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 13 /*!\defgroup encoder Encoder Algorithm Interface 14 * \ingroup codec 15 * This abstraction allows applications using this encoder to easily support 16 * multiple video formats with minimal code duplication. This section describes 17 * the interface common to all encoders. 18 * @{ 19 */ 20 21 /*!\file 22 * \brief Describes the encoder algorithm interface to applications. 23 * 24 * This file describes the interface between an application and a 25 * video encoder algorithm. 26 * 27 */ 28 #ifdef __cplusplus 29 extern "C" { 30 #endif 31 32 #include "./vpx_codec.h" // IWYU pragma: export 33 #include "./vpx_ext_ratectrl.h" 34 35 /*! Temporal Scalability: Maximum length of the sequence defining frame 36 * layer membership 37 */ 38 #define VPX_TS_MAX_PERIODICITY 16 39 40 /*! Temporal Scalability: Maximum number of coding layers */ 41 #define VPX_TS_MAX_LAYERS 5 42 43 /*! Temporal+Spatial Scalability: Maximum number of coding layers */ 44 #define VPX_MAX_LAYERS 12 // 3 temporal + 4 spatial layers are allowed. 45 46 /*! Spatial Scalability: Maximum number of coding layers */ 47 #define VPX_SS_MAX_LAYERS 5 48 49 /*! Spatial Scalability: Default number of coding layers */ 50 #define VPX_SS_DEFAULT_LAYERS 1 51 52 /*!\brief Current ABI version number 53 * 54 * \internal 55 * If this file is altered in any way that changes the ABI, this value 56 * must be bumped. Examples include, but are not limited to, changing 57 * types, removing or reassigning enums, adding/removing/rearranging 58 * fields to structures 59 * 60 * \note 61 * VPX_ENCODER_ABI_VERSION has a VPX_EXT_RATECTRL_ABI_VERSION component 62 * because the VP9E_SET_EXTERNAL_RATE_CONTROL codec control uses 63 * vpx_rc_funcs_t. 64 */ 65 #define VPX_ENCODER_ABI_VERSION \ 66 (18 + VPX_CODEC_ABI_VERSION + \ 67 VPX_EXT_RATECTRL_ABI_VERSION) /**<\hideinitializer*/ 68 69 /*! \brief Encoder capabilities bitfield 70 * 71 * Each encoder advertises the capabilities it supports as part of its 72 * ::vpx_codec_iface_t interface structure. Capabilities are extra 73 * interfaces or functionality, and are not required to be supported 74 * by an encoder. 75 * 76 * The available flags are specified by VPX_CODEC_CAP_* defines. 77 */ 78 #define VPX_CODEC_CAP_PSNR 0x10000 /**< Can issue PSNR packets */ 79 80 /*! Can output one partition at a time. Each partition is returned in its 81 * own VPX_CODEC_CX_FRAME_PKT, with the FRAME_IS_FRAGMENT flag set for 82 * every partition but the last. In this mode all frames are always 83 * returned partition by partition. 84 */ 85 #define VPX_CODEC_CAP_OUTPUT_PARTITION 0x20000 86 87 /*! \brief Initialization-time Feature Enabling 88 * 89 * Certain codec features must be known at initialization time, to allow 90 * for proper memory allocation. 91 * 92 * The available flags are specified by VPX_CODEC_USE_* defines. 93 */ 94 #define VPX_CODEC_USE_PSNR 0x10000 /**< Calculate PSNR on each frame */ 95 /*!\brief Make the encoder output one partition at a time. */ 96 #define VPX_CODEC_USE_OUTPUT_PARTITION 0x20000 97 #define VPX_CODEC_USE_HIGHBITDEPTH 0x40000 /**< Use high bitdepth */ 98 99 /*!\brief Generic fixed size buffer structure 100 * 101 * This structure is able to hold a reference to any fixed size buffer. 102 */ 103 typedef struct vpx_fixed_buf { 104 void *buf; /**< Pointer to the data */ 105 size_t sz; /**< Length of the buffer, in chars */ 106 } vpx_fixed_buf_t; /**< alias for struct vpx_fixed_buf */ 107 108 /*!\brief Time Stamp Type 109 * 110 * An integer, which when multiplied by the stream's time base, provides 111 * the absolute time of a sample. 112 */ 113 typedef int64_t vpx_codec_pts_t; 114 115 /*!\brief Compressed Frame Flags 116 * 117 * This type represents a bitfield containing information about a compressed 118 * frame that may be useful to an application. The most significant 16 bits 119 * can be used by an algorithm to provide additional detail, for example to 120 * support frame types that are codec specific (MPEG-1 D-frames for example) 121 */ 122 typedef uint32_t vpx_codec_frame_flags_t; 123 #define VPX_FRAME_IS_KEY 0x1u /**< frame is the start of a GOP */ 124 /*!\brief frame can be dropped without affecting the stream (no future frame 125 * depends on this one) */ 126 #define VPX_FRAME_IS_DROPPABLE 0x2u 127 /*!\brief frame should be decoded but will not be shown */ 128 #define VPX_FRAME_IS_INVISIBLE 0x4u 129 /*!\brief this is a fragment of the encoded frame */ 130 #define VPX_FRAME_IS_FRAGMENT 0x8u 131 132 /*!\brief Error Resilient flags 133 * 134 * These flags define which error resilient features to enable in the 135 * encoder. The flags are specified through the 136 * vpx_codec_enc_cfg::g_error_resilient variable. 137 */ 138 typedef uint32_t vpx_codec_er_flags_t; 139 /*!\brief Improve resiliency against losses of whole frames */ 140 #define VPX_ERROR_RESILIENT_DEFAULT 0x1u 141 /*!\brief The frame partitions are independently decodable by the bool decoder, 142 * meaning that partitions can be decoded even though earlier partitions have 143 * been lost. Note that intra prediction is still done over the partition 144 * boundary. 145 * \note This is only supported by VP8.*/ 146 #define VPX_ERROR_RESILIENT_PARTITIONS 0x2u 147 148 /*!\brief Encoder output packet variants 149 * 150 * This enumeration lists the different kinds of data packets that can be 151 * returned by calls to vpx_codec_get_cx_data(). Algorithms \ref MAY 152 * extend this list to provide additional functionality. 153 */ 154 enum vpx_codec_cx_pkt_kind { 155 VPX_CODEC_CX_FRAME_PKT, /**< Compressed video frame */ 156 VPX_CODEC_STATS_PKT, /**< Two-pass statistics for this frame */ 157 VPX_CODEC_FPMB_STATS_PKT, /**< first pass mb statistics for this frame */ 158 VPX_CODEC_PSNR_PKT, /**< PSNR statistics for this frame */ 159 VPX_CODEC_CUSTOM_PKT = 256 /**< Algorithm extensions */ 160 }; 161 162 /*!\brief Encoder output packet 163 * 164 * This structure contains the different kinds of output data the encoder 165 * may produce while compressing a frame. 166 */ 167 typedef struct vpx_codec_cx_pkt { 168 enum vpx_codec_cx_pkt_kind kind; /**< packet variant */ 169 union { 170 struct { 171 void *buf; /**< compressed data buffer */ 172 size_t sz; /**< length of compressed data */ 173 /*!\brief time stamp to show frame (in timebase units) */ 174 vpx_codec_pts_t pts; 175 /*!\brief duration to show frame (in timebase units) */ 176 unsigned long duration; 177 vpx_codec_frame_flags_t flags; /**< flags for this frame */ 178 /*!\brief the partition id defines the decoding order of the partitions. 179 * Only applicable when "output partition" mode is enabled. First 180 * partition has id 0.*/ 181 int partition_id; 182 /*!\brief Width and height of frames in this packet. VP8 will only use the 183 * first one.*/ 184 unsigned int width[VPX_SS_MAX_LAYERS]; /**< frame width */ 185 unsigned int height[VPX_SS_MAX_LAYERS]; /**< frame height */ 186 /*!\brief Flag to indicate if spatial layer frame in this packet is 187 * encoded or dropped. VP8 will always be set to 1.*/ 188 uint8_t spatial_layer_encoded[VPX_SS_MAX_LAYERS]; 189 } frame; /**< data for compressed frame packet */ 190 vpx_fixed_buf_t twopass_stats; /**< data for two-pass packet */ 191 vpx_fixed_buf_t firstpass_mb_stats; /**< first pass mb packet */ 192 struct vpx_psnr_pkt { 193 unsigned int samples[4]; /**< Number of samples, total/y/u/v */ 194 uint64_t sse[4]; /**< sum squared error, total/y/u/v */ 195 double psnr[4]; /**< PSNR, total/y/u/v */ 196 } psnr; /**< data for PSNR packet */ 197 vpx_fixed_buf_t raw; /**< data for arbitrary packets */ 198 199 /* This packet size is fixed to allow codecs to extend this 200 * interface without having to manage storage for raw packets, 201 * i.e., if it's smaller than 128 bytes, you can store in the 202 * packet list directly. 203 */ 204 char pad[128 - sizeof(enum vpx_codec_cx_pkt_kind)]; /**< fixed sz */ 205 } data; /**< packet data */ 206 } vpx_codec_cx_pkt_t; /**< alias for struct vpx_codec_cx_pkt */ 207 208 /*!\brief Encoder return output buffer callback 209 * 210 * This callback function, when registered, returns with packets when each 211 * spatial layer is encoded. 212 */ 213 typedef void (*vpx_codec_enc_output_cx_pkt_cb_fn_t)(vpx_codec_cx_pkt_t *pkt, 214 void *user_data); 215 216 /*!\brief Callback function pointer / user data pair storage */ 217 typedef struct vpx_codec_enc_output_cx_cb_pair { 218 vpx_codec_enc_output_cx_pkt_cb_fn_t output_cx_pkt; /**< Callback function */ 219 void *user_priv; /**< Pointer to private data */ 220 } vpx_codec_priv_output_cx_pkt_cb_pair_t; 221 222 /*!\brief Rational Number 223 * 224 * This structure holds a fractional value. 225 */ 226 typedef struct vpx_rational { 227 int num; /**< fraction numerator */ 228 int den; /**< fraction denominator */ 229 } vpx_rational_t; /**< alias for struct vpx_rational */ 230 231 /*!\brief Multi-pass Encoding Pass */ 232 typedef enum vpx_enc_pass { 233 VPX_RC_ONE_PASS, /**< Single pass mode */ 234 VPX_RC_FIRST_PASS, /**< First pass of multi-pass mode */ 235 VPX_RC_LAST_PASS /**< Final pass of multi-pass mode */ 236 } vpx_enc_pass; 237 238 /*!\brief Rate control mode */ 239 enum vpx_rc_mode { 240 VPX_VBR, /**< Variable Bit Rate (VBR) mode */ 241 VPX_CBR, /**< Constant Bit Rate (CBR) mode */ 242 VPX_CQ, /**< Constrained Quality (CQ) mode */ 243 VPX_Q, /**< Constant Quality (Q) mode */ 244 }; 245 246 /*!\brief Keyframe placement mode. 247 * 248 * This enumeration determines whether keyframes are placed automatically by 249 * the encoder or whether this behavior is disabled. Older releases of this 250 * SDK were implemented such that VPX_KF_FIXED meant keyframes were disabled. 251 * This name is confusing for this behavior, so the new symbols to be used 252 * are VPX_KF_AUTO and VPX_KF_DISABLED. 253 */ 254 enum vpx_kf_mode { 255 VPX_KF_FIXED, /**< deprecated, implies VPX_KF_DISABLED */ 256 VPX_KF_AUTO, /**< Encoder determines optimal placement automatically */ 257 VPX_KF_DISABLED = 0 /**< Encoder does not place keyframes. */ 258 }; 259 260 /*!\brief Encoded Frame Flags 261 * 262 * This type indicates a bitfield to be passed to vpx_codec_encode(), defining 263 * per-frame boolean values. By convention, bits common to all codecs will be 264 * named VPX_EFLAG_*, and bits specific to an algorithm will be named 265 * /algo/_eflag_*. The lower order 16 bits are reserved for common use. 266 */ 267 typedef long vpx_enc_frame_flags_t; 268 #define VPX_EFLAG_FORCE_KF (1 << 0) /**< Force this frame to be a keyframe */ 269 270 /*!\brief Encoder configuration structure 271 * 272 * This structure contains the encoder settings that have common representations 273 * across all codecs. This doesn't imply that all codecs support all features, 274 * however. 275 */ 276 typedef struct vpx_codec_enc_cfg { 277 /* 278 * generic settings (g) 279 */ 280 281 /*!\brief Deprecated: Algorithm specific "usage" value 282 * 283 * This value must be zero. 284 */ 285 unsigned int g_usage; 286 287 /*!\brief Maximum number of threads to use 288 * 289 * For multi-threaded implementations, use no more than this number of 290 * threads. The codec may use fewer threads than allowed. The value 291 * 0 is equivalent to the value 1. 292 */ 293 unsigned int g_threads; 294 295 /*!\brief Bitstream profile to use 296 * 297 * Some codecs support a notion of multiple bitstream profiles. Typically 298 * this maps to a set of features that are turned on or off. Often the 299 * profile to use is determined by the features of the intended decoder. 300 * Consult the documentation for the codec to determine the valid values 301 * for this parameter, or set to zero for a sane default. 302 */ 303 unsigned int g_profile; /**< profile of bitstream to use */ 304 305 /*!\brief Width of the frame 306 * 307 * This value identifies the presentation resolution of the frame, 308 * in pixels. Note that the frames passed as input to the encoder must 309 * have this resolution. Frames will be presented by the decoder in this 310 * resolution, independent of any spatial resampling the encoder may do. 311 */ 312 unsigned int g_w; 313 314 /*!\brief Height of the frame 315 * 316 * This value identifies the presentation resolution of the frame, 317 * in pixels. Note that the frames passed as input to the encoder must 318 * have this resolution. Frames will be presented by the decoder in this 319 * resolution, independent of any spatial resampling the encoder may do. 320 */ 321 unsigned int g_h; 322 323 /*!\brief Bit-depth of the codec 324 * 325 * This value identifies the bit_depth of the codec, 326 * Only certain bit-depths are supported as identified in the 327 * vpx_bit_depth_t enum. 328 */ 329 vpx_bit_depth_t g_bit_depth; 330 331 /*!\brief Bit-depth of the input frames 332 * 333 * This value identifies the bit_depth of the input frames in bits. 334 * Note that the frames passed as input to the encoder must have 335 * this bit-depth. 336 */ 337 unsigned int g_input_bit_depth; 338 339 /*!\brief Stream timebase units 340 * 341 * Indicates the smallest interval of time, in seconds, used by the stream. 342 * For fixed frame rate material, or variable frame rate material where 343 * frames are timed at a multiple of a given clock (ex: video capture), 344 * the \ref RECOMMENDED method is to set the timebase to the reciprocal 345 * of the frame rate (ex: 1001/30000 for 29.970 Hz NTSC). This allows the 346 * pts to correspond to the frame number, which can be handy. For 347 * re-encoding video from containers with absolute time timestamps, the 348 * \ref RECOMMENDED method is to set the timebase to that of the parent 349 * container or multimedia framework (ex: 1/1000 for ms, as in FLV). 350 */ 351 struct vpx_rational g_timebase; 352 353 /*!\brief Enable error resilient modes. 354 * 355 * The error resilient bitfield indicates to the encoder which features 356 * it should enable to take measures for streaming over lossy or noisy 357 * links. 358 */ 359 vpx_codec_er_flags_t g_error_resilient; 360 361 /*!\brief Multi-pass Encoding Mode 362 * 363 * This value should be set to the current phase for multi-pass encoding. 364 * For single pass, set to #VPX_RC_ONE_PASS. 365 */ 366 enum vpx_enc_pass g_pass; 367 368 /*!\brief Allow lagged encoding 369 * 370 * If set, this value allows the encoder to consume a number of input 371 * frames before producing output frames. This allows the encoder to 372 * base decisions for the current frame on future frames. This does 373 * increase the latency of the encoding pipeline, so it is not appropriate 374 * in all situations (ex: realtime encoding). 375 * 376 * Note that this is a maximum value -- the encoder may produce frames 377 * sooner than the given limit. Set this value to 0 to disable this 378 * feature. 379 */ 380 unsigned int g_lag_in_frames; 381 382 /* 383 * rate control settings (rc) 384 */ 385 386 /*!\brief Temporal resampling configuration, if supported by the codec. 387 * 388 * Temporal resampling allows the codec to "drop" frames as a strategy to 389 * meet its target data rate. This can cause temporal discontinuities in 390 * the encoded video, which may appear as stuttering during playback. This 391 * trade-off is often acceptable, but for many applications is not. It can 392 * be disabled in these cases. 393 * 394 * This threshold is described as a percentage of the target data buffer. 395 * When the data buffer falls below this percentage of fullness, a 396 * dropped frame is indicated. Set the threshold to zero (0) to disable 397 * this feature. 398 */ 399 unsigned int rc_dropframe_thresh; 400 401 /*!\brief Enable/disable spatial resampling, if supported by the codec. 402 * 403 * Spatial resampling allows the codec to compress a lower resolution 404 * version of the frame, which is then upscaled by the encoder to the 405 * correct presentation resolution. This increases visual quality at 406 * low data rates, at the expense of CPU time on the encoder/decoder. 407 */ 408 unsigned int rc_resize_allowed; 409 410 /*!\brief Internal coded frame width. 411 * 412 * If spatial resampling is enabled this specifies the width of the 413 * encoded frame. 414 */ 415 unsigned int rc_scaled_width; 416 417 /*!\brief Internal coded frame height. 418 * 419 * If spatial resampling is enabled this specifies the height of the 420 * encoded frame. 421 */ 422 unsigned int rc_scaled_height; 423 424 /*!\brief Spatial resampling up watermark. 425 * 426 * This threshold is described as a percentage of the target data buffer. 427 * When the data buffer rises above this percentage of fullness, the 428 * encoder will step up to a higher resolution version of the frame. 429 */ 430 unsigned int rc_resize_up_thresh; 431 432 /*!\brief Spatial resampling down watermark. 433 * 434 * This threshold is described as a percentage of the target data buffer. 435 * When the data buffer falls below this percentage of fullness, the 436 * encoder will step down to a lower resolution version of the frame. 437 */ 438 unsigned int rc_resize_down_thresh; 439 440 /*!\brief Rate control algorithm to use. 441 * 442 * Indicates whether the end usage of this stream is to be streamed over 443 * a bandwidth constrained link, indicating that Constant Bit Rate (CBR) 444 * mode should be used, or whether it will be played back on a high 445 * bandwidth link, as from a local disk, where higher variations in 446 * bitrate are acceptable. 447 */ 448 enum vpx_rc_mode rc_end_usage; 449 450 /*!\brief Two-pass stats buffer. 451 * 452 * A buffer containing all of the stats packets produced in the first 453 * pass, concatenated. 454 */ 455 vpx_fixed_buf_t rc_twopass_stats_in; 456 457 /*!\brief first pass mb stats buffer. 458 * 459 * A buffer containing all of the first pass mb stats packets produced 460 * in the first pass, concatenated. 461 */ 462 vpx_fixed_buf_t rc_firstpass_mb_stats_in; 463 464 /*!\brief Target data rate 465 * 466 * Target bitrate to use for this stream, in kilobits per second. 467 * Internally capped to the smaller of the uncompressed bitrate and 468 * 1000000 kilobits per second. 469 */ 470 unsigned int rc_target_bitrate; 471 472 /* 473 * quantizer settings 474 */ 475 476 /*!\brief Minimum (Best Quality) Quantizer 477 * 478 * The quantizer is the most direct control over the quality of the 479 * encoded image. The range of valid values for the quantizer is codec 480 * specific. Consult the documentation for the codec to determine the 481 * values to use. 482 */ 483 unsigned int rc_min_quantizer; 484 485 /*!\brief Maximum (Worst Quality) Quantizer 486 * 487 * The quantizer is the most direct control over the quality of the 488 * encoded image. The range of valid values for the quantizer is codec 489 * specific. Consult the documentation for the codec to determine the 490 * values to use. 491 */ 492 unsigned int rc_max_quantizer; 493 494 /* 495 * bitrate tolerance 496 */ 497 498 /*!\brief Rate control adaptation undershoot control 499 * 500 * VP8: Expressed as a percentage of the target bitrate, 501 * controls the maximum allowed adaptation speed of the codec. 502 * This factor controls the maximum amount of bits that can 503 * be subtracted from the target bitrate in order to compensate 504 * for prior overshoot. 505 * VP9: Expressed as a percentage of the target bitrate, a threshold 506 * undershoot level (current rate vs target) beyond which more aggressive 507 * corrective measures are taken. 508 * * 509 * Valid values in the range VP8:0-100 VP9: 0-100. 510 */ 511 unsigned int rc_undershoot_pct; 512 513 /*!\brief Rate control adaptation overshoot control 514 * 515 * VP8: Expressed as a percentage of the target bitrate, 516 * controls the maximum allowed adaptation speed of the codec. 517 * This factor controls the maximum amount of bits that can 518 * be added to the target bitrate in order to compensate for 519 * prior undershoot. 520 * VP9: Expressed as a percentage of the target bitrate, a threshold 521 * overshoot level (current rate vs target) beyond which more aggressive 522 * corrective measures are taken. 523 * 524 * Valid values in the range VP8:0-100 VP9: 0-100. 525 */ 526 unsigned int rc_overshoot_pct; 527 528 /* 529 * decoder buffer model parameters 530 */ 531 532 /*!\brief Decoder Buffer Size 533 * 534 * This value indicates the amount of data that may be buffered by the 535 * decoding application. Note that this value is expressed in units of 536 * time (milliseconds). For example, a value of 5000 indicates that the 537 * client will buffer (at least) 5000ms worth of encoded data. Use the 538 * target bitrate (#rc_target_bitrate) to convert to bits/bytes, if 539 * necessary. 540 */ 541 unsigned int rc_buf_sz; 542 543 /*!\brief Decoder Buffer Initial Size 544 * 545 * This value indicates the amount of data that will be buffered by the 546 * decoding application prior to beginning playback. This value is 547 * expressed in units of time (milliseconds). Use the target bitrate 548 * (#rc_target_bitrate) to convert to bits/bytes, if necessary. 549 */ 550 unsigned int rc_buf_initial_sz; 551 552 /*!\brief Decoder Buffer Optimal Size 553 * 554 * This value indicates the amount of data that the encoder should try 555 * to maintain in the decoder's buffer. This value is expressed in units 556 * of time (milliseconds). Use the target bitrate (#rc_target_bitrate) 557 * to convert to bits/bytes, if necessary. 558 */ 559 unsigned int rc_buf_optimal_sz; 560 561 /* 562 * 2 pass rate control parameters 563 */ 564 565 /*!\brief Two-pass mode CBR/VBR bias 566 * 567 * Bias, expressed on a scale of 0 to 100, for determining target size 568 * for the current frame. The value 0 indicates the optimal CBR mode 569 * value should be used. The value 100 indicates the optimal VBR mode 570 * value should be used. Values in between indicate which way the 571 * encoder should "lean." 572 */ 573 unsigned int rc_2pass_vbr_bias_pct; 574 575 /*!\brief Two-pass mode per-GOP minimum bitrate 576 * 577 * This value, expressed as a percentage of the target bitrate, indicates 578 * the minimum bitrate to be used for a single GOP (aka "section") 579 */ 580 unsigned int rc_2pass_vbr_minsection_pct; 581 582 /*!\brief Two-pass mode per-GOP maximum bitrate 583 * 584 * This value, expressed as a percentage of the target bitrate, indicates 585 * the maximum bitrate to be used for a single GOP (aka "section") 586 */ 587 unsigned int rc_2pass_vbr_maxsection_pct; 588 589 /*!\brief Two-pass corpus vbr mode complexity control 590 * Used only in VP9: A value representing the corpus midpoint complexity 591 * for corpus vbr mode. This value defaults to 0 which disables corpus vbr 592 * mode in favour of normal vbr mode. 593 */ 594 unsigned int rc_2pass_vbr_corpus_complexity; 595 596 /* 597 * keyframing settings (kf) 598 */ 599 600 /*!\brief Keyframe placement mode 601 * 602 * This value indicates whether the encoder should place keyframes at a 603 * fixed interval, or determine the optimal placement automatically 604 * (as governed by the #kf_min_dist and #kf_max_dist parameters) 605 */ 606 enum vpx_kf_mode kf_mode; 607 608 /*!\brief Keyframe minimum interval 609 * 610 * This value, expressed as a number of frames, prevents the encoder from 611 * placing a keyframe nearer than kf_min_dist to the previous keyframe. At 612 * least kf_min_dist frames non-keyframes will be coded before the next 613 * keyframe. Set kf_min_dist equal to kf_max_dist for a fixed interval. 614 */ 615 unsigned int kf_min_dist; 616 617 /*!\brief Keyframe maximum interval 618 * 619 * This value, expressed as a number of frames, forces the encoder to code 620 * a keyframe if one has not been coded in the last kf_max_dist frames. 621 * A value of 0 implies all frames will be keyframes. Set kf_min_dist 622 * equal to kf_max_dist for a fixed interval. 623 */ 624 unsigned int kf_max_dist; 625 626 /* 627 * Spatial scalability settings (ss) 628 */ 629 630 /*!\brief Number of spatial coding layers. 631 * 632 * This value specifies the number of spatial coding layers to be used. 633 */ 634 unsigned int ss_number_layers; 635 636 /*!\brief Enable auto alt reference flags for each spatial layer. 637 * 638 * These values specify if auto alt reference frame is enabled for each 639 * spatial layer. 640 */ 641 int ss_enable_auto_alt_ref[VPX_SS_MAX_LAYERS]; 642 643 /*!\brief Target bitrate for each spatial layer. 644 * 645 * These values specify the target coding bitrate to be used for each 646 * spatial layer. (in kbps) 647 */ 648 unsigned int ss_target_bitrate[VPX_SS_MAX_LAYERS]; 649 650 /*!\brief Number of temporal coding layers. 651 * 652 * This value specifies the number of temporal layers to be used. 653 */ 654 unsigned int ts_number_layers; 655 656 /*!\brief Target bitrate for each temporal layer. 657 * 658 * These values specify the target coding bitrate to be used for each 659 * temporal layer. (in kbps) 660 */ 661 unsigned int ts_target_bitrate[VPX_TS_MAX_LAYERS]; 662 663 /*!\brief Frame rate decimation factor for each temporal layer. 664 * 665 * These values specify the frame rate decimation factors to apply 666 * to each temporal layer. 667 */ 668 unsigned int ts_rate_decimator[VPX_TS_MAX_LAYERS]; 669 670 /*!\brief Length of the sequence defining frame temporal layer membership. 671 * 672 * This value specifies the length of the sequence that defines the 673 * membership of frames to temporal layers. For example, if the 674 * ts_periodicity = 8, then the frames are assigned to coding layers with a 675 * repeated sequence of length 8. 676 */ 677 unsigned int ts_periodicity; 678 679 /*!\brief Template defining the membership of frames to temporal layers. 680 * 681 * This array defines the membership of frames to temporal coding layers. 682 * For a 2-layer encoding that assigns even numbered frames to one temporal 683 * layer (0) and odd numbered frames to a second temporal layer (1) with 684 * ts_periodicity=8, then ts_layer_id = (0,1,0,1,0,1,0,1). 685 */ 686 unsigned int ts_layer_id[VPX_TS_MAX_PERIODICITY]; 687 688 /*!\brief Target bitrate for each spatial/temporal layer. 689 * 690 * These values specify the target coding bitrate to be used for each 691 * spatial/temporal layer. (in kbps) 692 * 693 */ 694 unsigned int layer_target_bitrate[VPX_MAX_LAYERS]; 695 696 /*!\brief Temporal layering mode indicating which temporal layering scheme to 697 * use. 698 * 699 * The value (refer to VP9E_TEMPORAL_LAYERING_MODE) specifies the 700 * temporal layering mode to use. 701 * 702 */ 703 int temporal_layering_mode; 704 705 /*!\brief A flag indicating whether to use external rate control parameters. 706 * By default is 0. If set to 1, the following parameters will be used in the 707 * rate control system. 708 */ 709 int use_vizier_rc_params; 710 711 /*!\brief Active worst quality factor. 712 * 713 * Rate control parameters, set from external experiment results. 714 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 715 * used. Otherwise, the default value is used. 716 * 717 */ 718 vpx_rational_t active_wq_factor; 719 720 /*!\brief Error per macroblock adjustment factor. 721 * 722 * Rate control parameters, set from external experiment results. 723 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 724 * used. Otherwise, the default value is used. 725 * 726 */ 727 vpx_rational_t err_per_mb_factor; 728 729 /*!\brief Second reference default decay limit. 730 * 731 * Rate control parameters, set from external experiment results. 732 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 733 * used. Otherwise, the default value is used. 734 * 735 */ 736 vpx_rational_t sr_default_decay_limit; 737 738 /*!\brief Second reference difference factor. 739 * 740 * Rate control parameters, set from external experiment results. 741 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 742 * used. Otherwise, the default value is used. 743 * 744 */ 745 vpx_rational_t sr_diff_factor; 746 747 /*!\brief Keyframe error per macroblock adjustment factor. 748 * 749 * Rate control parameters, set from external experiment results. 750 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 751 * used. Otherwise, the default value is used. 752 * 753 */ 754 vpx_rational_t kf_err_per_mb_factor; 755 756 /*!\brief Keyframe minimum boost adjustment factor. 757 * 758 * Rate control parameters, set from external experiment results. 759 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 760 * used. Otherwise, the default value is used. 761 * 762 */ 763 vpx_rational_t kf_frame_min_boost_factor; 764 765 /*!\brief Keyframe maximum boost adjustment factor, for the first keyframe 766 * in a chunk. 767 * 768 * Rate control parameters, set from external experiment results. 769 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 770 * used. Otherwise, the default value is used. 771 * 772 */ 773 vpx_rational_t kf_frame_max_boost_first_factor; 774 775 /*!\brief Keyframe maximum boost adjustment factor, for subsequent keyframes. 776 * 777 * Rate control parameters, set from external experiment results. 778 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 779 * used. Otherwise, the default value is used. 780 * 781 */ 782 vpx_rational_t kf_frame_max_boost_subs_factor; 783 784 /*!\brief Keyframe maximum total boost adjustment factor. 785 * 786 * Rate control parameters, set from external experiment results. 787 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 788 * used. Otherwise, the default value is used. 789 * 790 */ 791 vpx_rational_t kf_max_total_boost_factor; 792 793 /*!\brief Golden frame maximum total boost adjustment factor. 794 * 795 * Rate control parameters, set from external experiment results. 796 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 797 * used. Otherwise, the default value is used. 798 * 799 */ 800 vpx_rational_t gf_max_total_boost_factor; 801 802 /*!\brief Golden frame maximum boost adjustment factor. 803 * 804 * Rate control parameters, set from external experiment results. 805 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 806 * used. Otherwise, the default value is used. 807 * 808 */ 809 vpx_rational_t gf_frame_max_boost_factor; 810 811 /*!\brief Zero motion power factor. 812 * 813 * Rate control parameters, set from external experiment results. 814 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 815 * used. Otherwise, the default value is used. 816 * 817 */ 818 vpx_rational_t zm_factor; 819 820 /*!\brief Rate-distortion multiplier for inter frames. 821 * The multiplier is a crucial parameter in the calculation of rate distortion 822 * cost. It is often related to the qp (qindex) value. 823 * Rate control parameters, could be set from external experiment results. 824 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 825 * used. Otherwise, the default value is used. 826 * 827 */ 828 vpx_rational_t rd_mult_inter_qp_fac; 829 830 /*!\brief Rate-distortion multiplier for alt-ref frames. 831 * The multiplier is a crucial parameter in the calculation of rate distortion 832 * cost. It is often related to the qp (qindex) value. 833 * Rate control parameters, could be set from external experiment results. 834 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 835 * used. Otherwise, the default value is used. 836 * 837 */ 838 vpx_rational_t rd_mult_arf_qp_fac; 839 840 /*!\brief Rate-distortion multiplier for key frames. 841 * The multiplier is a crucial parameter in the calculation of rate distortion 842 * cost. It is often related to the qp (qindex) value. 843 * Rate control parameters, could be set from external experiment results. 844 * Only when |use_vizier_rc_params| is set to 1, the pass in value will be 845 * used. Otherwise, the default value is used. 846 * 847 */ 848 vpx_rational_t rd_mult_key_qp_fac; 849 } vpx_codec_enc_cfg_t; /**< alias for struct vpx_codec_enc_cfg */ 850 851 /*!\brief vp9 svc extra configure parameters 852 * 853 * This defines max/min quantizers and scale factors for each layer 854 * 855 */ 856 typedef struct vpx_svc_parameters { 857 int max_quantizers[VPX_MAX_LAYERS]; /**< Max Q for each layer */ 858 int min_quantizers[VPX_MAX_LAYERS]; /**< Min Q for each layer */ 859 int scaling_factor_num[VPX_MAX_LAYERS]; /**< Scaling factor-numerator */ 860 int scaling_factor_den[VPX_MAX_LAYERS]; /**< Scaling factor-denominator */ 861 int speed_per_layer[VPX_MAX_LAYERS]; /**< Speed setting for each sl */ 862 int temporal_layering_mode; /**< Temporal layering mode */ 863 int loopfilter_ctrl[VPX_MAX_LAYERS]; /**< Loopfilter ctrl for each sl */ 864 } vpx_svc_extra_cfg_t; 865 866 /*!\brief Initialize an encoder instance 867 * 868 * Initializes an encoder context using the given interface. Applications 869 * should call the vpx_codec_enc_init convenience macro instead of this 870 * function directly, to ensure that the ABI version number parameter 871 * is properly initialized. 872 * 873 * If the library was configured with --disable-multithread, this call 874 * is not thread safe and should be guarded with a lock if being used 875 * in a multithreaded context. 876 * 877 * If vpx_codec_enc_init_ver() fails, it is not necessary to call 878 * vpx_codec_destroy() on the encoder context. 879 * 880 * \param[in] ctx Pointer to this instance's context. 881 * \param[in] iface Pointer to the algorithm interface to use. 882 * \param[in] cfg Configuration to use, if known. May be NULL. 883 * \param[in] flags Bitfield of VPX_CODEC_USE_* flags 884 * \param[in] ver ABI version number. Must be set to 885 * VPX_ENCODER_ABI_VERSION 886 * \retval #VPX_CODEC_OK 887 * The decoder algorithm initialized. 888 * \retval #VPX_CODEC_MEM_ERROR 889 * Memory allocation failed. 890 */ 891 vpx_codec_err_t vpx_codec_enc_init_ver(vpx_codec_ctx_t *ctx, 892 vpx_codec_iface_t *iface, 893 const vpx_codec_enc_cfg_t *cfg, 894 vpx_codec_flags_t flags, int ver); 895 896 /*!\brief Convenience macro for vpx_codec_enc_init_ver() 897 * 898 * Ensures the ABI version parameter is properly set. 899 */ 900 #define vpx_codec_enc_init(ctx, iface, cfg, flags) \ 901 vpx_codec_enc_init_ver(ctx, iface, cfg, flags, VPX_ENCODER_ABI_VERSION) 902 903 /*!\brief Initialize multi-encoder instance 904 * 905 * Initializes multi-encoder context using the given interface. 906 * Applications should call the vpx_codec_enc_init_multi convenience macro 907 * instead of this function directly, to ensure that the ABI version number 908 * parameter is properly initialized. 909 * 910 * \param[in] ctx Pointer to this instance's context. 911 * \param[in] iface Pointer to the algorithm interface to use. 912 * \param[in] cfg Configuration to use, if known. May be NULL. 913 * \param[in] num_enc Total number of encoders. 914 * \param[in] flags Bitfield of VPX_CODEC_USE_* flags 915 * \param[in] dsf Pointer to down-sampling factors. 916 * \param[in] ver ABI version number. Must be set to 917 * VPX_ENCODER_ABI_VERSION 918 * \retval #VPX_CODEC_OK 919 * The encoder algorithm has been initialized. 920 * \retval #VPX_CODEC_MEM_ERROR 921 * Memory allocation failed. 922 */ 923 vpx_codec_err_t vpx_codec_enc_init_multi_ver( 924 vpx_codec_ctx_t *ctx, vpx_codec_iface_t *iface, vpx_codec_enc_cfg_t *cfg, 925 int num_enc, vpx_codec_flags_t flags, vpx_rational_t *dsf, int ver); 926 927 /*!\brief Convenience macro for vpx_codec_enc_init_multi_ver() 928 * 929 * Ensures the ABI version parameter is properly set. 930 */ 931 #define vpx_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf) \ 932 vpx_codec_enc_init_multi_ver(ctx, iface, cfg, num_enc, flags, dsf, \ 933 VPX_ENCODER_ABI_VERSION) 934 935 /*!\brief Get a default configuration 936 * 937 * Initializes a encoder configuration structure with default values. Supports 938 * the notion of "usages" so that an algorithm may offer different default 939 * settings depending on the user's intended goal. This function \ref SHOULD 940 * be called by all applications to initialize the configuration structure 941 * before specializing the configuration with application specific values. 942 * 943 * \param[in] iface Pointer to the algorithm interface to use. 944 * \param[out] cfg Configuration buffer to populate. 945 * \param[in] usage Must be set to 0. 946 * 947 * \retval #VPX_CODEC_OK 948 * The configuration was populated. 949 * \retval #VPX_CODEC_INCAPABLE 950 * Interface is not an encoder interface. 951 * \retval #VPX_CODEC_INVALID_PARAM 952 * A parameter was NULL, or the usage value was not recognized. 953 */ 954 vpx_codec_err_t vpx_codec_enc_config_default(vpx_codec_iface_t *iface, 955 vpx_codec_enc_cfg_t *cfg, 956 unsigned int usage); 957 958 /*!\brief Set or change configuration 959 * 960 * Reconfigures an encoder instance according to the given configuration. 961 * 962 * \param[in] ctx Pointer to this instance's context 963 * \param[in] cfg Configuration buffer to use 964 * 965 * \retval #VPX_CODEC_OK 966 * The configuration was populated. 967 * \retval #VPX_CODEC_INCAPABLE 968 * Interface is not an encoder interface. 969 * \retval #VPX_CODEC_INVALID_PARAM 970 * A parameter was NULL, or the usage value was not recognized. 971 */ 972 vpx_codec_err_t vpx_codec_enc_config_set(vpx_codec_ctx_t *ctx, 973 const vpx_codec_enc_cfg_t *cfg); 974 975 /*!\brief Get global stream headers 976 * 977 * Retrieves a stream level global header packet, if supported by the codec. 978 * 979 * \li VP8: Unsupported 980 * \li VP9: Returns a buffer of <tt>ID (1 byte)|Length (1 byte)|Length 981 * bytes</tt> values. The function should be called after encoding to retrieve 982 * the most accurate information. 983 * 984 * \param[in] ctx Pointer to this instance's context 985 * 986 * \retval NULL 987 * Encoder does not support global header 988 * \retval Non-NULL 989 * Pointer to buffer containing global header packet. The buffer pointer 990 * and its contents are only valid for the lifetime of \a ctx. The contents 991 * may change in subsequent calls to the function. 992 * \sa 993 * https://www.webmproject.org/docs/container/#vp9-codec-feature-metadata-codecprivate 994 */ 995 vpx_fixed_buf_t *vpx_codec_get_global_headers(vpx_codec_ctx_t *ctx); 996 997 /*!\brief Encode Deadline 998 * 999 * This type indicates a deadline, in microseconds, to be passed to 1000 * vpx_codec_encode(). 1001 */ 1002 typedef unsigned long vpx_enc_deadline_t; 1003 /*!\brief deadline parameter analogous to VPx REALTIME mode. */ 1004 #define VPX_DL_REALTIME 1ul 1005 /*!\brief deadline parameter analogous to VPx GOOD QUALITY mode. */ 1006 #define VPX_DL_GOOD_QUALITY 1000000ul 1007 /*!\brief deadline parameter analogous to VPx BEST QUALITY mode. */ 1008 #define VPX_DL_BEST_QUALITY 0ul 1009 /*!\brief Encode a frame 1010 * 1011 * Encodes a video frame at the given "presentation time." The presentation 1012 * time stamp (PTS) \ref MUST be strictly increasing. 1013 * 1014 * The encoder supports the notion of a soft real-time deadline. Given a 1015 * non-zero value to the deadline parameter, the encoder will make a "best 1016 * effort" guarantee to return before the given time slice expires. It is 1017 * implicit that limiting the available time to encode will degrade the 1018 * output quality. The encoder can be given an unlimited time to produce the 1019 * best possible frame by specifying a deadline of '0'. This deadline 1020 * supersedes the VPx notion of "best quality, good quality, realtime". 1021 * Applications that wish to map these former settings to the new deadline 1022 * based system can use the symbols #VPX_DL_REALTIME, #VPX_DL_GOOD_QUALITY, 1023 * and #VPX_DL_BEST_QUALITY. 1024 * 1025 * When the last frame has been passed to the encoder, this function should 1026 * continue to be called, with the img parameter set to NULL. This will 1027 * signal the end-of-stream condition to the encoder and allow it to encode 1028 * any held buffers. Encoding is complete when vpx_codec_encode() is called 1029 * and vpx_codec_get_cx_data() returns no data. 1030 * 1031 * \param[in] ctx Pointer to this instance's context 1032 * \param[in] img Image data to encode, NULL to flush. 1033 * Encoding sample values outside the range 1034 * [0..(1<<img->bit_depth)-1] is undefined behavior. 1035 * \param[in] pts Presentation time stamp, in timebase units. 1036 * \param[in] duration Duration to show frame, in timebase units. 1037 * \param[in] flags Flags to use for encoding this frame. 1038 * \param[in] deadline Time to spend encoding, in microseconds. (0=infinite) 1039 * 1040 * \retval #VPX_CODEC_OK 1041 * The configuration was populated. 1042 * \retval #VPX_CODEC_INCAPABLE 1043 * Interface is not an encoder interface. 1044 * \retval #VPX_CODEC_INVALID_PARAM 1045 * A parameter was NULL, the image format is unsupported, etc. 1046 */ 1047 vpx_codec_err_t vpx_codec_encode(vpx_codec_ctx_t *ctx, const vpx_image_t *img, 1048 vpx_codec_pts_t pts, unsigned long duration, 1049 vpx_enc_frame_flags_t flags, 1050 vpx_enc_deadline_t deadline); 1051 1052 /*!\brief Set compressed data output buffer 1053 * 1054 * Sets the buffer that the codec should output the compressed data 1055 * into. This call effectively sets the buffer pointer returned in the 1056 * next VPX_CODEC_CX_FRAME_PKT packet. Subsequent packets will be 1057 * appended into this buffer. The buffer is preserved across frames, 1058 * so applications must periodically call this function after flushing 1059 * the accumulated compressed data to disk or to the network to reset 1060 * the pointer to the buffer's head. 1061 * 1062 * `pad_before` bytes will be skipped before writing the compressed 1063 * data, and `pad_after` bytes will be appended to the packet. The size 1064 * of the packet will be the sum of the size of the actual compressed 1065 * data, pad_before, and pad_after. The padding bytes will be preserved 1066 * (not overwritten). 1067 * 1068 * Note that calling this function does not guarantee that the returned 1069 * compressed data will be placed into the specified buffer. In the 1070 * event that the encoded data will not fit into the buffer provided, 1071 * the returned packet \ref MAY point to an internal buffer, as it would 1072 * if this call were never used. In this event, the output packet will 1073 * NOT have any padding, and the application must free space and copy it 1074 * to the proper place. This is of particular note in configurations 1075 * that may output multiple packets for a single encoded frame (e.g., lagged 1076 * encoding) or if the application does not reset the buffer periodically. 1077 * 1078 * Applications may restore the default behavior of the codec providing 1079 * the compressed data buffer by calling this function with a NULL 1080 * buffer. 1081 * 1082 * Applications \ref MUSTNOT call this function during iteration of 1083 * vpx_codec_get_cx_data(). 1084 * 1085 * \param[in] ctx Pointer to this instance's context 1086 * \param[in] buf Buffer to store compressed data into 1087 * \param[in] pad_before Bytes to skip before writing compressed data 1088 * \param[in] pad_after Bytes to skip after writing compressed data 1089 * 1090 * \retval #VPX_CODEC_OK 1091 * The buffer was set successfully. 1092 * \retval #VPX_CODEC_INVALID_PARAM 1093 * A parameter was NULL, the image format is unsupported, etc. 1094 * 1095 * \note 1096 * `duration` and `deadline` are of the unsigned long type, which can be 32 1097 * or 64 bits. `duration` and `deadline` must be less than or equal to 1098 * UINT32_MAX so that their ranges are independent of the size of unsigned 1099 * long. 1100 */ 1101 vpx_codec_err_t vpx_codec_set_cx_data_buf(vpx_codec_ctx_t *ctx, 1102 const vpx_fixed_buf_t *buf, 1103 unsigned int pad_before, 1104 unsigned int pad_after); 1105 1106 /*!\brief Encoded data iterator 1107 * 1108 * Iterates over a list of data packets to be passed from the encoder to the 1109 * application. The different kinds of packets available are enumerated in 1110 * #vpx_codec_cx_pkt_kind. 1111 * 1112 * #VPX_CODEC_CX_FRAME_PKT packets should be passed to the application's 1113 * muxer. Multiple compressed frames may be in the list. 1114 * #VPX_CODEC_STATS_PKT packets should be appended to a global buffer. 1115 * 1116 * The application \ref MUST silently ignore any packet kinds that it does 1117 * not recognize or support. 1118 * 1119 * The data buffers returned from this function are only guaranteed to be 1120 * valid until the application makes another call to any vpx_codec_* function. 1121 * 1122 * \param[in] ctx Pointer to this instance's context 1123 * \param[in,out] iter Iterator storage, initialized to NULL 1124 * 1125 * \return Returns a pointer to an output data packet (compressed frame data, 1126 * two-pass statistics, etc.) or NULL to signal end-of-list. 1127 * 1128 */ 1129 const vpx_codec_cx_pkt_t *vpx_codec_get_cx_data(vpx_codec_ctx_t *ctx, 1130 vpx_codec_iter_t *iter); 1131 1132 /*!\brief Get Preview Frame 1133 * 1134 * Returns an image that can be used as a preview. Shows the image as it would 1135 * exist at the decompressor. The application \ref MUST NOT write into this 1136 * image buffer. 1137 * 1138 * \param[in] ctx Pointer to this instance's context 1139 * 1140 * \return Returns a pointer to a preview image, or NULL if no image is 1141 * available. 1142 * 1143 */ 1144 const vpx_image_t *vpx_codec_get_preview_frame(vpx_codec_ctx_t *ctx); 1145 1146 /*!@} - end defgroup encoder*/ 1147 #ifdef __cplusplus 1148 } 1149 #endif 1150 #endif // VPX_VPX_VPX_ENCODER_H_ 1151