1 /* 2 * Copyright (c) 2016, Alliance for Open Media. All rights reserved 3 * 4 * This source code is subject to the terms of the BSD 2 Clause License and 5 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License 6 * was not distributed with this source code in the LICENSE file, you can 7 * obtain it at www.aomedia.org/license/software. If the Alliance for Open 8 * Media Patent License 1.0 was not distributed with this source code in the 9 * PATENTS file, you can obtain it at www.aomedia.org/license/patent. 10 */ 11 #ifndef AOM_AOM_AOM_ENCODER_H_ 12 #define AOM_AOM_AOM_ENCODER_H_ 13 14 /*!\defgroup encoder Encoder Algorithm Interface 15 * \ingroup codec 16 * This abstraction allows applications using this encoder to easily support 17 * multiple video formats with minimal code duplication. This section describes 18 * the interface common to all encoders. 19 * @{ 20 */ 21 22 /*!\file 23 * \brief Describes the encoder algorithm interface to applications. 24 * 25 * This file describes the interface between an application and a 26 * video encoder algorithm. 27 * 28 */ 29 #ifdef __cplusplus 30 extern "C" { 31 #endif 32 33 #include "aom/aom_codec.h" 34 #include "aom/aom_external_partition.h" 35 36 /*!\brief Current ABI version number 37 * 38 * \hideinitializer 39 * \internal 40 * If this file is altered in any way that changes the ABI, this value 41 * must be bumped. Examples include, but are not limited to, changing 42 * types, removing or reassigning enums, adding/removing/rearranging 43 * fields to structures 44 * 45 * Note: In the definition of AOM_ENCODER_ABI_VERSION, 3 is the value of 46 * AOM_EXT_PART_ABI_VERSION in libaom v3.2.0. The old value of 47 * AOM_EXT_PART_ABI_VERSION is used so as to not break the ABI version check in 48 * aom_codec_enc_init_ver() when an application compiled against libaom v3.2.0 49 * passes the old value of AOM_ENCODER_ABI_VERSION to aom_codec_enc_init_ver(). 50 * The external partition API is still experimental. When it is declared stable, 51 * we will replace 3 with AOM_EXT_PART_ABI_VERSION in the definition of 52 * AOM_ENCODER_ABI_VERSION. 53 */ 54 #define AOM_ENCODER_ABI_VERSION \ 55 (10 + AOM_CODEC_ABI_VERSION + /*AOM_EXT_PART_ABI_VERSION=*/3) 56 57 /*! \brief Encoder capabilities bitfield 58 * 59 * Each encoder advertises the capabilities it supports as part of its 60 * ::aom_codec_iface_t interface structure. Capabilities are extra 61 * interfaces or functionality, and are not required to be supported 62 * by an encoder. 63 * 64 * The available flags are specified by AOM_CODEC_CAP_* defines. 65 */ 66 #define AOM_CODEC_CAP_PSNR 0x10000 /**< Can issue PSNR packets */ 67 68 /*! Can support input images at greater than 8 bitdepth. 69 */ 70 #define AOM_CODEC_CAP_HIGHBITDEPTH 0x40000 71 72 /*! \brief Initialization-time Feature Enabling 73 * 74 * Certain codec features must be known at initialization time, to allow 75 * for proper memory allocation. 76 * 77 * The available flags are specified by AOM_CODEC_USE_* defines. 78 */ 79 #define AOM_CODEC_USE_PSNR 0x10000 /**< Calculate PSNR on each frame */ 80 #define AOM_CODEC_USE_HIGHBITDEPTH 0x40000 /**< Use high bitdepth */ 81 82 /*!\brief Generic fixed size buffer structure 83 * 84 * This structure is able to hold a reference to any fixed size buffer. 85 */ 86 typedef struct aom_fixed_buf { 87 void *buf; /**< Pointer to the data. Does NOT own the data! */ 88 size_t sz; /**< Length of the buffer, in chars */ 89 } aom_fixed_buf_t; /**< alias for struct aom_fixed_buf */ 90 91 /*!\brief Error Resilient flags 92 * 93 * These flags define which error resilient features to enable in the 94 * encoder. The flags are specified through the 95 * aom_codec_enc_cfg::g_error_resilient variable. 96 */ 97 typedef uint32_t aom_codec_er_flags_t; 98 /*!\brief Improve resiliency against losses of whole frames */ 99 #define AOM_ERROR_RESILIENT_DEFAULT 0x1 100 101 /*!\brief Encoder output packet variants 102 * 103 * This enumeration lists the different kinds of data packets that can be 104 * returned by calls to aom_codec_get_cx_data(). Algorithms \ref MAY 105 * extend this list to provide additional functionality. 106 */ 107 enum aom_codec_cx_pkt_kind { 108 AOM_CODEC_CX_FRAME_PKT, /**< Compressed video frame */ 109 AOM_CODEC_STATS_PKT, /**< Two-pass statistics for this frame */ 110 AOM_CODEC_FPMB_STATS_PKT, /**< first pass mb statistics for this frame */ 111 AOM_CODEC_PSNR_PKT, /**< PSNR statistics for this frame */ 112 AOM_CODEC_CUSTOM_PKT = 256 /**< Algorithm extensions */ 113 }; 114 115 /*!\brief Encoder output packet 116 * 117 * This structure contains the different kinds of output data the encoder 118 * may produce while compressing a frame. 119 */ 120 typedef struct aom_codec_cx_pkt { 121 enum aom_codec_cx_pkt_kind kind; /**< packet variant */ 122 union { 123 struct { 124 void *buf; /**< compressed data buffer */ 125 size_t sz; /**< length of compressed data */ 126 /*!\brief time stamp to show frame (in timebase units) */ 127 aom_codec_pts_t pts; 128 /*!\brief duration to show frame (in timebase units) */ 129 unsigned long duration; 130 aom_codec_frame_flags_t flags; /**< flags for this frame */ 131 /*!\brief the partition id defines the decoding order of the partitions. 132 * Only applicable when "output partition" mode is enabled. First 133 * partition has id 0.*/ 134 int partition_id; 135 /*!\brief size of the visible frame in this packet */ 136 size_t vis_frame_size; 137 } frame; /**< data for compressed frame packet */ 138 aom_fixed_buf_t twopass_stats; /**< data for two-pass packet */ 139 aom_fixed_buf_t firstpass_mb_stats; /**< first pass mb packet */ 140 struct aom_psnr_pkt { 141 unsigned int samples[4]; /**< Number of samples, total/y/u/v */ 142 uint64_t sse[4]; /**< sum squared error, total/y/u/v */ 143 double psnr[4]; /**< PSNR, total/y/u/v */ 144 /*!\brief Number of samples, total/y/u/v when 145 * input bit-depth < stream bit-depth.*/ 146 unsigned int samples_hbd[4]; 147 /*!\brief sum squared error, total/y/u/v when 148 * input bit-depth < stream bit-depth.*/ 149 uint64_t sse_hbd[4]; 150 /*!\brief PSNR, total/y/u/v when 151 * input bit-depth < stream bit-depth.*/ 152 double psnr_hbd[4]; 153 } psnr; /**< data for PSNR packet */ 154 aom_fixed_buf_t raw; /**< data for arbitrary packets */ 155 } data; /**< packet data */ 156 } aom_codec_cx_pkt_t; /**< alias for struct aom_codec_cx_pkt */ 157 158 /*!\brief Rational Number 159 * 160 * This structure holds a fractional value. 161 */ 162 typedef struct aom_rational { 163 int num; /**< fraction numerator */ 164 int den; /**< fraction denominator */ 165 } aom_rational_t; /**< alias for struct aom_rational */ 166 167 /*!\brief Multi-pass Encoding Pass 168 * 169 * AOM_RC_LAST_PASS is kept for backward compatibility. 170 * If passes is not given and pass==2, the codec will assume passes=2. 171 * For new code, it is recommended to use AOM_RC_SECOND_PASS and set 172 * the "passes" member to 2 via the key & val API for two-pass encoding. 173 */ 174 enum aom_enc_pass { 175 AOM_RC_ONE_PASS = 0, /**< Single pass mode */ 176 AOM_RC_FIRST_PASS = 1, /**< First pass of multi-pass mode */ 177 AOM_RC_SECOND_PASS = 2, /**< Second pass of multi-pass mode */ 178 AOM_RC_THIRD_PASS = 3, /**< Third pass of multi-pass mode */ 179 AOM_RC_LAST_PASS = 2, /**< Final pass of two-pass mode */ 180 }; 181 182 /*!\brief Rate control mode */ 183 enum aom_rc_mode { 184 AOM_VBR, /**< Variable Bit Rate (VBR) mode */ 185 AOM_CBR, /**< Constant Bit Rate (CBR) mode */ 186 AOM_CQ, /**< Constrained Quality (CQ) mode */ 187 AOM_Q, /**< Constant Quality (Q) mode */ 188 }; 189 190 /*!\brief Keyframe placement mode. 191 * 192 * This enumeration determines whether keyframes are placed automatically by 193 * the encoder or whether this behavior is disabled. Older releases of this 194 * SDK were implemented such that AOM_KF_FIXED meant keyframes were disabled. 195 * This name is confusing for this behavior, so the new symbols to be used 196 * are AOM_KF_AUTO and AOM_KF_DISABLED. 197 */ 198 enum aom_kf_mode { 199 AOM_KF_FIXED, /**< deprecated, implies AOM_KF_DISABLED */ 200 AOM_KF_AUTO, /**< Encoder determines optimal placement automatically */ 201 AOM_KF_DISABLED = 0 /**< Encoder does not place keyframes. */ 202 }; 203 204 /*!\brief Frame super-resolution mode. */ 205 typedef enum { 206 /**< Frame super-resolution is disabled for all frames. */ 207 AOM_SUPERRES_NONE, 208 /**< All frames are coded at the specified scale and super-resolved. */ 209 AOM_SUPERRES_FIXED, 210 /**< All frames are coded at a random scale and super-resolved. */ 211 AOM_SUPERRES_RANDOM, 212 /**< Super-resolution scale for each frame is determined based on the q index 213 of that frame. */ 214 AOM_SUPERRES_QTHRESH, 215 /**< Full-resolution or super-resolution and the scale (in case of 216 super-resolution) are automatically selected for each frame. */ 217 AOM_SUPERRES_AUTO, 218 } aom_superres_mode; 219 220 /*!\brief Encoder Config Options 221 * 222 * This type allows to enumerate and control flags defined for encoder control 223 * via config file at runtime. 224 */ 225 typedef struct cfg_options { 226 /*!\brief Indicate init by cfg file 227 * 0 or 1 228 */ 229 unsigned int init_by_cfg_file; 230 /*!\brief Superblock size 231 * 0, 64 or 128 232 */ 233 unsigned int super_block_size; 234 /*!\brief max partition size 235 * 8, 16, 32, 64, 128 236 */ 237 unsigned int max_partition_size; 238 /*!\brief min partition size 239 * 8, 16, 32, 64, 128 240 */ 241 unsigned int min_partition_size; 242 /*!\brief disable AB Shape partition type 243 * 244 */ 245 unsigned int disable_ab_partition_type; 246 /*!\brief disable rectangular partition type 247 * 248 */ 249 unsigned int disable_rect_partition_type; 250 /*!\brief disable 1:4/4:1 partition type 251 * 252 */ 253 unsigned int disable_1to4_partition_type; 254 /*!\brief disable flip and identity transform type 255 * 256 */ 257 unsigned int disable_flip_idtx; 258 /*!\brief disable CDEF filter 259 * 260 */ 261 unsigned int disable_cdef; 262 /*!\brief disable Loop Restoration Filter 263 * 264 */ 265 unsigned int disable_lr; 266 /*!\brief disable OBMC 267 * 268 */ 269 unsigned int disable_obmc; 270 /*!\brief disable Warped Motion 271 * 272 */ 273 unsigned int disable_warp_motion; 274 /*!\brief disable global motion 275 * 276 */ 277 unsigned int disable_global_motion; 278 /*!\brief disable dist weighted compound 279 * 280 */ 281 unsigned int disable_dist_wtd_comp; 282 /*!\brief disable diff weighted compound 283 * 284 */ 285 unsigned int disable_diff_wtd_comp; 286 /*!\brief disable inter/intra compound 287 * 288 */ 289 unsigned int disable_inter_intra_comp; 290 /*!\brief disable masked compound 291 * 292 */ 293 unsigned int disable_masked_comp; 294 /*!\brief disable one sided compound 295 * 296 */ 297 unsigned int disable_one_sided_comp; 298 /*!\brief disable Palette 299 * 300 */ 301 unsigned int disable_palette; 302 /*!\brief disable Intra Block Copy 303 * 304 */ 305 unsigned int disable_intrabc; 306 /*!\brief disable chroma from luma 307 * 308 */ 309 unsigned int disable_cfl; 310 /*!\brief disable intra smooth mode 311 * 312 */ 313 unsigned int disable_smooth_intra; 314 /*!\brief disable filter intra 315 * 316 */ 317 unsigned int disable_filter_intra; 318 /*!\brief disable dual filter 319 * 320 */ 321 unsigned int disable_dual_filter; 322 /*!\brief disable intra angle delta 323 * 324 */ 325 unsigned int disable_intra_angle_delta; 326 /*!\brief disable intra edge filter 327 * 328 */ 329 unsigned int disable_intra_edge_filter; 330 /*!\brief disable 64x64 transform 331 * 332 */ 333 unsigned int disable_tx_64x64; 334 /*!\brief disable smooth inter/intra 335 * 336 */ 337 unsigned int disable_smooth_inter_intra; 338 /*!\brief disable inter/inter wedge comp 339 * 340 */ 341 unsigned int disable_inter_inter_wedge; 342 /*!\brief disable inter/intra wedge comp 343 * 344 */ 345 unsigned int disable_inter_intra_wedge; 346 /*!\brief disable paeth intra 347 * 348 */ 349 unsigned int disable_paeth_intra; 350 /*!\brief disable trellis quantization 351 * 352 */ 353 unsigned int disable_trellis_quant; 354 /*!\brief disable ref frame MV 355 * 356 */ 357 unsigned int disable_ref_frame_mv; 358 /*!\brief use reduced reference frame set 359 * 360 */ 361 unsigned int reduced_reference_set; 362 /*!\brief use reduced transform type set 363 * 364 */ 365 unsigned int reduced_tx_type_set; 366 } cfg_options_t; 367 368 /*!\brief Encoded Frame Flags 369 * 370 * This type indicates a bitfield to be passed to aom_codec_encode(), defining 371 * per-frame boolean values. By convention, bits common to all codecs will be 372 * named AOM_EFLAG_*, and bits specific to an algorithm will be named 373 * /algo/_eflag_*. The lower order 16 bits are reserved for common use. 374 */ 375 typedef long aom_enc_frame_flags_t; 376 /*!\brief Force this frame to be a keyframe */ 377 #define AOM_EFLAG_FORCE_KF (1 << 0) 378 379 /*!\brief Encoder configuration structure 380 * 381 * This structure contains the encoder settings that have common representations 382 * across all codecs. This doesn't imply that all codecs support all features, 383 * however. 384 */ 385 typedef struct aom_codec_enc_cfg { 386 /* 387 * generic settings (g) 388 */ 389 390 /*!\brief Algorithm specific "usage" value 391 * 392 * Algorithms may define multiple values for usage, which may convey the 393 * intent of how the application intends to use the stream. If this value 394 * is non-zero, consult the documentation for the codec to determine its 395 * meaning. 396 */ 397 unsigned int g_usage; 398 399 /*!\brief Maximum number of threads to use 400 * 401 * For multi-threaded implementations, use no more than this number of 402 * threads. The codec may use fewer threads than allowed. The value 403 * 0 is equivalent to the value 1. 404 */ 405 unsigned int g_threads; 406 407 /*!\brief Bitstream profile to use 408 * 409 * Some codecs support a notion of multiple bitstream profiles. Typically 410 * this maps to a set of features that are turned on or off. Often the 411 * profile to use is determined by the features of the intended decoder. 412 * Consult the documentation for the codec to determine the valid values 413 * for this parameter, or set to zero for a sane default. 414 */ 415 unsigned int g_profile; /**< profile of bitstream to use */ 416 417 /*!\brief Width of the frame 418 * 419 * This value identifies the presentation resolution of the frame, 420 * in pixels. Note that the frames passed as input to the encoder must 421 * have this resolution. Frames will be presented by the decoder in this 422 * resolution, independent of any spatial resampling the encoder may do. 423 */ 424 unsigned int g_w; 425 426 /*!\brief Height of the frame 427 * 428 * This value identifies the presentation resolution of the frame, 429 * in pixels. Note that the frames passed as input to the encoder must 430 * have this resolution. Frames will be presented by the decoder in this 431 * resolution, independent of any spatial resampling the encoder may do. 432 */ 433 unsigned int g_h; 434 435 /*!\brief Max number of frames to encode 436 * 437 * If force video mode is off (the default) and g_limit is 1, the encoder 438 * will encode a still picture (still_picture is set to 1 in the sequence 439 * header OBU). If in addition full_still_picture_hdr is 0 (the default), 440 * the encoder will use a reduced header (reduced_still_picture_header is 441 * set to 1 in the sequence header OBU) for the still picture. 442 */ 443 unsigned int g_limit; 444 445 /*!\brief Forced maximum width of the frame 446 * 447 * If this value is non-zero then it is used to force the maximum frame 448 * width written in write_sequence_header(). 449 */ 450 unsigned int g_forced_max_frame_width; 451 452 /*!\brief Forced maximum height of the frame 453 * 454 * If this value is non-zero then it is used to force the maximum frame 455 * height written in write_sequence_header(). 456 */ 457 unsigned int g_forced_max_frame_height; 458 459 /*!\brief Bit-depth of the codec 460 * 461 * This value identifies the bit_depth of the codec, 462 * Only certain bit-depths are supported as identified in the 463 * aom_bit_depth_t enum. 464 */ 465 aom_bit_depth_t g_bit_depth; 466 467 /*!\brief Bit-depth of the input frames 468 * 469 * This value identifies the bit_depth of the input frames in bits. 470 * Note that the frames passed as input to the encoder must have 471 * this bit-depth. 472 */ 473 unsigned int g_input_bit_depth; 474 475 /*!\brief Stream timebase units 476 * 477 * Indicates the smallest interval of time, in seconds, used by the stream. 478 * For fixed frame rate material, or variable frame rate material where 479 * frames are timed at a multiple of a given clock (ex: video capture), 480 * the \ref RECOMMENDED method is to set the timebase to the reciprocal 481 * of the frame rate (ex: 1001/30000 for 29.970 Hz NTSC). This allows the 482 * pts to correspond to the frame number, which can be handy. For 483 * re-encoding video from containers with absolute time timestamps, the 484 * \ref RECOMMENDED method is to set the timebase to that of the parent 485 * container or multimedia framework (ex: 1/1000 for ms, as in FLV). 486 */ 487 struct aom_rational g_timebase; 488 489 /*!\brief Enable error resilient modes. 490 * 491 * The error resilient bitfield indicates to the encoder which features 492 * it should enable to take measures for streaming over lossy or noisy 493 * links. 494 */ 495 aom_codec_er_flags_t g_error_resilient; 496 497 /*!\brief Multi-pass Encoding Mode 498 * 499 * This value should be set to the current phase for multi-pass encoding. 500 * For single pass, set to #AOM_RC_ONE_PASS. 501 */ 502 enum aom_enc_pass g_pass; 503 504 /*!\brief Allow lagged encoding 505 * 506 * If set, this value allows the encoder to consume a number of input 507 * frames before producing output frames. This allows the encoder to 508 * base decisions for the current frame on future frames. This does 509 * increase the latency of the encoding pipeline, so it is not appropriate 510 * in all situations (ex: realtime encoding). 511 * 512 * Note that this is a maximum value -- the encoder may produce frames 513 * sooner than the given limit. Set this value to 0 to disable this 514 * feature. 515 */ 516 unsigned int g_lag_in_frames; 517 518 /* 519 * rate control settings (rc) 520 */ 521 522 /*!\brief Temporal resampling configuration, if supported by the codec. 523 * 524 * Temporal resampling allows the codec to "drop" frames as a strategy to 525 * meet its target data rate. This can cause temporal discontinuities in 526 * the encoded video, which may appear as stuttering during playback. This 527 * trade-off is often acceptable, but for many applications is not. It can 528 * be disabled in these cases. 529 * 530 * Note that not all codecs support this feature. All aom AVx codecs do. 531 * For other codecs, consult the documentation for that algorithm. 532 * 533 * This threshold is described as a percentage of the target data buffer. 534 * When the data buffer falls below this percentage of fullness, a 535 * dropped frame is indicated. Set the threshold to zero (0) to disable 536 * this feature. 537 */ 538 unsigned int rc_dropframe_thresh; 539 540 /*!\brief Mode for spatial resampling, if supported by the codec. 541 * 542 * Spatial resampling allows the codec to compress a lower resolution 543 * version of the frame, which is then upscaled by the decoder to the 544 * correct presentation resolution. This increases visual quality at 545 * low data rates, at the expense of CPU time on the encoder/decoder. 546 */ 547 unsigned int rc_resize_mode; 548 549 /*!\brief Frame resize denominator. 550 * 551 * The denominator for resize to use, assuming 8 as the numerator. 552 * 553 * Valid denominators are 8 - 16 for now. 554 */ 555 unsigned int rc_resize_denominator; 556 557 /*!\brief Keyframe resize denominator. 558 * 559 * The denominator for resize to use, assuming 8 as the numerator. 560 * 561 * Valid denominators are 8 - 16 for now. 562 */ 563 unsigned int rc_resize_kf_denominator; 564 565 /*!\brief Frame super-resolution scaling mode. 566 * 567 * Similar to spatial resampling, frame super-resolution integrates 568 * upscaling after the encode/decode process. Taking control of upscaling and 569 * using restoration filters should allow it to outperform normal resizing. 570 */ 571 aom_superres_mode rc_superres_mode; 572 573 /*!\brief Frame super-resolution denominator. 574 * 575 * The denominator for superres to use. If fixed it will only change if the 576 * cumulative scale change over resizing and superres is greater than 1/2; 577 * this forces superres to reduce scaling. 578 * 579 * Valid denominators are 8 to 16. 580 * 581 * Used only by AOM_SUPERRES_FIXED. 582 */ 583 unsigned int rc_superres_denominator; 584 585 /*!\brief Keyframe super-resolution denominator. 586 * 587 * The denominator for superres to use. If fixed it will only change if the 588 * cumulative scale change over resizing and superres is greater than 1/2; 589 * this forces superres to reduce scaling. 590 * 591 * Valid denominators are 8 - 16 for now. 592 */ 593 unsigned int rc_superres_kf_denominator; 594 595 /*!\brief Frame super-resolution q threshold. 596 * 597 * The q level threshold after which superres is used. 598 * Valid values are 1 to 63. 599 * 600 * Used only by AOM_SUPERRES_QTHRESH 601 */ 602 unsigned int rc_superres_qthresh; 603 604 /*!\brief Keyframe super-resolution q threshold. 605 * 606 * The q level threshold after which superres is used for key frames. 607 * Valid values are 1 to 63. 608 * 609 * Used only by AOM_SUPERRES_QTHRESH 610 */ 611 unsigned int rc_superres_kf_qthresh; 612 613 /*!\brief Rate control algorithm to use. 614 * 615 * Indicates whether the end usage of this stream is to be streamed over 616 * a bandwidth constrained link, indicating that Constant Bit Rate (CBR) 617 * mode should be used, or whether it will be played back on a high 618 * bandwidth link, as from a local disk, where higher variations in 619 * bitrate are acceptable. 620 */ 621 enum aom_rc_mode rc_end_usage; 622 623 /*!\brief Two-pass stats buffer. 624 * 625 * A buffer containing all of the stats packets produced in the first 626 * pass, concatenated. 627 */ 628 aom_fixed_buf_t rc_twopass_stats_in; 629 630 /*!\brief first pass mb stats buffer. 631 * 632 * A buffer containing all of the first pass mb stats packets produced 633 * in the first pass, concatenated. 634 */ 635 aom_fixed_buf_t rc_firstpass_mb_stats_in; 636 637 /*!\brief Target data rate 638 * 639 * Target bitrate to use for this stream, in kilobits per second. 640 */ 641 unsigned int rc_target_bitrate; 642 643 /* 644 * quantizer settings 645 */ 646 647 /*!\brief Minimum (Best Quality) Quantizer 648 * 649 * The quantizer is the most direct control over the quality of the 650 * encoded image. The range of valid values for the quantizer is codec 651 * specific. Consult the documentation for the codec to determine the 652 * values to use. To determine the range programmatically, call 653 * aom_codec_enc_config_default() with a usage value of 0. 654 */ 655 unsigned int rc_min_quantizer; 656 657 /*!\brief Maximum (Worst Quality) Quantizer 658 * 659 * The quantizer is the most direct control over the quality of the 660 * encoded image. The range of valid values for the quantizer is codec 661 * specific. Consult the documentation for the codec to determine the 662 * values to use. To determine the range programmatically, call 663 * aom_codec_enc_config_default() with a usage value of 0. 664 */ 665 unsigned int rc_max_quantizer; 666 667 /* 668 * bitrate tolerance 669 */ 670 671 /*!\brief Rate control adaptation undershoot control 672 * 673 * This value, controls the tolerance of the VBR algorithm to undershoot 674 * and is used as a trigger threshold for more aggressive adaptation of Q. 675 * 676 * Valid values in the range 0-100. 677 */ 678 unsigned int rc_undershoot_pct; 679 680 /*!\brief Rate control adaptation overshoot control 681 * 682 * This value, controls the tolerance of the VBR algorithm to overshoot 683 * and is used as a trigger threshold for more aggressive adaptation of Q. 684 * 685 * Valid values in the range 0-100. 686 */ 687 unsigned int rc_overshoot_pct; 688 689 /* 690 * decoder buffer model parameters 691 */ 692 693 /*!\brief Decoder Buffer Size 694 * 695 * This value indicates the amount of data that may be buffered by the 696 * decoding application. Note that this value is expressed in units of 697 * time (milliseconds). For example, a value of 5000 indicates that the 698 * client will buffer (at least) 5000ms worth of encoded data. Use the 699 * target bitrate (#rc_target_bitrate) to convert to bits/bytes, if 700 * necessary. 701 */ 702 unsigned int rc_buf_sz; 703 704 /*!\brief Decoder Buffer Initial Size 705 * 706 * This value indicates the amount of data that will be buffered by the 707 * decoding application prior to beginning playback. This value is 708 * expressed in units of time (milliseconds). Use the target bitrate 709 * (#rc_target_bitrate) to convert to bits/bytes, if necessary. 710 */ 711 unsigned int rc_buf_initial_sz; 712 713 /*!\brief Decoder Buffer Optimal Size 714 * 715 * This value indicates the amount of data that the encoder should try 716 * to maintain in the decoder's buffer. This value is expressed in units 717 * of time (milliseconds). Use the target bitrate (#rc_target_bitrate) 718 * to convert to bits/bytes, if necessary. 719 */ 720 unsigned int rc_buf_optimal_sz; 721 722 /* 723 * 2 pass rate control parameters 724 */ 725 726 /*!\brief Two-pass mode CBR/VBR bias 727 * 728 * Bias, expressed on a scale of 0 to 100, for determining target size 729 * for the current frame. The value 0 indicates the optimal CBR mode 730 * value should be used. The value 100 indicates the optimal VBR mode 731 * value should be used. Values in between indicate which way the 732 * encoder should "lean." 733 */ 734 unsigned int rc_2pass_vbr_bias_pct; 735 736 /*!\brief Two-pass mode per-GOP minimum bitrate 737 * 738 * This value, expressed as a percentage of the target bitrate, indicates 739 * the minimum bitrate to be used for a single GOP (aka "section") 740 */ 741 unsigned int rc_2pass_vbr_minsection_pct; 742 743 /*!\brief Two-pass mode per-GOP maximum bitrate 744 * 745 * This value, expressed as a percentage of the target bitrate, indicates 746 * the maximum bitrate to be used for a single GOP (aka "section") 747 */ 748 unsigned int rc_2pass_vbr_maxsection_pct; 749 750 /* 751 * keyframing settings (kf) 752 */ 753 754 /*!\brief Option to enable forward reference key frame 755 * 756 */ 757 int fwd_kf_enabled; 758 759 /*!\brief Keyframe placement mode 760 * 761 * This value indicates whether the encoder should place keyframes at a 762 * fixed interval, or determine the optimal placement automatically 763 * (as governed by the #kf_min_dist and #kf_max_dist parameters) 764 */ 765 enum aom_kf_mode kf_mode; 766 767 /*!\brief Keyframe minimum interval 768 * 769 * This value, expressed as a number of frames, prevents the encoder from 770 * placing a keyframe nearer than kf_min_dist to the previous keyframe. At 771 * least kf_min_dist frames non-keyframes will be coded before the next 772 * keyframe. Set kf_min_dist equal to kf_max_dist for a fixed interval. 773 */ 774 unsigned int kf_min_dist; 775 776 /*!\brief Keyframe maximum interval 777 * 778 * This value, expressed as a number of frames, forces the encoder to code 779 * a keyframe if one has not been coded in the last kf_max_dist frames. 780 * A value of 0 implies all frames will be keyframes. Set kf_min_dist 781 * equal to kf_max_dist for a fixed interval. 782 */ 783 unsigned int kf_max_dist; 784 785 /*!\brief sframe interval 786 * 787 * This value, expressed as a number of frames, forces the encoder to code 788 * an S-Frame every sframe_dist frames. 789 */ 790 unsigned int sframe_dist; 791 792 /*!\brief sframe insertion mode 793 * 794 * This value must be set to 1 or 2, and tells the encoder how to insert 795 * S-Frames. It will only have an effect if sframe_dist != 0. 796 * 797 * If altref is enabled: 798 * - if sframe_mode == 1, the considered frame will be made into an 799 * S-Frame only if it is an altref frame 800 * - if sframe_mode == 2, the next altref frame will be made into an 801 * S-Frame. 802 * 803 * Otherwise: the considered frame will be made into an S-Frame. 804 */ 805 unsigned int sframe_mode; 806 807 /*!\brief Tile coding mode 808 * 809 * This value indicates the tile coding mode. 810 * A value of 0 implies a normal non-large-scale tile coding. A value of 1 811 * implies a large-scale tile coding. 812 */ 813 unsigned int large_scale_tile; 814 815 /*!\brief Monochrome mode 816 * 817 * If this is nonzero, the encoder will generate a monochrome stream 818 * with no chroma planes. 819 */ 820 unsigned int monochrome; 821 822 /*!\brief full_still_picture_hdr 823 * 824 * If this is nonzero, the encoder will generate a full header 825 * (reduced_still_picture_header is set to 0 in the sequence header OBU) even 826 * for still picture encoding. If this is zero (the default), a reduced 827 * header (reduced_still_picture_header is set to 1 in the sequence header 828 * OBU) is used for still picture encoding. This flag has no effect when a 829 * regular video with more than a single frame is encoded. 830 */ 831 unsigned int full_still_picture_hdr; 832 833 /*!\brief Bitstream syntax mode 834 * 835 * This value indicates the bitstream syntax mode. 836 * A value of 0 indicates bitstream is saved as Section 5 bitstream. A value 837 * of 1 indicates the bitstream is saved in Annex-B format 838 */ 839 unsigned int save_as_annexb; 840 841 /*!\brief Number of explicit tile widths specified 842 * 843 * This value indicates the number of tile widths specified 844 * A value of 0 implies no tile widths are specified. 845 * Tile widths are given in the array tile_widths[] 846 */ 847 int tile_width_count; 848 849 /*!\brief Number of explicit tile heights specified 850 * 851 * This value indicates the number of tile heights specified 852 * A value of 0 implies no tile heights are specified. 853 * Tile heights are given in the array tile_heights[] 854 */ 855 int tile_height_count; 856 857 /*!\brief Maximum number of tile widths in tile widths array 858 * 859 * This define gives the maximum number of elements in the tile_widths array. 860 */ 861 #define MAX_TILE_WIDTHS 64 // maximum tile width array length 862 863 /*!\brief Array of specified tile widths 864 * 865 * This array specifies tile widths (and may be empty) 866 * The number of widths specified is given by tile_width_count 867 */ 868 int tile_widths[MAX_TILE_WIDTHS]; 869 870 /*!\brief Maximum number of tile heights in tile heights array. 871 * 872 * This define gives the maximum number of elements in the tile_heights array. 873 */ 874 #define MAX_TILE_HEIGHTS 64 // maximum tile height array length 875 876 /*!\brief Array of specified tile heights 877 * 878 * This array specifies tile heights (and may be empty) 879 * The number of heights specified is given by tile_height_count 880 */ 881 int tile_heights[MAX_TILE_HEIGHTS]; 882 883 /*!\brief Whether encoder should use fixed QP offsets. 884 * 885 * If a value of 1 is provided, encoder will use fixed QP offsets for frames 886 * at different levels of the pyramid. 887 * If a value of 0 is provided, encoder will NOT use fixed QP offsets. 888 * Note: This option is only relevant for --end-usage=q. 889 */ 890 unsigned int use_fixed_qp_offsets; 891 892 /*!\brief Deprecated and ignored. DO NOT USE. 893 * 894 * TODO(aomedia:3269): Remove fixed_qp_offsets in libaom v4.0.0. 895 */ 896 int fixed_qp_offsets[5]; 897 898 /*!\brief Options defined per config file 899 * 900 */ 901 cfg_options_t encoder_cfg; 902 } aom_codec_enc_cfg_t; /**< alias for struct aom_codec_enc_cfg */ 903 904 /*!\brief Initialize an encoder instance 905 * 906 * Initializes a encoder context using the given interface. Applications 907 * should call the aom_codec_enc_init convenience macro instead of this 908 * function directly, to ensure that the ABI version number parameter 909 * is properly initialized. 910 * 911 * If the library was configured with -DCONFIG_MULTITHREAD=0, this call 912 * is not thread safe and should be guarded with a lock if being used 913 * in a multithreaded context. 914 * 915 * \param[in] ctx Pointer to this instance's context. 916 * \param[in] iface Pointer to the algorithm interface to use. 917 * \param[in] cfg Configuration to use, if known. 918 * \param[in] flags Bitfield of AOM_CODEC_USE_* flags 919 * \param[in] ver ABI version number. Must be set to 920 * AOM_ENCODER_ABI_VERSION 921 * \retval #AOM_CODEC_OK 922 * The decoder algorithm initialized. 923 * \retval #AOM_CODEC_MEM_ERROR 924 * Memory allocation failed. 925 */ 926 aom_codec_err_t aom_codec_enc_init_ver(aom_codec_ctx_t *ctx, 927 aom_codec_iface_t *iface, 928 const aom_codec_enc_cfg_t *cfg, 929 aom_codec_flags_t flags, int ver); 930 931 /*!\brief Convenience macro for aom_codec_enc_init_ver() 932 * 933 * Ensures the ABI version parameter is properly set. 934 */ 935 #define aom_codec_enc_init(ctx, iface, cfg, flags) \ 936 aom_codec_enc_init_ver(ctx, iface, cfg, flags, AOM_ENCODER_ABI_VERSION) 937 938 /*!\brief Get the default configuration for a usage. 939 * 940 * Initializes an encoder configuration structure with default values. Supports 941 * the notion of "usages" so that an algorithm may offer different default 942 * settings depending on the user's intended goal. This function \ref SHOULD 943 * be called by all applications to initialize the configuration structure 944 * before specializing the configuration with application specific values. 945 * 946 * \param[in] iface Pointer to the algorithm interface to use. 947 * \param[out] cfg Configuration buffer to populate. 948 * \param[in] usage Algorithm specific usage value. For AV1, must be 949 * set to AOM_USAGE_GOOD_QUALITY (0), 950 * AOM_USAGE_REALTIME (1), or AOM_USAGE_ALL_INTRA (2). 951 * 952 * \retval #AOM_CODEC_OK 953 * The configuration was populated. 954 * \retval #AOM_CODEC_INCAPABLE 955 * Interface is not an encoder interface. 956 * \retval #AOM_CODEC_INVALID_PARAM 957 * A parameter was NULL, or the usage value was not recognized. 958 */ 959 aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface, 960 aom_codec_enc_cfg_t *cfg, 961 unsigned int usage); 962 963 /*!\brief Set or change configuration 964 * 965 * Reconfigures an encoder instance according to the given configuration. 966 * 967 * \param[in] ctx Pointer to this instance's context 968 * \param[in] cfg Configuration buffer to use 969 * 970 * \retval #AOM_CODEC_OK 971 * The configuration was populated. 972 * \retval #AOM_CODEC_INCAPABLE 973 * Interface is not an encoder interface. 974 * \retval #AOM_CODEC_INVALID_PARAM 975 * A parameter was NULL, or the usage value was not recognized. 976 */ 977 aom_codec_err_t aom_codec_enc_config_set(aom_codec_ctx_t *ctx, 978 const aom_codec_enc_cfg_t *cfg); 979 980 /*!\brief Get global stream headers 981 * 982 * Retrieves a stream level global header packet, if supported by the codec. 983 * Calls to this function should be deferred until all configuration information 984 * has been passed to libaom. Otherwise the global header data may be 985 * invalidated by additional configuration changes. 986 * 987 * The AV1 implementation of this function returns an OBU. The OBU returned is 988 * in Low Overhead Bitstream Format. Specifically, the obu_has_size_field bit is 989 * set, and the buffer contains the obu_size field for the returned OBU. 990 * 991 * \param[in] ctx Pointer to this instance's context 992 * 993 * \retval NULL 994 * Encoder does not support global header, or an error occurred while 995 * generating the global header. 996 * 997 * \retval Non-NULL 998 * Pointer to buffer containing global header packet. The caller owns the 999 * memory associated with this buffer, and must free the 'buf' member of the 1000 * aom_fixed_buf_t as well as the aom_fixed_buf_t pointer. Memory returned 1001 * must be freed via call to free(). 1002 */ 1003 aom_fixed_buf_t *aom_codec_get_global_headers(aom_codec_ctx_t *ctx); 1004 1005 /*!\brief usage parameter analogous to AV1 GOOD QUALITY mode. */ 1006 #define AOM_USAGE_GOOD_QUALITY (0) 1007 /*!\brief usage parameter analogous to AV1 REALTIME mode. */ 1008 #define AOM_USAGE_REALTIME (1) 1009 /*!\brief usage parameter analogous to AV1 all intra mode. */ 1010 #define AOM_USAGE_ALL_INTRA (2) 1011 1012 /*!\brief Encode a frame 1013 * 1014 * Encodes a video frame at the given "presentation time." The presentation 1015 * time stamp (PTS) \ref MUST be strictly increasing. 1016 * 1017 * When the last frame has been passed to the encoder, this function should 1018 * continue to be called in a loop, with the img parameter set to NULL. This 1019 * will signal the end-of-stream condition to the encoder and allow it to 1020 * encode any held buffers. Encoding is complete when aom_codec_encode() is 1021 * called with img set to NULL and aom_codec_get_cx_data() returns no data. 1022 * 1023 * \param[in] ctx Pointer to this instance's context 1024 * \param[in] img Image data to encode, NULL to flush. 1025 * Encoding sample values outside the range 1026 * [0..(1<<img->bit_depth)-1] is undefined behavior. 1027 * \param[in] pts Presentation time stamp, in timebase units. If img 1028 * is NULL, pts is ignored. 1029 * \param[in] duration Duration to show frame, in timebase units. If img 1030 * is not NULL, duration must be nonzero. If img is 1031 * NULL, duration is ignored. 1032 * \param[in] flags Flags to use for encoding this frame. 1033 * 1034 * \retval #AOM_CODEC_OK 1035 * The configuration was populated. 1036 * \retval #AOM_CODEC_INCAPABLE 1037 * Interface is not an encoder interface. 1038 * \retval #AOM_CODEC_INVALID_PARAM 1039 * A parameter was NULL, the image format is unsupported, etc. 1040 */ 1041 aom_codec_err_t aom_codec_encode(aom_codec_ctx_t *ctx, const aom_image_t *img, 1042 aom_codec_pts_t pts, unsigned long duration, 1043 aom_enc_frame_flags_t flags); 1044 1045 /*!\brief Set compressed data output buffer 1046 * 1047 * Sets the buffer that the codec should output the compressed data 1048 * into. This call effectively sets the buffer pointer returned in the 1049 * next AOM_CODEC_CX_FRAME_PKT packet. Subsequent packets will be 1050 * appended into this buffer. The buffer is preserved across frames, 1051 * so applications must periodically call this function after flushing 1052 * the accumulated compressed data to disk or to the network to reset 1053 * the pointer to the buffer's head. 1054 * 1055 * `pad_before` bytes will be skipped before writing the compressed 1056 * data, and `pad_after` bytes will be appended to the packet. The size 1057 * of the packet will be the sum of the size of the actual compressed 1058 * data, pad_before, and pad_after. The padding bytes will be preserved 1059 * (not overwritten). 1060 * 1061 * Note that calling this function does not guarantee that the returned 1062 * compressed data will be placed into the specified buffer. In the 1063 * event that the encoded data will not fit into the buffer provided, 1064 * the returned packet \ref MAY point to an internal buffer, as it would 1065 * if this call were never used. In this event, the output packet will 1066 * NOT have any padding, and the application must free space and copy it 1067 * to the proper place. This is of particular note in configurations 1068 * that may output multiple packets for a single encoded frame (e.g., lagged 1069 * encoding) or if the application does not reset the buffer periodically. 1070 * 1071 * Applications may restore the default behavior of the codec providing 1072 * the compressed data buffer by calling this function with a NULL 1073 * buffer. 1074 * 1075 * Applications \ref MUSTNOT call this function during iteration of 1076 * aom_codec_get_cx_data(). 1077 * 1078 * \param[in] ctx Pointer to this instance's context 1079 * \param[in] buf Buffer to store compressed data into 1080 * \param[in] pad_before Bytes to skip before writing compressed data 1081 * \param[in] pad_after Bytes to skip after writing compressed data 1082 * 1083 * \retval #AOM_CODEC_OK 1084 * The buffer was set successfully. 1085 * \retval #AOM_CODEC_INVALID_PARAM 1086 * A parameter was NULL, the image format is unsupported, etc. 1087 */ 1088 aom_codec_err_t aom_codec_set_cx_data_buf(aom_codec_ctx_t *ctx, 1089 const aom_fixed_buf_t *buf, 1090 unsigned int pad_before, 1091 unsigned int pad_after); 1092 1093 /*!\brief Encoded data iterator 1094 * 1095 * Iterates over a list of data packets to be passed from the encoder to the 1096 * application. The different kinds of packets available are enumerated in 1097 * #aom_codec_cx_pkt_kind. 1098 * 1099 * #AOM_CODEC_CX_FRAME_PKT packets should be passed to the application's 1100 * muxer. Multiple compressed frames may be in the list. 1101 * #AOM_CODEC_STATS_PKT packets should be appended to a global buffer. 1102 * 1103 * The application \ref MUST silently ignore any packet kinds that it does 1104 * not recognize or support. 1105 * 1106 * The data buffers returned from this function are only guaranteed to be 1107 * valid until the application makes another call to any aom_codec_* function. 1108 * 1109 * \param[in] ctx Pointer to this instance's context 1110 * \param[in,out] iter Iterator storage, initialized to NULL 1111 * 1112 * \return Returns a pointer to an output data packet (compressed frame data, 1113 * two-pass statistics, etc.) or NULL to signal end-of-list. 1114 * 1115 */ 1116 const aom_codec_cx_pkt_t *aom_codec_get_cx_data(aom_codec_ctx_t *ctx, 1117 aom_codec_iter_t *iter); 1118 1119 /*!\brief Get Preview Frame 1120 * 1121 * Returns an image that can be used as a preview. Shows the image as it would 1122 * exist at the decompressor. The application \ref MUST NOT write into this 1123 * image buffer. 1124 * 1125 * \param[in] ctx Pointer to this instance's context 1126 * 1127 * \return Returns a pointer to a preview image, or NULL if no image is 1128 * available. 1129 * 1130 */ 1131 const aom_image_t *aom_codec_get_preview_frame(aom_codec_ctx_t *ctx); 1132 1133 /*!@} - end defgroup encoder*/ 1134 #ifdef __cplusplus 1135 } 1136 #endif 1137 #endif // AOM_AOM_AOM_ENCODER_H_ 1138