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 35 /*!\brief Current ABI version number 36 * 37 * \internal 38 * If this file is altered in any way that changes the ABI, this value 39 * must be bumped. Examples include, but are not limited to, changing 40 * types, removing or reassigning enums, adding/removing/rearranging 41 * fields to structures 42 */ 43 #define AOM_ENCODER_ABI_VERSION \ 44 (8 + AOM_CODEC_ABI_VERSION) /**<\hideinitializer*/ 45 46 /*! \brief Encoder capabilities bitfield 47 * 48 * Each encoder advertises the capabilities it supports as part of its 49 * ::aom_codec_iface_t interface structure. Capabilities are extra 50 * interfaces or functionality, and are not required to be supported 51 * by an encoder. 52 * 53 * The available flags are specified by AOM_CODEC_CAP_* defines. 54 */ 55 #define AOM_CODEC_CAP_PSNR 0x10000 /**< Can issue PSNR packets */ 56 57 /*! Can support input images at greater than 8 bitdepth. 58 */ 59 #define AOM_CODEC_CAP_HIGHBITDEPTH 0x40000 60 61 /*! \brief Initialization-time Feature Enabling 62 * 63 * Certain codec features must be known at initialization time, to allow 64 * for proper memory allocation. 65 * 66 * The available flags are specified by AOM_CODEC_USE_* defines. 67 */ 68 #define AOM_CODEC_USE_PSNR 0x10000 /**< Calculate PSNR on each frame */ 69 /*!\brief Make the encoder output one partition at a time. */ 70 #define AOM_CODEC_USE_HIGHBITDEPTH 0x40000 /**< Use high bitdepth */ 71 72 /*!\brief Generic fixed size buffer structure 73 * 74 * This structure is able to hold a reference to any fixed size buffer. 75 */ 76 typedef struct aom_fixed_buf { 77 void *buf; /**< Pointer to the data. Does NOT own the data! */ 78 size_t sz; /**< Length of the buffer, in chars */ 79 } aom_fixed_buf_t; /**< alias for struct aom_fixed_buf */ 80 81 /*!\brief Compressed Frame Flags 82 * 83 * This type represents a bitfield containing information about a compressed 84 * frame that may be useful to an application. The most significant 16 bits 85 * can be used by an algorithm to provide additional detail, for example to 86 * support frame types that are codec specific (MPEG-1 D-frames for example) 87 */ 88 typedef uint32_t aom_codec_frame_flags_t; 89 #define AOM_FRAME_IS_KEY 0x1 /**< frame is the start of a GOP */ 90 /*!\brief frame can be dropped without affecting the stream (no future frame 91 * depends on this one) */ 92 #define AOM_FRAME_IS_DROPPABLE 0x2 93 /*!\brief this is an INTRA_ONLY frame */ 94 #define AOM_FRAME_IS_INTRAONLY 0x10 95 /*!\brief this is an S-frame */ 96 #define AOM_FRAME_IS_SWITCH 0x20 97 /*!\brief this is an error-resilient frame */ 98 #define AOM_FRAME_IS_ERROR_RESILIENT 0x40 99 /*!\brief this is a key-frame dependent recovery-point frame */ 100 #define AOM_FRAME_IS_DELAYED_RANDOM_ACCESS_POINT 0x80 101 102 /*!\brief Error Resilient flags 103 * 104 * These flags define which error resilient features to enable in the 105 * encoder. The flags are specified through the 106 * aom_codec_enc_cfg::g_error_resilient variable. 107 */ 108 typedef uint32_t aom_codec_er_flags_t; 109 /*!\brief Improve resiliency against losses of whole frames */ 110 #define AOM_ERROR_RESILIENT_DEFAULT 0x1 111 112 /*!\brief Encoder output packet variants 113 * 114 * This enumeration lists the different kinds of data packets that can be 115 * returned by calls to aom_codec_get_cx_data(). Algorithms \ref MAY 116 * extend this list to provide additional functionality. 117 */ 118 enum aom_codec_cx_pkt_kind { 119 AOM_CODEC_CX_FRAME_PKT, /**< Compressed video frame */ 120 AOM_CODEC_STATS_PKT, /**< Two-pass statistics for this frame */ 121 AOM_CODEC_FPMB_STATS_PKT, /**< first pass mb statistics for this frame */ 122 AOM_CODEC_PSNR_PKT, /**< PSNR statistics for this frame */ 123 AOM_CODEC_CUSTOM_PKT = 256 /**< Algorithm extensions */ 124 }; 125 126 /*!\brief Encoder output packet 127 * 128 * This structure contains the different kinds of output data the encoder 129 * may produce while compressing a frame. 130 */ 131 typedef struct aom_codec_cx_pkt { 132 enum aom_codec_cx_pkt_kind kind; /**< packet variant */ 133 union { 134 struct { 135 void *buf; /**< compressed data buffer */ 136 size_t sz; /**< length of compressed data */ 137 /*!\brief time stamp to show frame (in timebase units) */ 138 aom_codec_pts_t pts; 139 /*!\brief duration to show frame (in timebase units) */ 140 unsigned long duration; 141 aom_codec_frame_flags_t flags; /**< flags for this frame */ 142 /*!\brief the partition id defines the decoding order of the partitions. 143 * Only applicable when "output partition" mode is enabled. First 144 * partition has id 0.*/ 145 int partition_id; 146 /*!\brief size of the visible frame in this packet */ 147 size_t vis_frame_size; 148 } frame; /**< data for compressed frame packet */ 149 aom_fixed_buf_t twopass_stats; /**< data for two-pass packet */ 150 aom_fixed_buf_t firstpass_mb_stats; /**< first pass mb packet */ 151 struct aom_psnr_pkt { 152 unsigned int samples[4]; /**< Number of samples, total/y/u/v */ 153 uint64_t sse[4]; /**< sum squared error, total/y/u/v */ 154 double psnr[4]; /**< PSNR, total/y/u/v */ 155 } psnr; /**< data for PSNR packet */ 156 aom_fixed_buf_t raw; /**< data for arbitrary packets */ 157 158 /* This packet size is fixed to allow codecs to extend this 159 * interface without having to manage storage for raw packets, 160 * i.e., if it's smaller than 128 bytes, you can store in the 161 * packet list directly. 162 */ 163 char pad[128 - sizeof(enum aom_codec_cx_pkt_kind)]; /**< fixed sz */ 164 } data; /**< packet data */ 165 } aom_codec_cx_pkt_t; /**< alias for struct aom_codec_cx_pkt */ 166 167 /*!\brief Rational Number 168 * 169 * This structure holds a fractional value. 170 */ 171 typedef struct aom_rational { 172 int num; /**< fraction numerator */ 173 int den; /**< fraction denominator */ 174 } aom_rational_t; /**< alias for struct aom_rational */ 175 176 /*!\brief Multi-pass Encoding Pass */ 177 enum aom_enc_pass { 178 AOM_RC_ONE_PASS, /**< Single pass mode */ 179 AOM_RC_FIRST_PASS, /**< First pass of multi-pass mode */ 180 AOM_RC_LAST_PASS /**< Final pass of multi-pass mode */ 181 }; 182 183 /*!\brief Rate control mode */ 184 enum aom_rc_mode { 185 AOM_VBR, /**< Variable Bit Rate (VBR) mode */ 186 AOM_CBR, /**< Constant Bit Rate (CBR) mode */ 187 AOM_CQ, /**< Constrained Quality (CQ) mode */ 188 AOM_Q, /**< Constant Quality (Q) mode */ 189 }; 190 191 /*!\brief Keyframe placement mode. 192 * 193 * This enumeration determines whether keyframes are placed automatically by 194 * the encoder or whether this behavior is disabled. Older releases of this 195 * SDK were implemented such that AOM_KF_FIXED meant keyframes were disabled. 196 * This name is confusing for this behavior, so the new symbols to be used 197 * are AOM_KF_AUTO and AOM_KF_DISABLED. 198 */ 199 enum aom_kf_mode { 200 AOM_KF_FIXED, /**< deprecated, implies AOM_KF_DISABLED */ 201 AOM_KF_AUTO, /**< Encoder determines optimal placement automatically */ 202 AOM_KF_DISABLED = 0 /**< Encoder does not place keyframes. */ 203 }; 204 205 /*!\brief Encoder Config Options 206 * 207 * This type allows to enumerate and control flags defined for encoder control 208 * via config file at runtime. 209 */ 210 typedef struct cfg_options { 211 /*!\brief Indicate init by cfg file 212 * 0 or 1 213 */ 214 unsigned int init_by_cfg_file; 215 /*!\brief Superblock size 216 * 0, 64 or 128 217 */ 218 unsigned int super_block_size; 219 /*!\brief max partition size 220 * 8, 16, 32, 64, 128 221 */ 222 unsigned int max_partition_size; 223 /*!\brief min partition size 224 * 8, 16, 32, 64, 128 225 */ 226 unsigned int min_partition_size; 227 /*!\brief disable AB Shape partition type 228 * 229 */ 230 unsigned int disable_ab_partition_type; 231 /*!\brief disable rectangular partition type 232 * 233 */ 234 unsigned int disable_rect_partition_type; 235 /*!\brief disable 1:4/4:1 partition type 236 * 237 */ 238 unsigned int disable_1to4_partition_type; 239 /*!\brief disable flip and identity transform type 240 * 241 */ 242 unsigned int disable_flip_idtx; 243 /*!\brief disable CDEF filter 244 * 245 */ 246 unsigned int disable_cdef; 247 /*!\brief disable Loop Restoration Filter 248 * 249 */ 250 unsigned int disable_lr; 251 /*!\brief disable OBMC 252 * 253 */ 254 unsigned int disable_obmc; 255 /*!\brief disable Warped Motion 256 * 257 */ 258 unsigned int disable_warp_motion; 259 /*!\brief disable global motion 260 * 261 */ 262 unsigned int disable_global_motion; 263 /*!\brief disable dist weighted compound 264 * 265 */ 266 unsigned int disable_dist_wtd_comp; 267 /*!\brief disable diff weighted compound 268 * 269 */ 270 unsigned int disable_diff_wtd_comp; 271 /*!\brief disable inter/intra compound 272 * 273 */ 274 unsigned int disable_inter_intra_comp; 275 /*!\brief disable masked compound 276 * 277 */ 278 unsigned int disable_masked_comp; 279 /*!\brief disable one sided compound 280 * 281 */ 282 unsigned int disable_one_sided_comp; 283 /*!\brief disable Palette 284 * 285 */ 286 unsigned int disable_palette; 287 /*!\brief disable Intra Block Copy 288 * 289 */ 290 unsigned int disable_intrabc; 291 /*!\brief disable chroma from luma 292 * 293 */ 294 unsigned int disable_cfl; 295 /*!\brief disable intra smooth mode 296 * 297 */ 298 unsigned int disable_smooth_intra; 299 /*!\brief disable filter intra 300 * 301 */ 302 unsigned int disable_filter_intra; 303 /*!\brief disable dual filter 304 * 305 */ 306 unsigned int disable_dual_filter; 307 /*!\brief disable intra angle delta 308 * 309 */ 310 unsigned int disable_intra_angle_delta; 311 /*!\brief disable intra edge filter 312 * 313 */ 314 unsigned int disable_intra_edge_filter; 315 /*!\brief disable 64x64 transform 316 * 317 */ 318 unsigned int disable_tx_64x64; 319 /*!\brief disable smooth inter/intra 320 * 321 */ 322 unsigned int disable_smooth_inter_intra; 323 /*!\brief disable inter/inter wedge comp 324 * 325 */ 326 unsigned int disable_inter_inter_wedge; 327 /*!\brief disable inter/intra wedge comp 328 * 329 */ 330 unsigned int disable_inter_intra_wedge; 331 /*!\brief disable paeth intra 332 * 333 */ 334 unsigned int disable_paeth_intra; 335 /*!\brief disable trellis quantization 336 * 337 */ 338 unsigned int disable_trellis_quant; 339 /*!\brief disable ref frame MV 340 * 341 */ 342 unsigned int disable_ref_frame_mv; 343 /*!\brief use reduced reference frame set 344 * 345 */ 346 unsigned int reduced_reference_set; 347 /*!\brief use reduced transform type set 348 * 349 */ 350 unsigned int reduced_tx_type_set; 351 } cfg_options_t; 352 353 /*!\brief Encoded Frame Flags 354 * 355 * This type indicates a bitfield to be passed to aom_codec_encode(), defining 356 * per-frame boolean values. By convention, bits common to all codecs will be 357 * named AOM_EFLAG_*, and bits specific to an algorithm will be named 358 * /algo/_eflag_*. The lower order 16 bits are reserved for common use. 359 */ 360 typedef long aom_enc_frame_flags_t; 361 #define AOM_EFLAG_FORCE_KF (1 << 0) /**< Force this frame to be a keyframe */ 362 363 /*!\brief Encoder configuration structure 364 * 365 * This structure contains the encoder settings that have common representations 366 * across all codecs. This doesn't imply that all codecs support all features, 367 * however. 368 */ 369 typedef struct aom_codec_enc_cfg { 370 /* 371 * generic settings (g) 372 */ 373 374 /*!\brief Algorithm specific "usage" value 375 * 376 * Algorithms may define multiple values for usage, which may convey the 377 * intent of how the application intends to use the stream. If this value 378 * is non-zero, consult the documentation for the codec to determine its 379 * meaning. 380 */ 381 unsigned int g_usage; 382 383 /*!\brief Maximum number of threads to use 384 * 385 * For multi-threaded implementations, use no more than this number of 386 * threads. The codec may use fewer threads than allowed. The value 387 * 0 is equivalent to the value 1. 388 */ 389 unsigned int g_threads; 390 391 /*!\brief Bitstream profile to use 392 * 393 * Some codecs support a notion of multiple bitstream profiles. Typically 394 * this maps to a set of features that are turned on or off. Often the 395 * profile to use is determined by the features of the intended decoder. 396 * Consult the documentation for the codec to determine the valid values 397 * for this parameter, or set to zero for a sane default. 398 */ 399 unsigned int g_profile; /**< profile of bitstream to use */ 400 401 /*!\brief Width of the frame 402 * 403 * This value identifies the presentation resolution of the frame, 404 * in pixels. Note that the frames passed as input to the encoder must 405 * have this resolution. Frames will be presented by the decoder in this 406 * resolution, independent of any spatial resampling the encoder may do. 407 */ 408 unsigned int g_w; 409 410 /*!\brief Height of the frame 411 * 412 * This value identifies the presentation resolution of the frame, 413 * in pixels. Note that the frames passed as input to the encoder must 414 * have this resolution. Frames will be presented by the decoder in this 415 * resolution, independent of any spatial resampling the encoder may do. 416 */ 417 unsigned int g_h; 418 419 /*!\brief Max number of frames to encode 420 * 421 */ 422 unsigned int g_limit; 423 424 /*!\brief Forced maximum width of the frame 425 * 426 * If this value is non-zero then it is used to force the maximum frame 427 * width written in write_sequence_header(). 428 */ 429 unsigned int g_forced_max_frame_width; 430 431 /*!\brief Forced maximum height of the frame 432 * 433 * If this value is non-zero then it is used to force the maximum frame 434 * height written in write_sequence_header(). 435 */ 436 unsigned int g_forced_max_frame_height; 437 438 /*!\brief Bit-depth of the codec 439 * 440 * This value identifies the bit_depth of the codec, 441 * Only certain bit-depths are supported as identified in the 442 * aom_bit_depth_t enum. 443 */ 444 aom_bit_depth_t g_bit_depth; 445 446 /*!\brief Bit-depth of the input frames 447 * 448 * This value identifies the bit_depth of the input frames in bits. 449 * Note that the frames passed as input to the encoder must have 450 * this bit-depth. 451 */ 452 unsigned int g_input_bit_depth; 453 454 /*!\brief Stream timebase units 455 * 456 * Indicates the smallest interval of time, in seconds, used by the stream. 457 * For fixed frame rate material, or variable frame rate material where 458 * frames are timed at a multiple of a given clock (ex: video capture), 459 * the \ref RECOMMENDED method is to set the timebase to the reciprocal 460 * of the frame rate (ex: 1001/30000 for 29.970 Hz NTSC). This allows the 461 * pts to correspond to the frame number, which can be handy. For 462 * re-encoding video from containers with absolute time timestamps, the 463 * \ref RECOMMENDED method is to set the timebase to that of the parent 464 * container or multimedia framework (ex: 1/1000 for ms, as in FLV). 465 */ 466 struct aom_rational g_timebase; 467 468 /*!\brief Enable error resilient modes. 469 * 470 * The error resilient bitfield indicates to the encoder which features 471 * it should enable to take measures for streaming over lossy or noisy 472 * links. 473 */ 474 aom_codec_er_flags_t g_error_resilient; 475 476 /*!\brief Multi-pass Encoding Mode 477 * 478 * This value should be set to the current phase for multi-pass encoding. 479 * For single pass, set to #AOM_RC_ONE_PASS. 480 */ 481 enum aom_enc_pass g_pass; 482 483 /*!\brief Allow lagged encoding 484 * 485 * If set, this value allows the encoder to consume a number of input 486 * frames before producing output frames. This allows the encoder to 487 * base decisions for the current frame on future frames. This does 488 * increase the latency of the encoding pipeline, so it is not appropriate 489 * in all situations (ex: realtime encoding). 490 * 491 * Note that this is a maximum value -- the encoder may produce frames 492 * sooner than the given limit. Set this value to 0 to disable this 493 * feature. 494 */ 495 unsigned int g_lag_in_frames; 496 497 /* 498 * rate control settings (rc) 499 */ 500 501 /*!\brief Temporal resampling configuration, if supported by the codec. 502 * 503 * Temporal resampling allows the codec to "drop" frames as a strategy to 504 * meet its target data rate. This can cause temporal discontinuities in 505 * the encoded video, which may appear as stuttering during playback. This 506 * trade-off is often acceptable, but for many applications is not. It can 507 * be disabled in these cases. 508 * 509 * Note that not all codecs support this feature. All aom AVx codecs do. 510 * For other codecs, consult the documentation for that algorithm. 511 * 512 * This threshold is described as a percentage of the target data buffer. 513 * When the data buffer falls below this percentage of fullness, a 514 * dropped frame is indicated. Set the threshold to zero (0) to disable 515 * this feature. 516 */ 517 unsigned int rc_dropframe_thresh; 518 519 /*!\brief Mode for spatial resampling, if supported by the codec. 520 * 521 * Spatial resampling allows the codec to compress a lower resolution 522 * version of the frame, which is then upscaled by the decoder to the 523 * correct presentation resolution. This increases visual quality at 524 * low data rates, at the expense of CPU time on the encoder/decoder. 525 */ 526 unsigned int rc_resize_mode; 527 528 /*!\brief Frame resize denominator. 529 * 530 * The denominator for resize to use, assuming 8 as the numerator. 531 * 532 * Valid denominators are 8 - 16 for now. 533 */ 534 unsigned int rc_resize_denominator; 535 536 /*!\brief Keyframe resize denominator. 537 * 538 * The denominator for resize to use, assuming 8 as the numerator. 539 * 540 * Valid denominators are 8 - 16 for now. 541 */ 542 unsigned int rc_resize_kf_denominator; 543 544 /*!\brief Frame super-resolution scaling mode. 545 * 546 * Similar to spatial resampling, frame super-resolution integrates 547 * upscaling after the encode/decode process. Taking control of upscaling and 548 * using restoration filters should allow it to outperform normal resizing. 549 * 550 * Valid values are 0 to 4 as defined in enum SUPERRES_MODE. 551 */ 552 unsigned int rc_superres_mode; 553 554 /*!\brief Frame super-resolution denominator. 555 * 556 * The denominator for superres to use. If fixed it will only change if the 557 * cumulative scale change over resizing and superres is greater than 1/2; 558 * this forces superres to reduce scaling. 559 * 560 * Valid denominators are 8 to 16. 561 * 562 * Used only by SUPERRES_FIXED. 563 */ 564 unsigned int rc_superres_denominator; 565 566 /*!\brief Keyframe super-resolution denominator. 567 * 568 * The denominator for superres to use. If fixed it will only change if the 569 * cumulative scale change over resizing and superres is greater than 1/2; 570 * this forces superres to reduce scaling. 571 * 572 * Valid denominators are 8 - 16 for now. 573 */ 574 unsigned int rc_superres_kf_denominator; 575 576 /*!\brief Frame super-resolution q threshold. 577 * 578 * The q level threshold after which superres is used. 579 * Valid values are 1 to 63. 580 * 581 * Used only by SUPERRES_QTHRESH 582 */ 583 unsigned int rc_superres_qthresh; 584 585 /*!\brief Keyframe super-resolution q threshold. 586 * 587 * The q level threshold after which superres is used for key frames. 588 * Valid values are 1 to 63. 589 * 590 * Used only by SUPERRES_QTHRESH 591 */ 592 unsigned int rc_superres_kf_qthresh; 593 594 /*!\brief Rate control algorithm to use. 595 * 596 * Indicates whether the end usage of this stream is to be streamed over 597 * a bandwidth constrained link, indicating that Constant Bit Rate (CBR) 598 * mode should be used, or whether it will be played back on a high 599 * bandwidth link, as from a local disk, where higher variations in 600 * bitrate are acceptable. 601 */ 602 enum aom_rc_mode rc_end_usage; 603 604 /*!\brief Two-pass stats buffer. 605 * 606 * A buffer containing all of the stats packets produced in the first 607 * pass, concatenated. 608 */ 609 aom_fixed_buf_t rc_twopass_stats_in; 610 611 /*!\brief first pass mb stats buffer. 612 * 613 * A buffer containing all of the first pass mb stats packets produced 614 * in the first pass, concatenated. 615 */ 616 aom_fixed_buf_t rc_firstpass_mb_stats_in; 617 618 /*!\brief Target data rate 619 * 620 * Target bandwidth to use for this stream, in kilobits per second. 621 */ 622 unsigned int rc_target_bitrate; 623 624 /* 625 * quantizer settings 626 */ 627 628 /*!\brief Minimum (Best Quality) Quantizer 629 * 630 * The quantizer is the most direct control over the quality of the 631 * encoded image. The range of valid values for the quantizer is codec 632 * specific. Consult the documentation for the codec to determine the 633 * values to use. To determine the range programmatically, call 634 * aom_codec_enc_config_default() with a usage value of 0. 635 */ 636 unsigned int rc_min_quantizer; 637 638 /*!\brief Maximum (Worst Quality) Quantizer 639 * 640 * The quantizer is the most direct control over the quality of the 641 * encoded image. The range of valid values for the quantizer is codec 642 * specific. Consult the documentation for the codec to determine the 643 * values to use. To determine the range programmatically, call 644 * aom_codec_enc_config_default() with a usage value of 0. 645 */ 646 unsigned int rc_max_quantizer; 647 648 /* 649 * bitrate tolerance 650 */ 651 652 /*!\brief Rate control adaptation undershoot control 653 * 654 * This value, expressed as a percentage of the target bitrate, 655 * controls the maximum allowed adaptation speed of the codec. 656 * This factor controls the maximum amount of bits that can 657 * be subtracted from the target bitrate in order to compensate 658 * for prior overshoot. 659 * 660 * Valid values in the range 0-1000. 661 */ 662 unsigned int rc_undershoot_pct; 663 664 /*!\brief Rate control adaptation overshoot control 665 * 666 * This value, expressed as a percentage of the target bitrate, 667 * controls the maximum allowed adaptation speed of the codec. 668 * This factor controls the maximum amount of bits that can 669 * be added to the target bitrate in order to compensate for 670 * prior undershoot. 671 * 672 * Valid values in the range 0-1000. 673 */ 674 unsigned int rc_overshoot_pct; 675 676 /* 677 * decoder buffer model parameters 678 */ 679 680 /*!\brief Decoder Buffer Size 681 * 682 * This value indicates the amount of data that may be buffered by the 683 * decoding application. Note that this value is expressed in units of 684 * time (milliseconds). For example, a value of 5000 indicates that the 685 * client will buffer (at least) 5000ms worth of encoded data. Use the 686 * target bitrate (#rc_target_bitrate) to convert to bits/bytes, if 687 * necessary. 688 */ 689 unsigned int rc_buf_sz; 690 691 /*!\brief Decoder Buffer Initial Size 692 * 693 * This value indicates the amount of data that will be buffered by the 694 * decoding application prior to beginning playback. This value is 695 * expressed in units of time (milliseconds). Use the target bitrate 696 * (#rc_target_bitrate) to convert to bits/bytes, if necessary. 697 */ 698 unsigned int rc_buf_initial_sz; 699 700 /*!\brief Decoder Buffer Optimal Size 701 * 702 * This value indicates the amount of data that the encoder should try 703 * to maintain in the decoder's buffer. This value is expressed in units 704 * of time (milliseconds). Use the target bitrate (#rc_target_bitrate) 705 * to convert to bits/bytes, if necessary. 706 */ 707 unsigned int rc_buf_optimal_sz; 708 709 /* 710 * 2 pass rate control parameters 711 */ 712 713 /*!\brief Two-pass mode CBR/VBR bias 714 * 715 * Bias, expressed on a scale of 0 to 100, for determining target size 716 * for the current frame. The value 0 indicates the optimal CBR mode 717 * value should be used. The value 100 indicates the optimal VBR mode 718 * value should be used. Values in between indicate which way the 719 * encoder should "lean." 720 */ 721 unsigned int rc_2pass_vbr_bias_pct; 722 723 /*!\brief Two-pass mode per-GOP minimum bitrate 724 * 725 * This value, expressed as a percentage of the target bitrate, indicates 726 * the minimum bitrate to be used for a single GOP (aka "section") 727 */ 728 unsigned int rc_2pass_vbr_minsection_pct; 729 730 /*!\brief Two-pass mode per-GOP maximum bitrate 731 * 732 * This value, expressed as a percentage of the target bitrate, indicates 733 * the maximum bitrate to be used for a single GOP (aka "section") 734 */ 735 unsigned int rc_2pass_vbr_maxsection_pct; 736 737 /* 738 * keyframing settings (kf) 739 */ 740 741 /*!\brief Option to enable forward reference key frame 742 * 743 */ 744 int fwd_kf_enabled; 745 746 /*!\brief Keyframe placement mode 747 * 748 * This value indicates whether the encoder should place keyframes at a 749 * fixed interval, or determine the optimal placement automatically 750 * (as governed by the #kf_min_dist and #kf_max_dist parameters) 751 */ 752 enum aom_kf_mode kf_mode; 753 754 /*!\brief Keyframe minimum interval 755 * 756 * This value, expressed as a number of frames, prevents the encoder from 757 * placing a keyframe nearer than kf_min_dist to the previous keyframe. At 758 * least kf_min_dist frames non-keyframes will be coded before the next 759 * keyframe. Set kf_min_dist equal to kf_max_dist for a fixed interval. 760 */ 761 unsigned int kf_min_dist; 762 763 /*!\brief Keyframe maximum interval 764 * 765 * This value, expressed as a number of frames, forces the encoder to code 766 * a keyframe if one has not been coded in the last kf_max_dist frames. 767 * A value of 0 implies all frames will be keyframes. Set kf_min_dist 768 * equal to kf_max_dist for a fixed interval. 769 */ 770 unsigned int kf_max_dist; 771 772 /*!\brief sframe interval 773 * 774 * This value, expressed as a number of frames, forces the encoder to code 775 * an S-Frame every sframe_dist frames. 776 */ 777 unsigned int sframe_dist; 778 779 /*!\brief sframe insertion mode 780 * 781 * This value must be set to 1 or 2, and tells the encoder how to insert 782 * S-Frames. It will only have an effect if sframe_dist != 0. 783 * 784 * If altref is enabled: 785 * - if sframe_mode == 1, the considered frame will be made into an 786 * S-Frame only if it is an altref frame 787 * - if sframe_mode == 2, the next altref frame will be made into an 788 * S-Frame. 789 * 790 * Otherwise: the considered frame will be made into an S-Frame. 791 */ 792 unsigned int sframe_mode; 793 794 /*!\brief Tile coding mode 795 * 796 * This value indicates the tile coding mode. 797 * A value of 0 implies a normal non-large-scale tile coding. A value of 1 798 * implies a large-scale tile coding. 799 */ 800 unsigned int large_scale_tile; 801 802 /*!\brief Monochrome mode 803 * 804 * If this is nonzero, the encoder will generate a monochrome stream 805 * with no chroma planes. 806 */ 807 unsigned int monochrome; 808 809 /*!\brief full_still_picture_hdr 810 * 811 * If this is nonzero, the encoder will generate a full header even for 812 * still picture encoding. if zero, a reduced header is used for still 813 * picture. This flag has no effect when a regular video with more than 814 * a single frame is encoded. 815 */ 816 unsigned int full_still_picture_hdr; 817 818 /*!\brief Bitstream syntax mode 819 * 820 * This value indicates the bitstream syntax mode. 821 * A value of 0 indicates bitstream is saved as Section 5 bitstream. A value 822 * of 1 indicates the bitstream is saved in Annex-B format 823 */ 824 unsigned int save_as_annexb; 825 826 /*!\brief Number of explicit tile widths specified 827 * 828 * This value indicates the number of tile widths specified 829 * A value of 0 implies no tile widths are specified. 830 * Tile widths are given in the array tile_widths[] 831 */ 832 int tile_width_count; 833 834 /*!\brief Number of explicit tile heights specified 835 * 836 * This value indicates the number of tile heights specified 837 * A value of 0 implies no tile heights are specified. 838 * Tile heights are given in the array tile_heights[] 839 */ 840 int tile_height_count; 841 842 /*!\brief Maximum number of tile widths in tile widths array 843 * 844 * This define gives the maximum number of elements in the tile_widths array. 845 */ 846 #define MAX_TILE_WIDTHS 64 // maximum tile width array length 847 848 /*!\brief Array of specified tile widths 849 * 850 * This array specifies tile widths (and may be empty) 851 * The number of widths specified is given by tile_width_count 852 */ 853 int tile_widths[MAX_TILE_WIDTHS]; 854 855 /*!\brief Maximum number of tile heights in tile heights array. 856 * 857 * This define gives the maximum number of elements in the tile_heights array. 858 */ 859 #define MAX_TILE_HEIGHTS 64 // maximum tile height array length 860 861 /*!\brief Array of specified tile heights 862 * 863 * This array specifies tile heights (and may be empty) 864 * The number of heights specified is given by tile_height_count 865 */ 866 int tile_heights[MAX_TILE_HEIGHTS]; 867 868 /*!\brief Whether encoder should use fixed QP offsets. 869 * 870 * If a value of 1 is provided, encoder will use fixed QP offsets for frames 871 * at different levels of the pyramid. 872 * - If 'fixed_qp_offsets' is also provided, encoder will use the given 873 * offsets 874 * - If not, encoder will select the fixed offsets based on the cq-level 875 * provided. 876 * If a value of 0 is provided and fixed_qp_offset are not provided, encoder 877 * will NOT use fixed QP offsets. 878 * Note: This option is only relevant for --end-usage=q. 879 */ 880 unsigned int use_fixed_qp_offsets; 881 882 /*!\brief Number of fixed QP offsets 883 * 884 * This defines the number of elements in the fixed_qp_offsets array. 885 */ 886 #define FIXED_QP_OFFSET_COUNT 5 887 888 /*!\brief Array of fixed QP offsets 889 * 890 * This array specifies fixed QP offsets (range: 0 to 63) for frames at 891 * different levels of the pyramid. It is a comma-separated list of 5 values: 892 * - QP offset for keyframe 893 * - QP offset for ALTREF frame 894 * - QP offset for 1st level internal ARF 895 * - QP offset for 2nd level internal ARF 896 * - QP offset for 3rd level internal ARF 897 * Notes: 898 * - QP offset for leaf level frames is not explicitly specified. These frames 899 * use the worst quality allowed (--cq-level). 900 * - This option is only relevant for --end-usage=q. 901 */ 902 int fixed_qp_offsets[FIXED_QP_OFFSET_COUNT]; 903 904 /*!\brief Options defined per config file 905 * 906 */ 907 cfg_options_t encoder_cfg; 908 } aom_codec_enc_cfg_t; /**< alias for struct aom_codec_enc_cfg */ 909 910 /*!\brief Initialize an encoder instance 911 * 912 * Initializes a encoder context using the given interface. Applications 913 * should call the aom_codec_enc_init convenience macro instead of this 914 * function directly, to ensure that the ABI version number parameter 915 * is properly initialized. 916 * 917 * If the library was configured with --disable-multithread, this call 918 * is not thread safe and should be guarded with a lock if being used 919 * in a multithreaded context. 920 * 921 * \param[in] ctx Pointer to this instance's context. 922 * \param[in] iface Pointer to the algorithm interface to use. 923 * \param[in] cfg Configuration to use, if known. 924 * \param[in] flags Bitfield of AOM_CODEC_USE_* flags 925 * \param[in] ver ABI version number. Must be set to 926 * AOM_ENCODER_ABI_VERSION 927 * \retval #AOM_CODEC_OK 928 * The decoder algorithm initialized. 929 * \retval #AOM_CODEC_MEM_ERROR 930 * Memory allocation failed. 931 */ 932 aom_codec_err_t aom_codec_enc_init_ver(aom_codec_ctx_t *ctx, 933 aom_codec_iface_t *iface, 934 const aom_codec_enc_cfg_t *cfg, 935 aom_codec_flags_t flags, int ver); 936 937 /*!\brief Convenience macro for aom_codec_enc_init_ver() 938 * 939 * Ensures the ABI version parameter is properly set. 940 */ 941 #define aom_codec_enc_init(ctx, iface, cfg, flags) \ 942 aom_codec_enc_init_ver(ctx, iface, cfg, flags, AOM_ENCODER_ABI_VERSION) 943 944 /*!\brief Get the default configuration for a usage. 945 * 946 * Initializes an encoder configuration structure with default values. Supports 947 * the notion of "usages" so that an algorithm may offer different default 948 * settings depending on the user's intended goal. This function \ref SHOULD 949 * be called by all applications to initialize the configuration structure 950 * before specializing the configuration with application specific values. 951 * 952 * \param[in] iface Pointer to the algorithm interface to use. 953 * \param[out] cfg Configuration buffer to populate. 954 * \param[in] usage Algorithm specific usage value. For AV1, must be 955 * set to AOM_USAGE_GOOD_QUALITY (0) or 956 * AOM_USAGE_REALTIME (1). 957 * 958 * \retval #AOM_CODEC_OK 959 * The configuration was populated. 960 * \retval #AOM_CODEC_INCAPABLE 961 * Interface is not an encoder interface. 962 * \retval #AOM_CODEC_INVALID_PARAM 963 * A parameter was NULL, or the usage value was not recognized. 964 */ 965 aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface, 966 aom_codec_enc_cfg_t *cfg, 967 unsigned int usage); 968 969 /*!\brief Set or change configuration 970 * 971 * Reconfigures an encoder instance according to the given configuration. 972 * 973 * \param[in] ctx Pointer to this instance's context 974 * \param[in] cfg Configuration buffer to use 975 * 976 * \retval #AOM_CODEC_OK 977 * The configuration was populated. 978 * \retval #AOM_CODEC_INCAPABLE 979 * Interface is not an encoder interface. 980 * \retval #AOM_CODEC_INVALID_PARAM 981 * A parameter was NULL, or the usage value was not recognized. 982 */ 983 aom_codec_err_t aom_codec_enc_config_set(aom_codec_ctx_t *ctx, 984 const aom_codec_enc_cfg_t *cfg); 985 986 /*!\brief Get global stream headers 987 * 988 * Retrieves a stream level global header packet, if supported by the codec. 989 * Calls to this function should be deferred until all configuration information 990 * has been passed to libaom. Otherwise the global header data may be 991 * invalidated by additional configuration changes. 992 * 993 * The AV1 implementation of this function returns an OBU. The OBU returned is 994 * in Low Overhead Bitstream Format. Specifically, the obu_has_size_field bit is 995 * set, and the buffer contains the obu_size field for the returned OBU. 996 * 997 * \param[in] ctx Pointer to this instance's context 998 * 999 * \retval NULL 1000 * Encoder does not support global header, or an error occurred while 1001 * generating the global header. 1002 * 1003 * \retval Non-NULL 1004 * Pointer to buffer containing global header packet. The caller owns the 1005 * memory associated with this buffer, and must free the 'buf' member of the 1006 * aom_fixed_buf_t as well as the aom_fixed_buf_t pointer. Memory returned 1007 * must be freed via call to free(). 1008 */ 1009 aom_fixed_buf_t *aom_codec_get_global_headers(aom_codec_ctx_t *ctx); 1010 1011 /*!\brief usage parameter analogous to AV1 GOOD QUALITY mode. */ 1012 #define AOM_USAGE_GOOD_QUALITY (0) 1013 /*!\brief usage parameter analogous to AV1 REALTIME mode. */ 1014 #define AOM_USAGE_REALTIME (1) 1015 1016 /*!\brief Encode a frame 1017 * 1018 * Encodes a video frame at the given "presentation time." The presentation 1019 * time stamp (PTS) \ref MUST be strictly increasing. 1020 * 1021 * When the last frame has been passed to the encoder, this function should 1022 * continue to be called, with the img parameter set to NULL. This will 1023 * signal the end-of-stream condition to the encoder and allow it to encode 1024 * any held buffers. Encoding is complete when aom_codec_encode() is called 1025 * and aom_codec_get_cx_data() returns no data. 1026 * 1027 * \param[in] ctx Pointer to this instance's context 1028 * \param[in] img Image data to encode, NULL to flush. 1029 * \param[in] pts Presentation time stamp, in timebase units. 1030 * \param[in] duration Duration to show frame, in timebase units. 1031 * \param[in] flags Flags to use for encoding this frame. 1032 * 1033 * \retval #AOM_CODEC_OK 1034 * The configuration was populated. 1035 * \retval #AOM_CODEC_INCAPABLE 1036 * Interface is not an encoder interface. 1037 * \retval #AOM_CODEC_INVALID_PARAM 1038 * A parameter was NULL, the image format is unsupported, etc. 1039 */ 1040 aom_codec_err_t aom_codec_encode(aom_codec_ctx_t *ctx, const aom_image_t *img, 1041 aom_codec_pts_t pts, unsigned long duration, 1042 aom_enc_frame_flags_t flags); 1043 1044 /*!\brief Set compressed data output buffer 1045 * 1046 * Sets the buffer that the codec should output the compressed data 1047 * into. This call effectively sets the buffer pointer returned in the 1048 * next AOM_CODEC_CX_FRAME_PKT packet. Subsequent packets will be 1049 * appended into this buffer. The buffer is preserved across frames, 1050 * so applications must periodically call this function after flushing 1051 * the accumulated compressed data to disk or to the network to reset 1052 * the pointer to the buffer's head. 1053 * 1054 * `pad_before` bytes will be skipped before writing the compressed 1055 * data, and `pad_after` bytes will be appended to the packet. The size 1056 * of the packet will be the sum of the size of the actual compressed 1057 * data, pad_before, and pad_after. The padding bytes will be preserved 1058 * (not overwritten). 1059 * 1060 * Note that calling this function does not guarantee that the returned 1061 * compressed data will be placed into the specified buffer. In the 1062 * event that the encoded data will not fit into the buffer provided, 1063 * the returned packet \ref MAY point to an internal buffer, as it would 1064 * if this call were never used. In this event, the output packet will 1065 * NOT have any padding, and the application must free space and copy it 1066 * to the proper place. This is of particular note in configurations 1067 * that may output multiple packets for a single encoded frame (e.g., lagged 1068 * encoding) or if the application does not reset the buffer periodically. 1069 * 1070 * Applications may restore the default behavior of the codec providing 1071 * the compressed data buffer by calling this function with a NULL 1072 * buffer. 1073 * 1074 * Applications \ref MUSTNOT call this function during iteration of 1075 * aom_codec_get_cx_data(). 1076 * 1077 * \param[in] ctx Pointer to this instance's context 1078 * \param[in] buf Buffer to store compressed data into 1079 * \param[in] pad_before Bytes to skip before writing compressed data 1080 * \param[in] pad_after Bytes to skip after writing compressed data 1081 * 1082 * \retval #AOM_CODEC_OK 1083 * The buffer was set successfully. 1084 * \retval #AOM_CODEC_INVALID_PARAM 1085 * A parameter was NULL, the image format is unsupported, etc. 1086 */ 1087 aom_codec_err_t aom_codec_set_cx_data_buf(aom_codec_ctx_t *ctx, 1088 const aom_fixed_buf_t *buf, 1089 unsigned int pad_before, 1090 unsigned int pad_after); 1091 1092 /*!\brief Encoded data iterator 1093 * 1094 * Iterates over a list of data packets to be passed from the encoder to the 1095 * application. The different kinds of packets available are enumerated in 1096 * #aom_codec_cx_pkt_kind. 1097 * 1098 * #AOM_CODEC_CX_FRAME_PKT packets should be passed to the application's 1099 * muxer. Multiple compressed frames may be in the list. 1100 * #AOM_CODEC_STATS_PKT packets should be appended to a global buffer. 1101 * 1102 * The application \ref MUST silently ignore any packet kinds that it does 1103 * not recognize or support. 1104 * 1105 * The data buffers returned from this function are only guaranteed to be 1106 * valid until the application makes another call to any aom_codec_* function. 1107 * 1108 * \param[in] ctx Pointer to this instance's context 1109 * \param[in,out] iter Iterator storage, initialized to NULL 1110 * 1111 * \return Returns a pointer to an output data packet (compressed frame data, 1112 * two-pass statistics, etc.) or NULL to signal end-of-list. 1113 * 1114 */ 1115 const aom_codec_cx_pkt_t *aom_codec_get_cx_data(aom_codec_ctx_t *ctx, 1116 aom_codec_iter_t *iter); 1117 1118 /*!\brief Get Preview Frame 1119 * 1120 * Returns an image that can be used as a preview. Shows the image as it would 1121 * exist at the decompressor. The application \ref MUST NOT write into this 1122 * image buffer. 1123 * 1124 * \param[in] ctx Pointer to this instance's context 1125 * 1126 * \return Returns a pointer to a preview image, or NULL if no image is 1127 * available. 1128 * 1129 */ 1130 const aom_image_t *aom_codec_get_preview_frame(aom_codec_ctx_t *ctx); 1131 1132 /*!@} - end defgroup encoder*/ 1133 #ifdef __cplusplus 1134 } 1135 #endif 1136 #endif // AOM_AOM_AOM_ENCODER_H_ 1137