High-level Algorithm

Topics
	Speed vs Quality Trade Off
	Source Frame Processing
	Rate Control
	Temporal Dependency Modelling
	Two Pass Mode
	The Look-Ahead Buffer
	Golden Frame Group

Functions
int	av1_pack_bitstream (struct AV1_COMP *const cpi, uint8_t *dst, size_t dst_size, size_t *size, int *const largest_tile_id)
	Pack the bitstream for one frame.
int	av1_encode_strategy (AV1_COMP *const cpi, size_t *const size, uint8_t *const dest, size_t dest_size, unsigned int *frame_flags, int64_t *const time_stamp, int64_t *const time_end, const aom_rational64_t *const timestamp_ratio, int *const pop_lookahead, int flush)
	Implement high-level encode strategy.
static void	encode_frame_internal (AV1_COMP *cpi)
	Encoder setup(only for the current frame), encoding, and recontruction for a single frame.
void	av1_encode_frame (AV1_COMP *cpi)
	Setup reference frame buffers and encode a frame.
static void	cdef_restoration_frame (AV1_COMP *cpi, AV1_COMMON *cm, MACROBLOCKD *xd, int use_restoration, int use_cdef, unsigned int skip_apply_postproc_filters)
	Select and apply cdef filters and switchable restoration filters.
static void	loopfilter_frame (AV1_COMP *cpi, AV1_COMMON *cm)
	Select and apply deblocking filters, cdef filters, and restoration filters.
static int	encode_without_recode (AV1_COMP *cpi)
	Encode a frame without the recode loop, usually used in one-pass encoding and realtime coding.
static int	encode_with_recode_loop (AV1_COMP *cpi, size_t *size, uint8_t *dest, size_t dest_size)
	Recode loop for encoding one frame. the purpose of encoding one frame for multiple times can be approaching a target bitrate or adjusting the usage of global motions.
static int	encode_with_recode_loop_and_filter (AV1_COMP *cpi, size_t *size, uint8_t *dest, size_t dest_size, int64_t *sse, int64_t *rate, int *largest_tile_id)
	Recode loop or a single loop for encoding one frame, followed by in-loop deblocking filters, CDEF filters, and restoration filters.
static int	encode_frame_to_data_rate (AV1_COMP *cpi, size_t *size, uint8_t *dest, size_t dest_size)
	Run the final pass encoding for 1-pass/2-pass encoding mode, and pack the bitstream.
int	av1_receive_raw_frame (AV1_COMP *cpi, aom_enc_frame_flags_t frame_flags, const YV12_BUFFER_CONFIG *sd, int64_t time_stamp, int64_t end_time_stamp)
	Obtain the raw frame data.
int	av1_get_compressed_data (AV1_COMP *cpi, AV1_COMP_DATA *const cpi_data)
	Encode a frame.
int	av1_encode (AV1_COMP *const cpi, uint8_t *const dest, size_t dest_size, const EncodeFrameInput *const frame_input, const EncodeFrameParams *const frame_params, size_t *const frame_size)
	Run 1-pass/2-pass encoding.

Detailed Description

This module describes sequence level/frame level algorithm in AV1. More details will be added.

Function Documentation

av1_encode_strategy()

int av1_encode_strategy	(	AV1_COMP *const	cpi,
		size_t *const	size,
		uint8_t *const	dest,
		size_t	dest_size,
		unsigned int *	frame_flags,
		int64_t *const	time_stamp,
		int64_t *const	time_end,
		const aom_rational64_t *const	timestamp_ratio,
		int *const	pop_lookahead,
		int	flush )

Implement high-level encode strategy.

This function will implement high-level encode strategy, choosing frame type, frame placement, etc. It populates an EncodeFrameParams struct with the results of these decisions and then encodes the frame. The caller should use the output parameters *time_stamp and *time_end only when this function returns AOM_CODEC_OK.

Parameters

[in]	cpi	Top-level encoder structure
[out]	size	Bitstream size
[out]	dest	Bitstream output buffer
[in]	dest_size	Bitstream output buffer size
[in]	frame_flags	Flags to decide how to encoding the frame
[out]	time_stamp	Time stamp of the frame
[out]	time_end	Time end
[in]	timestamp_ratio	Time base
[in]	pop_lookahead	Decide to pop the source frame from queue
[in]	flush	Decide to encode one frame or the rest of frames

Returns

Returns a value to indicate if the encoding is done successfully.

Return values

AOM_CODEC_OK
-1
AOM_CODEC_ERROR

References AOM_CODEC_ERROR, AOM_CODEC_OK, AOM_EFLAG_FORCE_KF, AOM_Q, AOM_RC_FIRST_PASS, AOM_RC_ONE_PASS, AOM_RC_SECOND_PASS, AOM_RC_THIRD_PASS, av1_cyclic_refresh_update_parameters(), av1_encode(), av1_get_second_pass_params(), av1_lookahead_depth(), av1_lookahead_peek(), av1_lookahead_pop_sz(), av1_set_rtc_reference_structure_one_layer(), AV1_COMP::common, AV1_COMP::compressor_stage, RateControlCfg::cq_level, AV1Common::cur_frame, AV1Common::current_frame, RateControlCfg::drop_frames_water_mark, AV1_COMP::droppable, AV1Common::error, AV1_COMP::ext_flags, AV1Common::features, AV1Common::film_grain_params, AV1_COMP::film_grain_table, TimeStamps::first_ts_start, AV1Common::frame_presentation_time, EncodeFrameParams::frame_type, AV1_COMP::framerate, AV1_COMP::gf_frame_index, AV1_PRIMARY::gf_group, has_no_stats_stage(), AV1_COMP::is_dropped_frame, KeyFrameCfg::key_freq_max, AV1EncoderConfig::kf_cfg, AV1_PRIMARY::lookahead, AV1Common::mi_params, RateControlCfg::mode, AV1_COMP::mv_stats, AV1_PRIMARY::mv_stats, AV1_COMP::orig_source, AV1_COMP::oxcf, AV1_PRIMARY::p_rc, AV1EncoderConfig::pass, AV1_COMP::ppi, TimeStamps::prev_ts_end, AV1Common::quant_params, PRIMARY_RATE_CONTROL::rate_correction_factors, AV1_COMP::rc, AV1EncoderConfig::rc_cfg, EncodeFrameParams::ref_frame_flags, ExternalFlags::ref_frame_flags, AV1Common::ref_frame_map, AV1_COMP::ref_refresh_index, AV1_COMP::refresh_frame, EncodeFrameParams::refresh_frame, ExternalFlags::refresh_frame, AV1_COMP::refresh_idx_available, AV1Common::remapped_ref_idx, EncodeFrameParams::remapped_ref_idx, SPEED_FEATURES::rt_sf, AV1_PRIMARY::rtc_ref, AV1_COMP::second_pass_log_stream, AV1Common::seq_params, AV1_COMP::sf, AV1_PRIMARY::show_existing_alt_ref, EncodeFrameParams::show_frame, AV1_COMP::skip_tpl_setup_stats, AV1_COMP::source, EncodeFrameParams::speed, AV1_COMP::svc, AV1_COMP::td, AV1_COMP::time_stamps, AV1_PRIMARY::tpl_data, TplParams::tpl_stats_pool, AV1_PRIMARY::twopass, AV1_COMP::twopass_frame, AV1_COMP::unfiltered_source, AV1_COMP::unscaled_source, ExtRefreshFrameFlagsInfo::update_pending, AV1_PRIMARY::use_svc, and CommonQuantParams::using_qmatrix.

Referenced by av1_get_compressed_data().

av1_encode_frame()

void av1_encode_frame

(

AV1_COMP *

cpi

)

Setup reference frame buffers and encode a frame.

Parameters

[in]

cpi

Top-level encoder structure

References AV1_COMP::alloc_pyramid, FeatureFlags::allow_warped_motion, AV1_COMP::common, AV1Common::cur_frame, AV1Common::current_frame, AV1_COMP::enc_seg, encode_frame_internal(), AV1Common::features, HIGH_LEVEL_SPEED_FEATURES::frame_parameter_update, SPEED_FEATURES::hl_sf, FeatureFlags::interp_filter, CommonTileParams::large_scale, EncSegmentationInfo::map, CommonModeInfoParams::mi_cols, AV1Common::mi_params, CommonModeInfoParams::mi_rows, AV1_COMP::oxcf, AV1_COMP::palette_pixel_num, FeatureFlags::reduced_tx_set_used, AV1_COMP::ref_frame_dist_info, AV1_COMP::ref_frame_flags, SPEED_FEATURES::rt_sf, AV1Common::seg, AV1_COMP::sf, AV1_COMP::source, FeatureFlags::switchable_motion_mode, AV1_COMP::td, AV1Common::tiles, and FeatureFlags::tx_mode.

Referenced by encode_with_recode_loop(), and encode_without_recode().

encode_without_recode()

static int encode_without_recode ( AV1_COMP * cpi )

static

Encode a frame without the recode loop, usually used in one-pass encoding and realtime coding.

Parameters

[in]

cpi

Top-level encoder structure

Returns

Returns a value to indicate if the encoding is done successfully.

Return values

AOM_CODEC_OK
AOM_CODEC_ERROR

References AV1_COMP::alloc_pyramid, AOM_CBR, AOM_CODEC_MEM_ERROR, AOM_CODEC_OK, av1_cyclic_refresh_setup(), av1_encode_frame(), av1_encodedframe_overshoot_cbr(), av1_set_speed_features_qindex_dependent(), av1_set_variance_partition_thresholds(), FeatureFlags::byte_alignment, AV1_COMP::common, AV1_COMP::consec_zero_mv, AV1_COMP::consec_zero_mv_alloc_size, AV1Common::cur_frame, AV1Common::current_frame, SVC::downsample_filter_phase, SVC::downsample_filter_type, RateControlCfg::drop_frames_water_mark, AV1_COMP::enc_quant_dequant_params, AV1Common::error, FeatureFlags::error_resilient_mode, AV1Common::features, AV1_COMP::gf_frame_index, AV1_PRIMARY::gf_group, RefreshFrameInfo::golden_frame, AV1Common::height, SPEED_FEATURES::inter_sf, AV1_COMP::last_source, CommonModeInfoParams::mi_cols, AV1Common::mi_params, CommonModeInfoParams::mi_rows, RateControlCfg::mode, AV1_COMP::orig_source, AV1_COMP::oxcf, SPEED_FEATURES::part_sf, AV1_COMP::ppi, AV1Common::prev_frame, FeatureFlags::primary_ref_frame, AV1Common::quant_params, AV1_COMP::rc, AV1EncoderConfig::rc_cfg, AV1_COMP::ref_frame_flags, AV1_COMP::refresh_frame, SPEED_FEATURES::rt_sf, AV1_PRIMARY::rtc_ref, AV1_COMP::scaled_last_source, AV1_COMP::scaled_last_source_available, AV1_COMP::scaled_source, AV1Common::seg, AV1Common::seq_params, AV1_COMP::sf, AV1_COMP::source, AV1_COMP::svc, AV1_COMP::unscaled_last_source, AV1_COMP::unscaled_source, AV1_PRIMARY::use_svc, and AV1Common::width.

Referenced by encode_with_recode_loop_and_filter().

encode_with_recode_loop()

static int encode_with_recode_loop ( AV1_COMP * cpi, size_t * size, uint8_t * dest, size_t dest_size )

static

Recode loop for encoding one frame. the purpose of encoding one frame for multiple times can be approaching a target bitrate or adjusting the usage of global motions.

Parameters

[in]	cpi	Top-level encoder structure
[in]	size	Bitstream size
[out]	dest	Bitstream output buffer
[in]	dest_size	Bitstream output buffer size

Returns

Returns a value to indicate if the encoding is done successfully.

Return values

AOM_CODEC_OK
-1
AOM_CODEC_ERROR

References AV1_COMP::alloc_pyramid, FeatureFlags::allow_high_precision_mv, AOM_CODEC_ERROR, AOM_CODEC_OK, AOM_Q, AOM_RC_THIRD_PASS, AOM_SUPERRES_NONE, av1_encode_frame(), av1_pack_bitstream(), av1_set_speed_features_qindex_dependent(), av1_set_variance_partition_thresholds(), RATE_CONTROL::coefficient_size, AV1_COMP::common, AV1Common::cur_frame, AV1Common::current_frame, AV1Common::delta_q_info, HIGH_LEVEL_SPEED_FEATURES::disable_extra_sc_testing, AV1_COMP::do_frame_data_update, AV1_COMP::ducky_encode_info, AV1_COMP::enc_quant_dequant_params, AV1Common::features, AV1_COMP::gf_frame_index, AV1_PRIMARY::gf_group, AV1_COMP::gm_info, AV1Common::height, HIGH_LEVEL_SPEED_FEATURES::high_precision_mv_usage, SPEED_FEATURES::hl_sf, InterpSearchFlags::interp_filter_search_mask, AV1_COMP::interp_search_flags, SPEED_FEATURES::interp_sf, AV1_COMP::last_source, RateControlCfg::min_cr, RateControlCfg::mode, AV1_COMP::mv_stats, AV1_COMP::num_frame_recode, NUM_RECODES_PER_FRAME, AV1_COMP::oxcf, SPEED_FEATURES::part_sf, AV1EncoderConfig::pass, AV1_COMP::ppi, AV1Common::prev_frame, RATE_CONTROL::projected_frame_size, AV1Common::quant_params, AV1_COMP::rc, AV1EncoderConfig::rc_cfg, AV1_COMP::rd, HIGH_LEVEL_SPEED_FEATURES::recode_loop, recode_loop_update_q(), AV1_COMP::scaled_last_source, AV1_COMP::scaled_source, GlobalMotionInfo::search_done, AV1Common::seg, AV1Common::seq_params, AV1_COMP::sf, AV1_COMP::source, AV1_COMP::superres_mode, AV1Common::superres_scale_denominator, AV1_COMP::unscaled_last_source, AV1_COMP::unscaled_source, AV1_COMP::use_ducky_encode, and AV1Common::width.

Referenced by encode_with_recode_loop_and_filter().

encode_with_recode_loop_and_filter()

static int encode_with_recode_loop_and_filter ( AV1_COMP * cpi, size_t * size, uint8_t * dest, size_t dest_size, int64_t * sse, int64_t * rate, int * largest_tile_id )

static

Recode loop or a single loop for encoding one frame, followed by in-loop deblocking filters, CDEF filters, and restoration filters.

Parameters

[in]	cpi	Top-level encoder structure
[in]	size	Bitstream size
[out]	dest	Bitstream output buffer
[in]	dest_size	Bitstream output buffer size
[in]	sse	Total distortion of the frame
[in]	rate	Total rate of the frame
[in]	largest_tile_id	Tile id of the last tile

Returns

Returns a value to indicate if the encoding is done successfully.

Return values

AOM_CODEC_OK
AOM_CODEC_ERROR

References FeatureFlags::allow_intrabc, AV1_COMP::ambient_err, AOM_CODEC_ERROR, AOM_CODEC_OK, av1_pack_bitstream(), av1_postencode_drop_cbr(), CommonQuantParams::base_qindex, RATE_CONTROL::coefficient_size, AV1_COMP::common, AV1Common::cur_frame, AV1_COMP::do_update_frame_probs_interpfilter, AV1_COMP::do_update_frame_probs_obmc, AV1_COMP::do_update_frame_probs_txtype, AV1_COMP::do_update_frame_probs_warp, AV1_COMP::do_update_vbr_bits_off_target_fast, AV1_COMP::ducky_encode_info, encode_with_recode_loop(), encode_without_recode(), AV1Common::features, RATE_CONTROL::frames_to_key, SPEED_FEATURES::hl_sf, loopfilter_frame(), AV1_COMP::mt_info, NUM_RECODES_PER_FRAME, AV1_COMP::oxcf, AV1_PRIMARY::p_rc, MultiThreadInfo::pipeline_lpf_mt_with_enc, AV1_COMP::ppi, AV1Common::quant_params, AV1_COMP::rc, AV1_COMP::rd, HIGH_LEVEL_SPEED_FEATURES::recode_loop, AV1Common::render_height, AV1Common::render_width, AV1_PRIMARY::rtc_ref, AV1Common::seq_params, AV1_COMP::sf, AV1_COMP::source, and AV1_COMP::use_ducky_encode.

Referenced by encode_frame_to_data_rate().

encode_frame_to_data_rate()

static int encode_frame_to_data_rate ( AV1_COMP * cpi, size_t * size, uint8_t * dest, size_t dest_size )

static

Run the final pass encoding for 1-pass/2-pass encoding mode, and pack the bitstream.

Parameters

[in]	cpi	Top-level encoder structure
[in]	size	Bitstream size
[out]	dest	Bitstream output buffer
[in]	dest_size	Bitstream output buffer size

Returns

Returns a value to indicate if the encoding is done successfully.

Return values

AOM_CODEC_OK
AOM_CODEC_ERROR

References AV1EncoderConfig::algo_cfg, FeatureFlags::allow_ref_frame_mvs, FeatureFlags::allow_screen_content_tools, FeatureFlags::allow_warped_motion, AOM_CBR, AOM_CODEC_ERROR, AOM_CODEC_MEM_ERROR, AOM_CODEC_OK, AOM_RC_THIRD_PASS, av1_pack_bitstream(), CommonQuantParams::base_qindex, AlgoCfg::cdf_update_mode, RATE_CONTROL::coefficient_size, AV1_COMP::common, AV1Common::cur_frame, FeatureFlags::cur_frame_force_integer_mv, AV1Common::current_frame, AV1Common::current_frame_id, FeatureFlags::disable_cdf_update, AV1_COMP::do_frame_data_update, AV1_COMP::ducky_encode_info, encode_with_recode_loop_and_filter(), AV1_COMP::existing_fb_idx_to_show, AV1Common::fc, AV1Common::features, AV1_COMP::force_intpel_info, AV1_COMP::frames_since_last_update, AV1_COMP::gf_frame_index, AV1_PRIMARY::gf_group, has_no_stats_stage(), AV1Common::height, AV1_COMP::is_dropped_frame, AV1EncoderConfig::kf_cfg, CommonTileParams::large_scale, AV1Common::last_frame_seg_map, AV1_COMP::last_frame_type, AV1_COMP::last_show_frame_buf, AV1Common::lf, RateControlCfg::mode, AV1_COMP::num_tg, AV1_COMP::oxcf, AV1EncoderConfig::pass, AV1_COMP::ppi, AV1Common::quant_params, AV1_COMP::rc, AV1EncoderConfig::rc_cfg, AV1_COMP::rd, AV1Common::ref_frame_id, FeatureFlags::refresh_frame_context, SPEED_FEATURES::rt_sf, AV1_PRIMARY::rtc_ref, AV1Common::seg, AV1Common::seq_params, AV1_COMP::sf, KeyFrameCfg::sframe_dist, AV1Common::show_frame, CommonTileParams::single_tile_decoding, AV1_COMP::source, AV1_COMP::superres_mode, AV1_COMP::svc, RateControlCfg::target_bandwidth, AV1_COMP::td, AV1_COMP::tile_data, AV1Common::tiles, AV1_COMP::unscaled_last_source, AV1_COMP::use_ducky_encode, and AV1Common::width.

Referenced by av1_encode().

av1_receive_raw_frame()

int av1_receive_raw_frame	(	AV1_COMP *	cpi,
		aom_enc_frame_flags_t	frame_flags,
		const YV12_BUFFER_CONFIG *	sd,
		int64_t	time_stamp,
		int64_t	end_time_stamp )

Obtain the raw frame data.

This function receives the raw frame data from input.

Parameters

[in]	cpi	Top-level encoder structure
[in]	frame_flags	Flags to decide how to encoding the frame
[in,out]	sd	Contain raw frame data
[in]	time_stamp	Time stamp of the frame
[in]	end_time_stamp	End time stamp

Returns

Returns a value to indicate if the frame data is received successfully.

Note

The caller can assume that a copy of this frame is made and not just a copy of the pointer.

< Y (Luminance) plane

< Y (Luminance) plane

References AV1_COMP::alloc_pyramid, AOM_BITS_10, AOM_CODEC_ERROR, AOM_CODEC_INVALID_PARAM, AOM_PLANE_Y, AOM_RC_FIRST_PASS, av1_lookahead_push(), AV1_COMP::common, AV1Common::error, AV1_PRIMARY::lookahead, AV1_COMP::oxcf, AV1EncoderConfig::pass, AV1_COMP::ppi, and AV1Common::seq_params.

av1_get_compressed_data()

int av1_get_compressed_data	(	AV1_COMP *	cpi,
		AV1_COMP_DATA *const	cpi_data )

Encode a frame.

This function encodes the raw frame data, and outputs the frame bit stream to the designated buffer. The caller should use the output parameters cpi_data->ts_frame_start and cpi_data->ts_frame_end only when this function returns AOM_CODEC_OK.

Parameters

[in]	cpi	Top-level encoder structure
[in,out]	cpi_data	Data corresponding to a frame encode

Returns

Returns a value to indicate if the encoding is done successfully.

Return values

AOM_CODEC_OK
-1	No frame encoded; more input is required.
A nonzero (positive) aom_codec_err_t code	The encoding failed with the error. Sets the error code and error message in cpi->common.error.

References AOM_CODEC_ERROR, AOM_CODEC_OK, av1_encode_strategy(), av1_one_pass_cbr_svc_start_layer(), AV1_COMP::available_bs_size, CommonQuantParams::base_qindex, AV1_COMP::common, AV1_COMP::compressor_stage, AV1_COMP::counts, AV1Common::current_frame, AV1_COMP_DATA::cx_data, AV1_COMP_DATA::cx_data_sz, AV1Common::error, AV1Common::features, AV1_COMP_DATA::flush, AV1_COMP_DATA::frame_size, AV1_COMP::gf_frame_index, AV1_PRIMARY::gf_group, AV1_COMP::is_dropped_frame, AV1_COMP_DATA::lib_flags, AV1_COMP::oxcf, AV1EncoderConfig::pass, AV1_COMP_DATA::pop_lookahead, AV1_COMP::ppi, AV1Common::quant_params, AV1_COMP::rc, FeatureFlags::refresh_frame_context, AV1Common::seq_params, AV1Common::show_existing_frame, AV1Common::show_frame, AV1Common::showable_frame, AV1_COMP::td, AV1_COMP_DATA::timestamp_ratio, AV1_COMP_DATA::ts_frame_end, AV1_COMP_DATA::ts_frame_start, and AV1_PRIMARY::use_svc.