, aom_rational_t *dsf, int ver);
clang-format's avatar
clang-format committed
723

724
/*!\brief Convenience macro for aom_codec_enc_init_multi_ver()
clang-format's avatar
clang-format committed
725 726 727
 *
 * Ensures the ABI version parameter is properly set.
 */
728 729 730
#define aom_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf) \
  aom_codec_enc_init_multi_ver(ctx, iface, cfg, num_enc, flags, dsf,   \
                               AOM_ENCODER_ABI_VERSION)
clang-format's avatar
clang-format committed
731 732 733 734 735 736 737 738 739 740 741

/*!\brief Get a default configuration
 *
 * Initializes a encoder configuration structure with default values. Supports
 * the notion of "usages" so that an algorithm may offer different default
 * settings depending on the user's intended goal. This function \ref SHOULD
 * be called by all applications to initialize the configuration structure
 * before specializing the configuration with application specific values.
 *
 * \param[in]    iface     Pointer to the algorithm interface to use.
 * \param[out]   cfg       Configuration buffer to populate.
742
 * \param[in]    reserved  Must set to 0.
clang-format's avatar
clang-format committed
743
 *
744
 * \retval #AOM_CODEC_OK
clang-format's avatar
clang-format committed
745
 *     The configuration was populated.
746
 * \retval #AOM_CODEC_INCAPABLE
clang-format's avatar
clang-format committed
747
 *     Interface is not an encoder interface.
748
 * \retval #AOM_CODEC_INVALID_PARAM
clang-format's avatar
clang-format committed
749 750
 *     A parameter was NULL, or the usage value was not recognized.
 */
751 752
aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface,
                                             aom_codec_enc_cfg_t *cfg,
clang-format's avatar
clang-format committed
753 754 755 756 757 758 759 760 761
                                             unsigned int reserved);

/*!\brief Set or change configuration
 *
 * Reconfigures an encoder instance according to the given configuration.
 *
 * \param[in]    ctx     Pointer to this instance's context
 * \param[in]    cfg     Configuration buffer to use
 *
762
 * \retval #AOM_CODEC_OK
clang-format's avatar
clang-format committed
763
 *     The configuration was populated.
764
 * \retval #AOM_CODEC_INCAPABLE
clang-format's avatar
clang-format committed
765
 *     Interface is not an encoder interface.
766
 * \retval #AOM_CODEC_INVALID_PARAM
clang-format's avatar
clang-format committed
767 768
 *     A parameter was NULL, or the usage value was not recognized.
 */
769 770
aom_codec_err_t aom_codec_enc_config_set(aom_codec_ctx_t *ctx,
                                         const aom_codec_enc_cfg_t *cfg);
clang-format's avatar
clang-format committed
771 772 773 774 775 776 777 778 779 780 781 782

/*!\brief Get global stream headers
 *
 * Retrieves a stream level global header packet, if supported by the codec.
 *
 * \param[in]    ctx     Pointer to this instance's context
 *
 * \retval NULL
 *     Encoder does not support global header
 * \retval Non-NULL
 *     Pointer to buffer containing global header packet
 */
783 784 785 786
aom_fixed_buf_t *aom_codec_get_global_headers(aom_codec_ctx_t *ctx);

/*!\brief deadline parameter analogous to  AVx GOOD QUALITY mode. */
#define AOM_DL_GOOD_QUALITY (1000000)
clang-format's avatar
clang-format committed
787 788 789 790 791 792 793 794 795 796 797
/*!\brief Encode a frame
 *
 * Encodes a video frame at the given "presentation time." The presentation
 * time stamp (PTS) \ref MUST be strictly increasing.
 *
 * The encoder supports the notion of a soft real-time deadline. Given a
 * non-zero value to the deadline parameter, the encoder will make a "best
 * effort" guarantee to  return before the given time slice expires. It is
 * implicit that limiting the available time to encode will degrade the
 * output quality. The encoder can be given an unlimited time to produce the
 * best possible frame by specifying a deadline of '0'. This deadline
798
 * supercedes the AVx notion of "best quality, good quality, realtime".
clang-format's avatar
clang-format committed
799
 * Applications that wish to map these former settings to the new deadline
Thomas Daede's avatar
Thomas Daede committed
800
 * based system can use the symbol #AOM_DL_GOOD_QUALITY.
clang-format's avatar
clang-format committed
801 802 803 804
 *
 * When the last frame has been passed to the encoder, this function should
 * continue to be called, with the img parameter set to NULL. This will
 * signal the end-of-stream condition to the encoder and allow it to encode
805 806
 * any held buffers. Encoding is complete when aom_codec_encode() is called
 * and aom_codec_get_cx_data() returns no data.
clang-format's avatar
clang-format committed
807 808 809 810 811 812 813 814
 *
 * \param[in]    ctx       Pointer to this instance's context
 * \param[in]    img       Image data to encode, NULL to flush.
 * \param[in]    pts       Presentation time stamp, in timebase units.
 * \param[in]    duration  Duration to show frame, in timebase units.
 * \param[in]    flags     Flags to use for encoding this frame.
 * \param[in]    deadline  Time to spend encoding, in microseconds. (0=infinite)
 *
815
 * \retval #AOM_CODEC_OK
clang-format's avatar
clang-format committed
816
 *     The configuration was populated.
817
 * \retval #AOM_CODEC_INCAPABLE
clang-format's avatar
clang-format committed
818
 *     Interface is not an encoder interface.
819
 * \retval #AOM_CODEC_INVALID_PARAM
clang-format's avatar
clang-format committed
820 821
 *     A parameter was NULL, the image format is unsupported, etc.
 */
822 823 824
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,
clang-format's avatar
clang-format committed
825 826 827 828 829 830
                                 unsigned long deadline);

/*!\brief Set compressed data output buffer
 *
 * Sets the buffer that the codec should output the compressed data
 * into. This call effectively sets the buffer pointer returned in the
831
 * next AOM_CODEC_CX_FRAME_PKT packet. Subsequent packets will be
clang-format's avatar
clang-format committed
832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857
 * appended into this buffer. The buffer is preserved across frames,
 * so applications must periodically call this function after flushing
 * the accumulated compressed data to disk or to the network to reset
 * the pointer to the buffer's head.
 *
 * `pad_before` bytes will be skipped before writing the compressed
 * data, and `pad_after` bytes will be appended to the packet. The size
 * of the packet will be the sum of the size of the actual compressed
 * data, pad_before, and pad_after. The padding bytes will be preserved
 * (not overwritten).
 *
 * Note that calling this function does not guarantee that the returned
 * compressed data will be placed into the specified buffer. In the
 * event that the encoded data will not fit into the buffer provided,
 * the returned packet \ref MAY point to an internal buffer, as it would
 * if this call were never used. In this event, the output packet will
 * NOT have any padding, and the application must free space and copy it
 * to the proper place. This is of particular note in configurations
 * that may output multiple packets for a single encoded frame (e.g., lagged
 * encoding) or if the application does not reset the buffer periodically.
 *
 * Applications may restore the default behavior of the codec providing
 * the compressed data buffer by calling this function with a NULL
 * buffer.
 *
 * Applications \ref MUSTNOT call this function during iteration of
858
 * aom_codec_get_cx_data().
clang-format's avatar
clang-format committed
859 860 861 862 863 864
 *
 * \param[in]    ctx         Pointer to this instance's context
 * \param[in]    buf         Buffer to store compressed data into
 * \param[in]    pad_before  Bytes to skip before writing compressed data
 * \param[in]    pad_after   Bytes to skip after writing compressed data
 *
865
 * \retval #AOM_CODEC_OK
clang-format's avatar
clang-format committed
866
 *     The buffer was set successfully.
867
 * \retval #AOM_CODEC_INVALID_PARAM
clang-format's avatar
clang-format committed
868 869
 *     A parameter was NULL, the image format is unsupported, etc.
 */
870 871
aom_codec_err_t aom_codec_set_cx_data_buf(aom_codec_ctx_t *ctx,
                                          const aom_fixed_buf_t *buf,
clang-format's avatar
clang-format committed
872 873
                                          unsigned int pad_before,
                                          unsigned int pad_after);
John Koleszar's avatar
John Koleszar committed
874

clang-format's avatar
clang-format committed
875 876 877 878
/*!\brief Encoded data iterator
 *
 * Iterates over a list of data packets to be passed from the encoder to the
 * application. The different kinds of packets available are enumerated in
879
 * #aom_codec_cx_pkt_kind.
clang-format's avatar
clang-format committed
880
 *
881
 * #AOM_CODEC_CX_FRAME_PKT packets should be passed to the application's
clang-format's avatar
clang-format committed
882
 * muxer. Multiple compressed frames may be in the list.
883
 * #AOM_CODEC_STATS_PKT packets should be appended to a global buffer.
clang-format's avatar
clang-format committed
884 885 886 887 888
 *
 * The application \ref MUST silently ignore any packet kinds that it does
 * not recognize or support.
 *
 * The data buffers returned from this function are only guaranteed to be
889
 * valid until the application makes another call to any aom_codec_* function.
clang-format's avatar
clang-format committed
890 891 892 893 894 895 896 897
 *
 * \param[in]     ctx      Pointer to this instance's context
 * \param[in,out] iter     Iterator storage, initialized to NULL
 *
 * \return Returns a pointer to an output data packet (compressed frame data,
 *         two-pass statistics, etc.) or NULL to signal end-of-list.
 *
 */
898 899
const aom_codec_cx_pkt_t *aom_codec_get_cx_data(aom_codec_ctx_t *ctx,
                                                aom_codec_iter_t *iter);
clang-format's avatar
clang-format committed
900 901 902 903 904 905 906 907 908 909 910 911 912

/*!\brief Get Preview Frame
 *
 * Returns an image that can be used as a preview. Shows the image as it would
 * exist at the decompressor. The application \ref MUST NOT write into this
 * image buffer.
 *
 * \param[in]     ctx      Pointer to this instance's context
 *
 * \return Returns a pointer to a preview image, or NULL if no image is
 *         available.
 *
 */
913
const aom_image_t *aom_codec_get_preview_frame(aom_codec_ctx_t *ctx);
John Koleszar's avatar
John Koleszar committed
914

clang-format's avatar
clang-format committed
915
/*!@} - end defgroup encoder*/
John Koleszar's avatar
John Koleszar committed
916 917 918
#ifdef __cplusplus
}
#endif
919
#endif  // AOM_AOM_ENCODER_H_