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 * \internal 39 * If this file is altered in any way that changes the ABI, this value 40 * must be bumped. Examples include, but are not limited to, changing 41 * types, removing or reassigning enums, adding/removing/rearranging 42 * fields to structures 43 */ 44 #define AOM_ENCODER_ABI_VERSION \ 45 (10 + AOM_CODEC_ABI_VERSION + AOM_EXT_PART_ABI_VERSION) /**<\hideinitializer*/ 46 47 /*! \brief Encoder capabilities bitfield 48 * 49 * Each encoder advertises the capabilities it supports as part of its 50 * ::aom_codec_iface_t interface structure. Capabilities are extra 51 * interfaces or functionality, and are not required to be supported 52 * by an encoder. 53 * 54 * The available flags are specified by AOM_CODEC_CAP_* defines. 55 */ 56 #define AOM_CODEC_CAP_PSNR 0x10000 /**< Can issue PSNR packets */ 57 58 /*! Can support input images at greater than 8 bitdepth. 59 */ 60 #define AOM_CODEC_CAP_HIGHBITDEPTH 0x40000 61 62 /*! \brief Initialization-time Feature Enabling 63 * 64 * Certain codec features must be known at initialization time, to allow 65 * for proper memory allocation. 66 * 67 * The available flags are specified by AOM_CODEC_USE_* defines. 68 */ 69 #define AOM_CODEC_USE_PSNR 0x10000 /**< Calculate PSNR on each frame */ 70 /*!\brief Make the encoder output one partition at a time. */ 71 #define AOM_CODEC_USE_HIGHBITDEPTH 0x40000 /**< Use high bitdepth */ 72 73 /*!\brief Generic fixed size buffer structure 74 * 75 * This structure is able to hold a reference to any fixed size buffer. 76 */ 77 typedef struct aom_fixed_buf { 78 void *buf; /**< Pointer to the data. Does NOT own the data! */ 79 size_t sz; /**< Length of the buffer, in chars */ 80 } aom_fixed_buf_t; /**< alias for struct aom_fixed_buf */ 81 82 /*!\brief Error Resilient flags 83 * 84 * These flags define which error resilient features to enable in the 85 * encoder. The flags are specified through the 86 * aom_codec_enc_cfg::g_error_resilient variable. 87 */ 88 typedef uint32_t aom_codec_er_flags_t; 89 /*!\brief Improve resiliency against losses of whole frames */ 90 #define AOM_ERROR_RESILIENT_DEFAULT 0x1 91 92 /*!\brief Encoder output packet variants 93 * 94 * This enumeration lists the different kinds of data packets that can be 95 * returned by calls to aom_codec_get_cx_data(). Algorithms \ref MAY 96 * extend this list to provide additional functionality. 97 */ 98 enum aom_codec_cx_pkt_kind { 99 AOM_CODEC_CX_FRAME_PKT, /**< Compressed video frame */ 100 AOM_CODEC_STATS_PKT, /**< Two-pass statistics for this frame */ 101 AOM_CODEC_FPMB_STATS_PKT, /**< first pass mb statistics for this frame */ 102 AOM_CODEC_PSNR_PKT, /**< PSNR statistics for this frame */ 103 AOM_CODEC_CUSTOM_PKT = 256 /**< Algorithm extensions */ 104 }; 105 106 /*!\brief Encoder output packet 107 * 108 * This structure contains the different kinds of output data the encoder 109 * may produce while compressing a frame. 110 */ 111 typedef struct aom_codec_cx_pkt { 112 enum aom_codec_cx_pkt_kind kind; /**< packet variant */ 113 union { 114 struct { 115 void *buf; /**< compressed data buffer */ 116 size_t sz; /**< length of compressed data */ 117 /*!\brief time stamp to show frame (in timebase units) */ 118 aom_codec_pts_t pts; 119 /*!\brief duration to show frame (in timebase units) */ 120 unsigned long duration; 121 aom_codec_frame_flags_t flags; /**< flags for this frame */ 122 /*!\brief the partition id defines the decoding order of the partitions. 123 * Only applicable when "output partition" mode is enabled. First 124 * partition has id 0.*/ 125 int partition_id; 126 /*!\brief size of the visible frame in this packet */ 127 size_t vis_frame_size; 128 } frame; /**< data for compressed frame packet */ 129 aom_fixed_buf_t twopass_stats; /**< data for two-pass packet */ 130 aom_fixed_buf_t firstpass_mb_stats; /**< first pass mb packet */ 131 struct aom_psnr_pkt { 132 unsigned int samples[4]; /**< Number of samples, total/y/u/v */ 133 uint64_t sse[4]; /**< sum squared error, total/y/u/v */ 134 double psnr[4]; /**< PSNR, total/y/u/v */ 135 /*!\brief Number of samples, total/y/u/v when 136 * input bit-depth < stream bit-depth.*/ 137 unsigned int samples_hbd[4]; 138 /*!\brief sum squared error, total/y/u/v when 139 * input bit-depth < stream bit-depth.*/ 140 uint64_t sse_hbd[4]; 141 /*!\brief PSNR, total/y/u/v when 142 * input bit-depth < stream bit-depth.*/ 143 double psnr_hbd[4]; 144 } psnr; /**< data for PSNR packet */ 145 aom_fixed_buf_t raw; /**< data for arbitrary packets */ 146 } data; /**< packet data */ 147 } aom_codec_cx_pkt_t; /**< alias for struct aom_codec_cx_pkt */ 148 149 /*!\brief Rational Number 150 * 151 * This structure holds a fractional value. 152 */ 153 typedef struct aom_rational { 154 int num; /**< fraction numerator */ 155 int den; /**< fraction denominator */ 156 } aom_rational_t; /**< alias for struct aom_rational */ 157 158 /*!\brief Multi-pass Encoding Pass 159 * 160 * AOM_RC_LAST_PASS is kept for backward compatibility. 161 * If passes is not given and pass==2, the codec will assume passes=2. 162 * For new code, it is recommended to use AOM_RC_SECOND_PASS and set 163 * the "passes" member to 2 via the key & val API for two-pass encoding. 164 */ 165 enum aom_enc_pass { 166 AOM_RC_ONE_PASS = 0, /**< Single pass mode */ 167 AOM_RC_FIRST_PASS = 1, /**< First pass of multi-pass mode */ 168 AOM_RC_SECOND_PASS = 2, /**< Second pass of multi-pass mode */ 169 AOM_RC_THIRD_PASS = 3, /**< Third pass of multi-pass mode */ 170 AOM_RC_LAST_PASS = 2, /**< Final pass of two-pass mode */ 171 }; 172 173 /*!\brief Rate control mode */ 174 enum aom_rc_mode { 175 AOM_VBR, /**< Variable Bit Rate (VBR) mode */ 176 AOM_CBR, /**< Constant Bit Rate (CBR) mode */ 177 AOM_CQ, /**< Constrained Quality (CQ) mode */ 178 AOM_Q, /**< Constant Quality (Q) mode */ 179 }; 180 181 /*!\brief Keyframe placement mode. 182 * 183 * This enumeration determines whether keyframes are placed automatically by 184 * the encoder or whether this behavior is disabled. Older releases of this 185 * SDK were implemented such that AOM_KF_FIXED meant keyframes were disabled. 186 * This name is confusing for this behavior, so the new symbols to be used 187 * are AOM_KF_AUTO and AOM_KF_DISABLED. 188 */ 189 enum aom_kf_mode { 190 AOM_KF_FIXED, /**< deprecated, implies AOM_KF_DISABLED */ 191 AOM_KF_AUTO, /**< Encoder determines optimal placement automatically */ 192 AOM_KF_DISABLED = 0 /**< Encoder does not place keyframes. */ 193 }; 194 195 /*!\brief Frame super-resolution mode. */ 196 typedef enum { 197 /**< Frame super-resolution is disabled for all frames. */ 198 AOM_SUPERRES_NONE, 199 /**< All frames are coded at the specified scale and super-resolved. */ 200 AOM_SUPERRES_FIXED, 201 /**< All frames are coded at a random scale and super-resolved. */ 202 AOM_SUPERRES_RANDOM, 203 /**< Super-resolution scale for each frame is determined based on the q index 204 of that frame. */ 205 AOM_SUPERRES_QTHRESH, 206 /**< Full-resolution or super-resolution and the scale (in case of 207 super-resolution) are automatically selected for each frame. */ 208 AOM_SUPERRES_AUTO, 209 } aom_superres_mode; 210 211 /*!\brief Encoder Config Options 212 * 213 * This type allows to enumerate and control flags defined for encoder control 214 * via config file at runtime. 215 */ 216 typedef struct cfg_options { 217 /*!\brief Indicate init by cfg file 218 * 0 or 1 219 */ 220 unsigned int init_by_cfg_file; 221 /*!\brief Superblock size 222 * 0, 64 or 128 223 */ 224 unsigned int super_block_size; 225 /*!\brief max partition size 226 * 8, 16, 32, 64, 128 227 */ 228 unsigned int max_partition_size; 229 /*!\brief min partition size 230 * 8, 16, 32, 64, 128 231 */ 232 unsigned int min_partition_size; 233 /*!\brief disable AB Shape partition type 234 * 235 */ 236 unsigned int disable_ab_partition_type; 237 /*!\brief disable rectangular partition type 238 * 239 */ 240 unsigned int disable_rect_partition_type; 241 /*!\brief disable 1:4/4:1 partition type 242 * 243 */ 244 unsigned int disable_1to4_partition_type; 245 /*!\brief disable flip and identity transform type 246 * 247 */ 248 unsigned int disable_flip_idtx; 249 /*!\brief disable CDEF filter 250 * 251 */ 252 unsigned int disable_cdef; 253 /*!\brief disable Loop Restoration Filter 254 * 255 */ 256 unsigned int disable_lr; 257 /*!\brief disable OBMC 258 * 259 */ 260 unsigned int disable_obmc; 261 /*!\brief disable Warped Motion 262 * 263 */ 264 unsigned int disable_warp_motion; 265 /*!\brief disable global motion 266 * 267 */ 268 unsigned int disable_global_motion; 269 /*!\brief disable dist weighted compound 270 * 271 */ 272 unsigned int disable_dist_wtd_comp; 273 /*!\brief disable diff weighted compound 274 * 275 */ 276 unsigned int disable_diff_wtd_comp; 277 /*!\brief disable inter/intra compound 278 * 279 */ 280 unsigned int disable_inter_intra_comp; 281 /*!\brief disable masked compound 282 * 283 */ 284 unsigned int disable_masked_comp; 285 /*!\brief disable one sided compound 286 * 287 */ 288 unsigned int disable_one_sided_comp; 289 /*!\brief disable Palette 290 * 291 */ 292 unsigned int disable_palette; 293 /*!\brief disable Intra Block Copy 294 * 295 */ 296 unsigned int disable_intrabc; 297 /*!\brief disable chroma from luma 298 * 299 */ 300 unsigned int disable_cfl; 301 /*!\brief disable intra smooth mode 302 * 303 */ 304 unsigned int disable_smooth_intra; 305 /*!\brief disable filter intra 306 * 307 */ 308 unsigned int disable_filter_intra; 309 /*!\brief disable dual filter 310 * 311 */ 312 unsigned int disable_dual_filter; 313 /*!\brief disable intra angle delta 314 * 315 */ 316 unsigned int disable_intra_angle_delta; 317 /*!\brief disable intra edge filter 318 * 319 */ 320 unsigned int disable_intra_edge_filter; 321 /*!\brief disable 64x64 transform 322 * 323 */ 324 unsigned int disable_tx_64x64; 325 /*!\brief disable smooth inter/intra 326 * 327 */ 328 unsigned int disable_smooth_inter_intra; 329 /*!\brief disable inter/inter wedge comp 330 * 331 */ 332 unsigned int disable_inter_inter_wedge; 333 /*!\brief disable inter/intra wedge comp 334 * 335 */ 336 unsigned int disable_inter_intra_wedge; 337 /*!\brief disable paeth intra 338 * 339 */ 340 unsigned int disable_paeth_intra; 341 /*!\brief disable trellis quantization 342 * 343 */ 344 unsigned int disable_trellis_quant; 345 /*!\brief disable ref frame MV 346 * 347 */ 348 unsigned int disable_ref_frame_mv; 349 /*!\brief use reduced reference frame set 350 * 351 */ 352 unsigned int reduced_reference_set; 353 /*!\brief use reduced transform type set 354 * 355 */ 356 unsigned int reduced_tx_type_set; 357 } cfg_options_t; 358 359 /*!\brief Encoded Frame Flags 360 * 361 * This type indicates a bitfield to be passed to aom_codec_encode(), defining 362 * per-frame boolean values. By convention, bits common to all codecs will be 363 * named AOM_EFLAG_*, and bits specific to an algorithm will be named 364 * /algo/_eflag_*. The lower order 16 bits are reserved for common use. 365 */ 366 typedef long aom_enc_frame_flags_t; 367 /*!\brief Force this frame to be a keyframe */ 368 #define AOM_EFLAG_FORCE_KF (1 << 0) 369 370 /*!\brief Encoder configuration structure 371 * 372 * This structure contains the encoder settings that have common representations 373 * across all codecs. This doesn't imply that all codecs support all features, 374 * however. 375 */ 376 typedef struct aom_codec_enc_cfg { 377 /* 378 * generic settings (g) 379 */ 380 381 /*!\brief Algorithm specific "usage" value 382 * 383 * Algorithms may define multiple values for usage, which may convey the 384 * intent of how the application intends to use the stream. If this value 385 * is non-zero, consult the documentation for the codec to determine its 386 * meaning. 387 */ 388 unsigned int g_usage; 389 390 /*!\brief Maximum number of threads to use 391 * 392 * For multi-threaded implementations, use no more than this number of 393 * threads. The codec may use fewer threads than allowed. The value 394 * 0 is equivalent to the value 1. 395 */ 396 unsigned int g_threads; 397 398 /*!\brief Bitstream profile to use 399 * 400 * Some codecs support a notion of multiple bitstream profiles. Typically 401 * this maps to a set of features that are turned on or off. Often the 402 * profile to use is determined by the features of the intended decoder. 403 * Consult the documentation for the codec to determine the valid values 404 * for this parameter, or set to zero for a sane default. 405 */ 406 unsigned int g_profile; /**< profile of bitstream to use */ 407 408 /*!\brief Width of the frame 409 * 410 * This value identifies the presentation resolution of the frame, 411 * in pixels. Note that the frames passed as input to the encoder must 412 * have this resolution. Frames will be presented by the decoder in this 413 * resolution, independent of any spatial resampling the encoder may do. 414 */ 415 unsigned int g_w; 416 417 /*!\brief Height 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_h; 425 426 /*!\brief Max number of frames to encode 427 * 428 */ 429 unsigned int g_limit; 430 431 /*!\brief Forced maximum width of the frame 432 * 433 * If this value is non-zero then it is used to force the maximum frame 434 * width written in write_sequence_header(). 435 */ 436 unsigned int g_forced_max_frame_width; 437 438 /*!\brief Forced maximum height of the frame 439 * 440 * If this value is non-zero then it is used to force the maximum frame 441 * height written in write_sequence_header(). 442 */ 443 unsigned int g_forced_max_frame_height; 444 445 /*!\brief Bit-depth of the codec 446 * 447 * This value identifies the bit_depth of the codec, 448 * Only certain bit-depths are supported as identified in the 449 * aom_bit_depth_t enum. 450 */ 451 aom_bit_depth_t g_bit_depth; 452 453 /*!\brief Bit-depth of the input frames 454 * 455 * This value identifies the bit_depth of the input frames in bits. 456 * Note that the frames passed as input to the encoder must have 457 * this bit-depth. 458 */ 459 unsigned int g_input_bit_depth; 460 461 /*!\brief Stream timebase units 462 * 463 * Indicates the smallest interval of time, in seconds, used by the stream. 464 * For fixed frame rate material, or variable frame rate material where 465 * frames are timed at a multiple of a given clock (ex: video capture), 466 * the \ref RECOMMENDED method is to set the timebase to the reciprocal 467 * of the frame rate (ex: 1001/30000 for 29.970 Hz NTSC). This allows the 468 * pts to correspond to the frame number, which can be handy. For 469 * re-encoding video from containers with absolute time timestamps, the 470 * \ref RECOMMENDED method is to set the timebase to that of the parent 471 * container or multimedia framework (ex: 1/1000 for ms, as in FLV). 472 */ 473 struct aom_rational g_timebase; 474 475 /*!\brief Enable error resilient modes. 476 * 477 * The error resilient bitfield indicates to the encoder which features 478 * it should enable to take measures for streaming over lossy or noisy 479 * links. 480 */ 481 aom_codec_er_flags_t g_error_resilient; 482 483 /*!\brief Multi-pass Encoding Mode 484 * 485 * This value should be set to the current phase for multi-pass encoding. 486 * For single pass, set to #AOM_RC_ONE_PASS. 487 */ 488 enum aom_enc_pass g_pass; 489 490 /*!\brief Allow lagged encoding 491 * 492 * If set, this value allows the encoder to consume a number of input 493 * frames before producing output frames. This allows the encoder to 494 * base decisions for the current frame on future frames. This does 495 * increase the latency of the encoding pipeline, so it is not appropriate 496 * in all situations (ex: realtime encoding). 497 * 498 * Note that this is a maximum value -- the encoder may produce frames 499 * sooner than the given limit. Set this value to 0 to disable this 500 * feature. 501 */ 502 unsigned int g_lag_in_frames; 503 504 /* 505 * rate control settings (rc) 506 */ 507 508 /*!\brief Temporal resampling configuration, if supported by the codec. 509 * 510 * Temporal resampling allows the codec to "drop" frames as a strategy to 511 * meet its target data rate. This can cause temporal discontinuities in 512 * the encoded video, which may appear as stuttering during playback. This 513 * trade-off is often acceptable, but for many applications is not. It can 514 * be disabled in these cases. 515 * 516 * Note that not all codecs support this feature. All aom AVx codecs do. 517 * For other codecs, consult the documentation for that algorithm. 518 * 519 * This threshold is described as a percentage of the target data buffer. 520 * When the data buffer falls below this percentage of fullness, a 521 * dropped frame is indicated. Set the threshold to zero (0) to disable 522 * this feature. 523 */ 524 unsigned int rc_dropframe_thresh; 525 526 /*!\brief Mode for spatial resampling, if supported by the codec. 527 * 528 * Spatial resampling allows the codec to compress a lower resolution 529 * version of the frame, which is then upscaled by the decoder to the 530 * correct presentation resolution. This increases visual quality at 531 * low data rates, at the expense of CPU time on the encoder/decoder. 532 */ 533 unsigned int rc_resize_mode; 534 535 /*!\brief Frame resize denominator. 536 * 537 * The denominator for resize to use, assuming 8 as the numerator. 538 * 539 * Valid denominators are 8 - 16 for now. 540 */ 541 unsigned int rc_resize_denominator; 542 543 /*!\brief Keyframe resize denominator. 544 * 545 * The denominator for resize to use, assuming 8 as the numerator. 546 * 547 * Valid denominators are 8 - 16 for now. 548 */ 549 unsigned int rc_resize_kf_denominator; 550 551 /*!\brief Frame super-resolution scaling mode. 552 * 553 * Similar to spatial resampling, frame super-resolution integrates 554 * upscaling after the encode/decode process. Taking control of upscaling and 555 * using restoration filters should allow it to outperform normal resizing. 556 */ 557 aom_superres_mode rc_superres_mode; 558 559 /*!\brief Frame super-resolution denominator. 560 * 561 * The denominator for superres to use. If fixed it will only change if the 562 * cumulative scale change over resizing and superres is greater than 1/2; 563 * this forces superres to reduce scaling. 564 * 565 * Valid denominators are 8 to 16. 566 * 567 * Used only by AOM_SUPERRES_FIXED. 568 */ 569 unsigned int rc_superres_denominator; 570 571 /*!\brief Keyframe super-resolution denominator. 572 * 573 * The denominator for superres to use. If fixed it will only change if the 574 * cumulative scale change over resizing and superres is greater than 1/2; 575 * this forces superres to reduce scaling. 576 * 577 * Valid denominators are 8 - 16 for now. 578 */ 579 unsigned int rc_superres_kf_denominator; 580 581 /*!\brief Frame super-resolution q threshold. 582 * 583 * The q level threshold after which superres is used. 584 * Valid values are 1 to 63. 585 * 586 * Used only by AOM_SUPERRES_QTHRESH 587 */ 588 unsigned int rc_superres_qthresh; 589 590 /*!\brief Keyframe super-resolution q threshold. 591 * 592 * The q level threshold after which superres is used for key frames. 593 * Valid values are 1 to 63. 594 * 595 * Used only by AOM_SUPERRES_QTHRESH 596 */ 597 unsigned int rc_superres_kf_qthresh; 598 599 /*!\brief Rate control algorithm to use. 600 * 601 * Indicates whether the end usage of this stream is to be streamed over 602 * a bandwidth constrained link, indicating that Constant Bit Rate (CBR) 603 * mode should be used, or whether it will be played back on a high 604 * bandwidth link, as from a local disk, where higher variations in 605 * bitrate are acceptable. 606 */ 607 enum aom_rc_mode rc_end_usage; 608 609 /*!\brief Two-pass stats buffer. 610 * 611 * A buffer containing all of the stats packets produced in the first 612 * pass, concatenated. 613 */ 614 aom_fixed_buf_t rc_twopass_stats_in; 615 616 /*!\brief first pass mb stats buffer. 617 * 618 * A buffer containing all of the first pass mb stats packets produced 619 * in the first pass, concatenated. 620 */ 621 aom_fixed_buf_t rc_firstpass_mb_stats_in; 622 623 /*!\brief Target data rate 624 * 625 * Target bitrate to use for this stream, in kilobits per second. 626 */ 627 unsigned int rc_target_bitrate; 628 629 /* 630 * quantizer settings 631 */ 632 633 /*!\brief Minimum (Best Quality) Quantizer 634 * 635 * The quantizer is the most direct control over the quality of the 636 * encoded image. The range of valid values for the quantizer is codec 637 * specific. Consult the documentation for the codec to determine the 638 * values to use. To determine the range programmatically, call 639 * aom_codec_enc_config_default() with a usage value of 0. 640 */ 641 unsigned int rc_min_quantizer; 642 643 /*!\brief Maximum (Worst Quality) Quantizer 644 * 645 * The quantizer is the most direct control over the quality of the 646 * encoded image. The range of valid values for the quantizer is codec 647 * specific. Consult the documentation for the codec to determine the 648 * values to use. To determine the range programmatically, call 649 * aom_codec_enc_config_default() with a usage value of 0. 650 */ 651 unsigned int rc_max_quantizer; 652 653 /* 654 * bitrate tolerance 655 */ 656 657 /*!\brief Rate control adaptation undershoot control 658 * 659 * This value, controls the tolerance of the VBR algorithm to undershoot 660 * and is used as a trigger threshold for more aggressive adaptation of Q. 661 * 662 * Valid values in the range 0-100. 663 */ 664 unsigned int rc_undershoot_pct; 665 666 /*!\brief Rate control adaptation overshoot control 667 * 668 * This value, controls the tolerance of the VBR algorithm to overshoot 669 * and is used as a trigger threshold for more aggressive adaptation of Q. 670 * 671 * Valid values in the range 0-100. 672 */ 673 unsigned int rc_overshoot_pct; 674 675 /* 676 * decoder buffer model parameters 677 */ 678 679 /*!\brief Decoder Buffer Size 680 * 681 * This value indicates the amount of data that may be buffered by the 682 * decoding application. Note that this value is expressed in units of 683 * time (milliseconds). For example, a value of 5000 indicates that the 684 * client will buffer (at least) 5000ms worth of encoded data. Use the 685 * target bitrate (#rc_target_bitrate) to convert to bits/bytes, if 686 * necessary. 687 */ 688 unsigned int rc_buf_sz; 689 690 /*!\brief Decoder Buffer Initial Size 691 * 692 * This value indicates the amount of data that will be buffered by the 693 * decoding application prior to beginning playback. This value is 694 * expressed in units of time (milliseconds). Use the target bitrate 695 * (#rc_target_bitrate) to convert to bits/bytes, if necessary. 696 */ 697 unsigned int rc_buf_initial_sz; 698 699 /*!\brief Decoder Buffer Optimal Size 700 * 701 * This value indicates the amount of data that the encoder should try 702 * to maintain in the decoder's buffer. This value is expressed in units 703 * of time (milliseconds). Use the target bitrate (#rc_target_bitrate) 704 * to convert to bits/bytes, if necessary. 705 */ 706 unsigned int rc_buf_optimal_sz; 707 708 /* 709 * 2 pass rate control parameters 710 */ 711 712 /*!\brief Two-pass mode CBR/VBR bias 713 * 714 * Bias, expressed on a scale of 0 to 100, for determining target size 715 * for the current frame. The value 0 indicates the optimal CBR mode 716 * value should be used. The value 100 indicates the optimal VBR mode 717 * value should be used. Values in between indicate which way the 718 * encoder should "lean." 719 */ 720 unsigned int rc_2pass_vbr_bias_pct; 721 722 /*!\brief Two-pass mode per-GOP minimum bitrate 723 * 724 * This value, expressed as a percentage of the target bitrate, indicates 725 * the minimum bitrate to be used for a single GOP (aka "section") 726 */ 727 unsigned int rc_2pass_vbr_minsection_pct; 728 729 /*!\brief Two-pass mode per-GOP maximum bitrate 730 * 731 * This value, expressed as a percentage of the target bitrate, indicates 732 * the maximum bitrate to be used for a single GOP (aka "section") 733 */ 734 unsigned int rc_2pass_vbr_maxsection_pct; 735 736 /* 737 * keyframing settings (kf) 738 */ 739 740 /*!\brief Option to enable forward reference key frame 741 * 742 */ 743 int fwd_kf_enabled; 744 745 /*!\brief Keyframe placement mode 746 * 747 * This value indicates whether the encoder should place keyframes at a 748 * fixed interval, or determine the optimal placement automatically 749 * (as governed by the #kf_min_dist and #kf_max_dist parameters) 750 */ 751 enum aom_kf_mode kf_mode; 752 753 /*!\brief Keyframe minimum interval 754 * 755 * This value, expressed as a number of frames, prevents the encoder from 756 * placing a keyframe nearer than kf_min_dist to the previous keyframe. At 757 * least kf_min_dist frames non-keyframes will be coded before the next 758 * keyframe. Set kf_min_dist equal to kf_max_dist for a fixed interval. 759 */ 760 unsigned int kf_min_dist; 761 762 /*!\brief Keyframe maximum interval 763 * 764 * This value, expressed as a number of frames, forces the encoder to code 765 * a keyframe if one has not been coded in the last kf_max_dist frames. 766 * A value of 0 implies all frames will be keyframes. Set kf_min_dist 767 * equal to kf_max_dist for a fixed interval. 768 */ 769 unsigned int kf_max_dist; 770 771 /*!\brief sframe interval 772 * 773 * This value, expressed as a number of frames, forces the encoder to code 774 * an S-Frame every sframe_dist frames. 775 */ 776 unsigned int sframe_dist; 777 778 /*!\brief sframe insertion mode 779 * 780 * This value must be set to 1 or 2, and tells the encoder how to insert 781 * S-Frames. It will only have an effect if sframe_dist != 0. 782 * 783 * If altref is enabled: 784 * - if sframe_mode == 1, the considered frame will be made into an 785 * S-Frame only if it is an altref frame 786 * - if sframe_mode == 2, the next altref frame will be made into an 787 * S-Frame. 788 * 789 * Otherwise: the considered frame will be made into an S-Frame. 790 */ 791 unsigned int sframe_mode; 792 793 /*!\brief Tile coding mode 794 * 795 * This value indicates the tile coding mode. 796 * A value of 0 implies a normal non-large-scale tile coding. A value of 1 797 * implies a large-scale tile coding. 798 */ 799 unsigned int large_scale_tile; 800 801 /*!\brief Monochrome mode 802 * 803 * If this is nonzero, the encoder will generate a monochrome stream 804 * with no chroma planes. 805 */ 806 unsigned int monochrome; 807 808 /*!\brief full_still_picture_hdr 809 * 810 * If this is nonzero, the encoder will generate a full header even for 811 * still picture encoding. if zero, a reduced header is used for still 812 * picture. This flag has no effect when a regular video with more than 813 * a single frame is encoded. 814 */ 815 unsigned int full_still_picture_hdr; 816 817 /*!\brief Bitstream syntax mode 818 * 819 * This value indicates the bitstream syntax mode. 820 * A value of 0 indicates bitstream is saved as Section 5 bitstream. A value 821 * of 1 indicates the bitstream is saved in Annex-B format 822 */ 823 unsigned int save_as_annexb; 824 825 /*!\brief Number of explicit tile widths specified 826 * 827 * This value indicates the number of tile widths specified 828 * A value of 0 implies no tile widths are specified. 829 * Tile widths are given in the array tile_widths[] 830 */ 831 int tile_width_count; 832 833 /*!\brief Number of explicit tile heights specified 834 * 835 * This value indicates the number of tile heights specified 836 * A value of 0 implies no tile heights are specified. 837 * Tile heights are given in the array tile_heights[] 838 */ 839 int tile_height_count; 840 841 /*!\brief Maximum number of tile widths in tile widths array 842 * 843 * This define gives the maximum number of elements in the tile_widths array. 844 */ 845 #define MAX_TILE_WIDTHS 64 // maximum tile width array length 846 847 /*!\brief Array of specified tile widths 848 * 849 * This array specifies tile widths (and may be empty) 850 * The number of widths specified is given by tile_width_count 851 */ 852 int tile_widths[MAX_TILE_WIDTHS]; 853 854 /*!\brief Maximum number of tile heights in tile heights array. 855 * 856 * This define gives the maximum number of elements in the tile_heights array. 857 */ 858 #define MAX_TILE_HEIGHTS 64 // maximum tile height array length 859 860 /*!\brief Array of specified tile heights 861 * 862 * This array specifies tile heights (and may be empty) 863 * The number of heights specified is given by tile_height_count 864 */ 865 int tile_heights[MAX_TILE_HEIGHTS]; 866 867 /*!\brief Whether encoder should use fixed QP offsets. 868 * 869 * If a value of 1 is provided, encoder will use fixed QP offsets for frames 870 * at different levels of the pyramid. 871 * - If 'fixed_qp_offsets' is also provided, encoder will use the given 872 * offsets 873 * - If not, encoder will select the fixed offsets based on the cq-level 874 * provided. 875 * If a value of 0 is provided and fixed_qp_offset are not provided, encoder 876 * will NOT use fixed QP offsets. 877 * Note: This option is only relevant for --end-usage=q. 878 */ 879 unsigned int use_fixed_qp_offsets; 880 881 /*!\brief Number of fixed QP offsets 882 * 883 * This defines the number of elements in the fixed_qp_offsets array. 884 */ 885 #define FIXED_QP_OFFSET_COUNT 5 886 887 /*!\brief Array of fixed QP offsets 888 * 889 * This array specifies fixed QP offsets (range: 0 to 63) for frames at 890 * different levels of the pyramid. It is a comma-separated list of 5 values: 891 * - QP offset for keyframe 892 * - QP offset for ALTREF frame 893 * - QP offset for 1st level internal ARF 894 * - QP offset for 2nd level internal ARF 895 * - QP offset for 3rd level internal ARF 896 * Notes: 897 * - QP offset for leaf level frames is not explicitly specified. These frames 898 * use the worst quality allowed (--cq-level). 899 * - This option is only relevant for --end-usage=q. 900 */ 901 int fixed_qp_offsets[FIXED_QP_OFFSET_COUNT]; 902 903 /*!\brief Options defined per config file 904 * 905 */ 906 cfg_options_t encoder_cfg; 907 } aom_codec_enc_cfg_t; /**< alias for struct aom_codec_enc_cfg */ 908 909 /*!\brief Initialize an encoder instance 910 * 911 * Initializes a encoder context using the given interface. Applications 912 * should call the aom_codec_enc_init convenience macro instead of this 913 * function directly, to ensure that the ABI version number parameter 914 * is properly initialized. 915 * 916 * If the library was configured with -DCONFIG_MULTITHREAD=0, this call 917 * is not thread safe and should be guarded with a lock if being used 918 * in a multithreaded context. 919 * 920 * \param[in] ctx Pointer to this instance's context. 921 * \param[in] iface Pointer to the algorithm interface to use. 922 * \param[in] cfg Configuration to use, if known. 923 * \param[in] flags Bitfield of AOM_CODEC_USE_* flags 924 * \param[in] ver ABI version number. Must be set to 925 * AOM_ENCODER_ABI_VERSION 926 * \retval #AOM_CODEC_OK 927 * The decoder algorithm initialized. 928 * \retval #AOM_CODEC_MEM_ERROR 929 * Memory allocation failed. 930 */ 931 aom_codec_err_t aom_codec_enc_init_ver(aom_codec_ctx_t *ctx, 932 aom_codec_iface_t *iface, 933 const aom_codec_enc_cfg_t *cfg, 934 aom_codec_flags_t flags, int ver); 935 936 /*!\brief Convenience macro for aom_codec_enc_init_ver() 937 * 938 * Ensures the ABI version parameter is properly set. 939 */ 940 #define aom_codec_enc_init(ctx, iface, cfg, flags) \ 941 aom_codec_enc_init_ver(ctx, iface, cfg, flags, AOM_ENCODER_ABI_VERSION) 942 943 /*!\brief Get the default configuration for a usage. 944 * 945 * Initializes an encoder configuration structure with default values. Supports 946 * the notion of "usages" so that an algorithm may offer different default 947 * settings depending on the user's intended goal. This function \ref SHOULD 948 * be called by all applications to initialize the configuration structure 949 * before specializing the configuration with application specific values. 950 * 951 * \param[in] iface Pointer to the algorithm interface to use. 952 * \param[out] cfg Configuration buffer to populate. 953 * \param[in] usage Algorithm specific usage value. For AV1, must be 954 * set to AOM_USAGE_GOOD_QUALITY (0), 955 * AOM_USAGE_REALTIME (1), or AOM_USAGE_ALL_INTRA (2). 956 * 957 * \retval #AOM_CODEC_OK 958 * The configuration was populated. 959 * \retval #AOM_CODEC_INCAPABLE 960 * Interface is not an encoder interface. 961 * \retval #AOM_CODEC_INVALID_PARAM 962 * A parameter was NULL, or the usage value was not recognized. 963 */ 964 aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface, 965 aom_codec_enc_cfg_t *cfg, 966 unsigned int usage); 967 968 /*!\brief Set or change configuration 969 * 970 * Reconfigures an encoder instance according to the given configuration. 971 * 972 * \param[in] ctx Pointer to this instance's context 973 * \param[in] cfg Configuration buffer to use 974 * 975 * \retval #AOM_CODEC_OK 976 * The configuration was populated. 977 * \retval #AOM_CODEC_INCAPABLE 978 * Interface is not an encoder interface. 979 * \retval #AOM_CODEC_INVALID_PARAM 980 * A parameter was NULL, or the usage value was not recognized. 981 */ 982 aom_codec_err_t aom_codec_enc_config_set(aom_codec_ctx_t *ctx, 983 const aom_codec_enc_cfg_t *cfg); 984 985 /*!\brief Get global stream headers 986 * 987 * Retrieves a stream level global header packet, if supported by the codec. 988 * Calls to this function should be deferred until all configuration information 989 * has been passed to libaom. Otherwise the global header data may be 990 * invalidated by additional configuration changes. 991 * 992 * The AV1 implementation of this function returns an OBU. The OBU returned is 993 * in Low Overhead Bitstream Format. Specifically, the obu_has_size_field bit is 994 * set, and the buffer contains the obu_size field for the returned OBU. 995 * 996 * \param[in] ctx Pointer to this instance's context 997 * 998 * \retval NULL 999 * Encoder does not support global header, or an error occurred while 1000 * generating the global header. 1001 * 1002 * \retval Non-NULL 1003 * Pointer to buffer containing global header packet. The caller owns the 1004 * memory associated with this buffer, and must free the 'buf' member of the 1005 * aom_fixed_buf_t as well as the aom_fixed_buf_t pointer. Memory returned 1006 * must be freed via call to free(). 1007 */ 1008 aom_fixed_buf_t *aom_codec_get_global_headers(aom_codec_ctx_t *ctx); 1009 1010 /*!\brief usage parameter analogous to AV1 GOOD QUALITY mode. */ 1011 #define AOM_USAGE_GOOD_QUALITY (0) 1012 /*!\brief usage parameter analogous to AV1 REALTIME mode. */ 1013 #define AOM_USAGE_REALTIME (1) 1014 /*!\brief usage parameter analogous to AV1 all intra mode. */ 1015 #define AOM_USAGE_ALL_INTRA (2) 1016 1017 /*!\brief Encode a frame 1018 * 1019 * Encodes a video frame at the given "presentation time." The presentation 1020 * time stamp (PTS) \ref MUST be strictly increasing. 1021 * 1022 * When the last frame has been passed to the encoder, this function should 1023 * continue to be called in a loop, with the img parameter set to NULL. This 1024 * will signal the end-of-stream condition to the encoder and allow it to 1025 * encode any held buffers. Encoding is complete when aom_codec_encode() is 1026 * called with img set to NULL and aom_codec_get_cx_data() returns no data. 1027 * 1028 * \param[in] ctx Pointer to this instance's context 1029 * \param[in] img Image data to encode, NULL to flush. 1030 * \param[in] pts Presentation time stamp, in timebase units. If img 1031 * is NULL, pts is ignored. 1032 * \param[in] duration Duration to show frame, in timebase units. If img 1033 * is not NULL, duration must be nonzero. If img is 1034 * NULL, duration is ignored. 1035 * \param[in] flags Flags to use for encoding this frame. 1036 * 1037 * \retval #AOM_CODEC_OK 1038 * The configuration was populated. 1039 * \retval #AOM_CODEC_INCAPABLE 1040 * Interface is not an encoder interface. 1041 * \retval #AOM_CODEC_INVALID_PARAM 1042 * A parameter was NULL, the image format is unsupported, etc. 1043 */ 1044 aom_codec_err_t aom_codec_encode(aom_codec_ctx_t *ctx, const aom_image_t *img, 1045 aom_codec_pts_t pts, unsigned long duration, 1046 aom_enc_frame_flags_t flags); 1047 1048 /*!\brief Set compressed data output buffer 1049 * 1050 * Sets the buffer that the codec should output the compressed data 1051 * into. This call effectively sets the buffer pointer returned in the 1052 * next AOM_CODEC_CX_FRAME_PKT packet. Subsequent packets will be 1053 * appended into this buffer. The buffer is preserved across frames, 1054 * so applications must periodically call this function after flushing 1055 * the accumulated compressed data to disk or to the network to reset 1056 * the pointer to the buffer's head. 1057 * 1058 * `pad_before` bytes will be skipped before writing the compressed 1059 * data, and `pad_after` bytes will be appended to the packet. The size 1060 * of the packet will be the sum of the size of the actual compressed 1061 * data, pad_before, and pad_after. The padding bytes will be preserved 1062 * (not overwritten). 1063 * 1064 * Note that calling this function does not guarantee that the returned 1065 * compressed data will be placed into the specified buffer. In the 1066 * event that the encoded data will not fit into the buffer provided, 1067 * the returned packet \ref MAY point to an internal buffer, as it would 1068 * if this call were never used. In this event, the output packet will 1069 * NOT have any padding, and the application must free space and copy it 1070 * to the proper place. This is of particular note in configurations 1071 * that may output multiple packets for a single encoded frame (e.g., lagged 1072 * encoding) or if the application does not reset the buffer periodically. 1073 * 1074 * Applications may restore the default behavior of the codec providing 1075 * the compressed data buffer by calling this function with a NULL 1076 * buffer. 1077 * 1078 * Applications \ref MUSTNOT call this function during iteration of 1079 * aom_codec_get_cx_data(). 1080 * 1081 * \param[in] ctx Pointer to this instance's context 1082 * \param[in] buf Buffer to store compressed data into 1083 * \param[in] pad_before Bytes to skip before writing compressed data 1084 * \param[in] pad_after Bytes to skip after writing compressed data 1085 * 1086 * \retval #AOM_CODEC_OK 1087 * The buffer was set successfully. 1088 * \retval #AOM_CODEC_INVALID_PARAM 1089 * A parameter was NULL, the image format is unsupported, etc. 1090 */ 1091 aom_codec_err_t aom_codec_set_cx_data_buf(aom_codec_ctx_t *ctx, 1092 const aom_fixed_buf_t *buf, 1093 unsigned int pad_before, 1094 unsigned int pad_after); 1095 1096 /*!\brief Encoded data iterator 1097 * 1098 * Iterates over a list of data packets to be passed from the encoder to the 1099 * application. The different kinds of packets available are enumerated in 1100 * #aom_codec_cx_pkt_kind. 1101 * 1102 * #AOM_CODEC_CX_FRAME_PKT packets should be passed to the application's 1103 * muxer. Multiple compressed frames may be in the list. 1104 * #AOM_CODEC_STATS_PKT packets should be appended to a global buffer. 1105 * 1106 * The application \ref MUST silently ignore any packet kinds that it does 1107 * not recognize or support. 1108 * 1109 * The data buffers returned from this function are only guaranteed to be 1110 * valid until the application makes another call to any aom_codec_* function. 1111 * 1112 * \param[in] ctx Pointer to this instance's context 1113 * \param[in,out] iter Iterator storage, initialized to NULL 1114 * 1115 * \return Returns a pointer to an output data packet (compressed frame data, 1116 * two-pass statistics, etc.) or NULL to signal end-of-list. 1117 * 1118 */ 1119 const aom_codec_cx_pkt_t *aom_codec_get_cx_data(aom_codec_ctx_t *ctx, 1120 aom_codec_iter_t *iter); 1121 1122 /*!\brief Get Preview Frame 1123 * 1124 * Returns an image that can be used as a preview. Shows the image as it would 1125 * exist at the decompressor. The application \ref MUST NOT write into this 1126 * image buffer. 1127 * 1128 * \param[in] ctx Pointer to this instance's context 1129 * 1130 * \return Returns a pointer to a preview image, or NULL if no image is 1131 * available. 1132 * 1133 */ 1134 const aom_image_t *aom_codec_get_preview_frame(aom_codec_ctx_t *ctx); 1135 1136 /*!@} - end defgroup encoder*/ 1137 #ifdef __cplusplus 1138 } 1139 #endif 1140 #endif // AOM_AOM_AOM_ENCODER_H_ 1141