ratectrl.h

/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */
 
#ifndef AOM_AV1_ENCODER_RATECTRL_H_
#define AOM_AV1_ENCODER_RATECTRL_H_
 
#include "aom/aom_codec.h"
#include "aom/aom_integer.h"
 
#include "aom_ports/mem.h"
 
#include "av1/common/av1_common_int.h"
#include "av1/common/blockd.h"
 
#ifdef __cplusplus
extern "C" {
#endif
 
// Bits Per MB at different Q (Multiplied by 512)
#define BPER_MB_NORMBITS 9
 
// Use this macro to turn on/off use of alt-refs in one-pass mode.
#define USE_ALTREF_FOR_ONE_PASS 1
 
// Threshold used to define if a KF group is static (e.g. a slide show).
// Essentially, this means that no frame in the group has more than 1% of MBs
// that are not marked as coded with 0,0 motion in the first pass.
#define STATIC_KF_GROUP_THRESH 99
#define STATIC_KF_GROUP_FLOAT_THRESH 0.99
 
// The maximum duration of a GF group that is static (e.g. a slide show).
#define MAX_STATIC_GF_GROUP_LENGTH 250
 
#define MIN_GF_INTERVAL 4
#define MAX_GF_INTERVAL 32
#define FIXED_GF_INTERVAL 16
#define MAX_GF_LENGTH_LAP 16
 
#define FIXED_GF_INTERVAL_RT 80
#define MAX_GF_INTERVAL_RT 160
 
#define MAX_NUM_GF_INTERVALS 15
 
#define MAX_ARF_LAYERS 6
// #define STRICT_RC
 
#define DEFAULT_KF_BOOST_RT 2300
#define DEFAULT_GF_BOOST_RT 2000
 
// A passive rate control strategy for screen content type in real-time mode.
// When it is turned on, the compression performance is improved by
// 7.8% (overall_psnr), 5.0% (VMAF) on average. Some clips see gains
// over 20% on metric.
// The downside is that it does not guarantee frame size.
// Since RT mode has a tight restriction on buffer overflow control, we
// turn it off by default.
#define RT_PASSIVE_STRATEGY 0
#define MAX_Q_HISTORY 1000
 
typedef struct {
  int resize_width;
  int resize_height;
  uint8_t superres_denom;
} size_params_type;
 
enum {
  INTER_NORMAL,
  GF_ARF_LOW,
  GF_ARF_STD,
  KF_STD,
  RATE_FACTOR_LEVELS
} UENUM1BYTE(RATE_FACTOR_LEVEL);
 
enum {
  KF_UPDATE,
  LF_UPDATE,
  GF_UPDATE,
  ARF_UPDATE,
  OVERLAY_UPDATE,
  INTNL_OVERLAY_UPDATE,  // Internal Overlay Frame
  INTNL_ARF_UPDATE,      // Internal Altref Frame
  FRAME_UPDATE_TYPES
} UENUM1BYTE(FRAME_UPDATE_TYPE);
 
enum {
  REFBUF_RESET,   // Clear reference frame buffer
  REFBUF_UPDATE,  // Refresh reference frame buffer
  REFBUF_STATES
} UENUM1BYTE(REFBUF_STATE);
 
typedef enum {
  NO_RESIZE = 0,
  DOWN_THREEFOUR = 1,  // From orig to 3/4.
  DOWN_ONEHALF = 2,    // From orig or 3/4 to 1/2.
  UP_THREEFOUR = -1,   // From 1/2 to 3/4.
  UP_ORIG = -2,        // From 1/2 or 3/4 to orig.
} RESIZE_ACTION;
 
typedef enum { ORIG = 0, THREE_QUARTER = 1, ONE_HALF = 2 } RESIZE_STATE;
 
#define MAX_FIRSTPASS_ANALYSIS_FRAMES 150
typedef enum region_types {
  STABLE_REGION = 0,
  HIGH_VAR_REGION = 1,
  SCENECUT_REGION = 2,
  BLENDING_REGION = 3,
} REGION_TYPES;
 
typedef struct regions {
  int start;
  int last;
  double avg_noise_var;
  double avg_cor_coeff;
  double avg_sr_fr_ratio;
  double avg_intra_err;
  double avg_coded_err;
  REGION_TYPES type;
} REGIONS;
 
typedef struct {
  // Rate targetting variables
 
  int base_frame_target;
  int this_frame_target;  // Actual frame target after rc adjustment.
 
  int projected_frame_size;
 
  int coefficient_size;
 
  int sb64_target_rate;
 
  int frames_since_golden;
 
  int frames_till_gf_update_due;
 
  int intervals_till_gf_calculate_due;
 
  int min_gf_interval;
  int max_gf_interval;
  int static_scene_max_gf_interval;
  int frames_to_key;
  int frames_since_key;
  int frames_to_fwd_kf;
  int is_src_frame_alt_ref;
  int sframe_due;
 
  int high_source_sad;
  uint64_t avg_source_sad;
  uint64_t prev_avg_source_sad;
  uint64_t frame_source_sad;
 
  int avg_frame_bandwidth;  // Average frame size target for clip
  int min_frame_bandwidth;  // Minimum allocation used for any frame
  int max_frame_bandwidth;  // Maximum burst rate allowed for a frame.
  int prev_avg_frame_bandwidth;
 
  int ni_av_qi;
  int ni_tot_qi;
 
  int decimation_factor;
  int decimation_count;
 
  int worst_quality;
  int best_quality;
 
  // rate control history for last frame(1) and the frame before(2).
  // -1: overshoot
  //  1: undershoot
  //  0: not initialized.
  int rc_1_frame;
  int rc_2_frame;
  int q_1_frame;
  int q_2_frame;
 
  int active_worst_quality;
 
  // Track amount of low motion in scene
  int avg_frame_low_motion;
 
  // signals if number of blocks with motion is high
  int high_num_blocks_with_motion;
 
  // For dynamic resize, 1 pass cbr.
  RESIZE_STATE resize_state;
  int resize_avg_qp;
  int resize_buffer_underflow;
  int resize_count;
 
  // Flag to disable content related qp adjustment.
  int rtc_external_ratectrl;
 
  // Stores fast_extra_bits of the current frame.
  int frame_level_fast_extra_bits;
 
  double frame_level_rate_correction_factors[RATE_FACTOR_LEVELS];
} RATE_CONTROL;
 
typedef struct {
  // Sub-gop level Rate targetting variables
 
  int64_t gf_group_bits;
 
  int kf_boost;
 
  int gfu_boost;
 
  int gf_intervals[MAX_NUM_GF_INTERVALS];
 
  int cur_gf_index;
 
  int num_regions;
 
  REGIONS regions[MAX_FIRSTPASS_ANALYSIS_FRAMES];
  int regions_offset;  // offset of regions from the last keyframe
  int frames_till_regions_update;
 
  int baseline_gf_interval;
 
  int constrained_gf_group;
 
  int this_key_frame_forced;
 
  int next_key_frame_forced;
  int64_t starting_buffer_level;
 
  int64_t optimal_buffer_level;
 
  int64_t maximum_buffer_size;
 
  int arf_q;
 
  float_t arf_boost_factor;
 
  int base_layer_qp;
 
  // Total number of stats used only for kf_boost calculation.
  int num_stats_used_for_kf_boost;
 
  // Total number of stats used only for gfu_boost calculation.
  int num_stats_used_for_gfu_boost;
 
  // Total number of stats required by gfu_boost calculation.
  int num_stats_required_for_gfu_boost;
 
  int enable_scenecut_detection;
 
  int use_arf_in_this_kf_group;
 
  int ni_frames;
 
  double tot_q;
  int last_kf_qindex;
 
  int avg_frame_qindex[FRAME_TYPES];
 
#if CONFIG_FRAME_PARALLEL_ENCODE && CONFIG_FPMT_TEST
 
  int temp_active_best_quality[MAX_ARF_LAYERS + 1];
 
  int temp_last_boosted_qindex;
 
  double temp_avg_q;
 
  int temp_last_q[FRAME_TYPES];
 
  int temp_projected_frame_size;
 
  int64_t temp_total_actual_bits;
 
  int64_t temp_buffer_level;
 
  int64_t temp_vbr_bits_off_target;
 
  int64_t temp_vbr_bits_off_target_fast;
 
  double temp_rate_correction_factors[RATE_FACTOR_LEVELS];
 
  int temp_rate_error_estimate;
 
  int temp_rolling_arf_group_target_bits;
 
  int temp_rolling_arf_group_actual_bits;
 
  int64_t temp_bits_left;
 
  int temp_extend_minq;
 
  int temp_extend_maxq;
 
  int temp_extend_minq_fast;
#endif
 
  int active_best_quality[MAX_ARF_LAYERS + 1];
 
  int last_boosted_qindex;
 
  double avg_q;
 
  int last_q[FRAME_TYPES];
 
  double rate_correction_factors[RATE_FACTOR_LEVELS];
 
  int64_t total_actual_bits;
 
  int64_t total_target_bits;
 
  int64_t buffer_level;
 
  int rate_error_estimate;
 
  int64_t vbr_bits_off_target;
 
  int64_t vbr_bits_off_target_fast;
 
  int64_t bits_off_target;
 
  int rolling_target_bits;
 
  int rolling_actual_bits;
 
  int q_history[MAX_Q_HISTORY];
} PRIMARY_RATE_CONTROL;
 
struct AV1_COMP;
struct AV1EncoderConfig;
struct GF_GROUP;
 
void av1_primary_rc_init(const struct AV1EncoderConfig *oxcf,
                         PRIMARY_RATE_CONTROL *p_rc);
 
void av1_rc_init(const struct AV1EncoderConfig *oxcf, RATE_CONTROL *rc);
 
int av1_estimate_bits_at_q(FRAME_TYPE frame_kind, int q, int mbs,
                           double correction_factor, aom_bit_depth_t bit_depth,
                           const int is_screen_content_type);
 
double av1_convert_qindex_to_q(int qindex, aom_bit_depth_t bit_depth);
 
void av1_rc_init_minq_luts(void);
 
int av1_rc_get_default_min_gf_interval(int width, int height, double framerate);
// Note av1_rc_get_default_max_gf_interval() requires the min_gf_interval to
// be passed in to ensure that the max_gf_interval returned is at least as bis
// as that.
int av1_rc_get_default_max_gf_interval(double framerate, int min_gf_interval);
 
// Generally at the high level, the following flow is expected
// to be enforced for rate control:
// First call per frame, one of:
//   av1_get_one_pass_rt_params()
//   av1_get_second_pass_params()
// depending on the usage to set the rate control encode parameters desired.
//
// Then, call encode_frame_to_data_rate() to perform the
// actual encode. This function will in turn call encode_frame()
// one or more times, followed by:
//   av1_rc_postencode_update_drop_frame()
//
// The majority of rate control parameters are only expected
// to be set in the av1_get_..._params() functions and
// updated during the av1_rc_postencode_update...() functions.
// The only exceptions are av1_rc_drop_frame() and
// av1_rc_update_rate_correction_factors() functions.
 
// Functions to set parameters for encoding before the actual
// encode_frame_to_data_rate() function.
struct EncodeFrameParams;
 
// Post encode update of the rate control parameters based
// on bytes used
void av1_rc_postencode_update(struct AV1_COMP *cpi, uint64_t bytes_used);
// Post encode update of the rate control parameters for dropped frames
void av1_rc_postencode_update_drop_frame(struct AV1_COMP *cpi);
 
void av1_rc_update_rate_correction_factors(struct AV1_COMP *cpi,
                                           int is_encode_stage, int width,
                                           int height);
// Decide if we should drop this frame: For 1-pass CBR.
// Changes only the decimation count in the rate control structure
int av1_rc_drop_frame(struct AV1_COMP *cpi);
 
// Computes frame size bounds.
void av1_rc_compute_frame_size_bounds(const struct AV1_COMP *cpi,
                                      int this_frame_target,
                                      int *frame_under_shoot_limit,
                                      int *frame_over_shoot_limit);
 
int av1_rc_pick_q_and_bounds(const struct AV1_COMP *cpi, int width, int height,
                             int gf_index, int *bottom_index, int *top_index);
 
int av1_rc_regulate_q(const struct AV1_COMP *cpi, int target_bits_per_frame,
                      int active_best_quality, int active_worst_quality,
                      int width, int height);
 
// Estimates bits per mb for a given qindex and correction factor.
int av1_rc_bits_per_mb(FRAME_TYPE frame_type, int qindex,
                       double correction_factor, aom_bit_depth_t bit_depth,
                       const int is_screen_content_type);
 
// Clamping utilities for bitrate targets for iframes and pframes.
int av1_rc_clamp_iframe_target_size(const struct AV1_COMP *const cpi,
                                    int64_t target);
int av1_rc_clamp_pframe_target_size(const struct AV1_COMP *const cpi,
                                    int target, uint8_t frame_update_type);
 
// Find q_index corresponding to desired_q, within [best_qindex, worst_qindex].
// To be precise, 'q_index' is the smallest integer, for which the corresponding
// q >= desired_q.
// If no such q index is found, returns 'worst_qindex'.
int av1_find_qindex(double desired_q, aom_bit_depth_t bit_depth,
                    int best_qindex, int worst_qindex);
 
// Computes a q delta (in "q index" terms) to get from a starting q value
// to a target q value
int av1_compute_qdelta(const RATE_CONTROL *rc, double qstart, double qtarget,
                       aom_bit_depth_t bit_depth);
 
// Computes a q delta (in "q index" terms) to get from a starting q value
// to a value that should equate to the given rate ratio.
int av1_compute_qdelta_by_rate(const RATE_CONTROL *rc, FRAME_TYPE frame_type,
                               int qindex, double rate_target_ratio,
                               const int is_screen_content_type,
                               aom_bit_depth_t bit_depth);
 
int av1_frame_type_qdelta(const struct AV1_COMP *cpi, int q);
 
void av1_rc_update_framerate(struct AV1_COMP *cpi, int width, int height);
 
void av1_rc_set_gf_interval_range(const struct AV1_COMP *const cpi,
                                  RATE_CONTROL *const rc);
 
void av1_set_target_rate(struct AV1_COMP *cpi, int width, int height);
 
int av1_resize_one_pass_cbr(struct AV1_COMP *cpi);
 
void av1_rc_set_frame_target(struct AV1_COMP *cpi, int target, int width,
                             int height);
 
void av1_adjust_gf_refresh_qp_one_pass_rt(struct AV1_COMP *cpi);
 
void av1_set_reference_structure_one_pass_rt(struct AV1_COMP *cpi,
                                             int gf_update);
 
int av1_calc_pframe_target_size_one_pass_vbr(
    const struct AV1_COMP *const cpi, FRAME_UPDATE_TYPE frame_update_type);
 
int av1_calc_iframe_target_size_one_pass_vbr(const struct AV1_COMP *const cpi);
 
int av1_calc_pframe_target_size_one_pass_cbr(
    const struct AV1_COMP *cpi, FRAME_UPDATE_TYPE frame_update_type);
 
int av1_calc_iframe_target_size_one_pass_cbr(const struct AV1_COMP *cpi);
 
void av1_get_one_pass_rt_params(struct AV1_COMP *cpi,
                                struct EncodeFrameParams *const frame_params,
                                unsigned int frame_flags);
 
int av1_encodedframe_overshoot_cbr(struct AV1_COMP *cpi, int *q);
 
int av1_q_mode_get_q_index(int base_q_index, int gf_update_type,
                           int gf_pyramid_level, int arf_q);
 
int av1_get_arf_q_index(int base_q_index, int gfu_boost, int bit_depth,
                        double arf_boost_factor);
 
#if !CONFIG_REALTIME_ONLY
struct TplDepFrame;
int av1_get_arf_q_index_q_mode(struct AV1_COMP *cpi,
                               struct TplDepFrame *tpl_frame);
#endif
#ifdef __cplusplus
}  // extern "C"
#endif
 
#endif  // AOM_AV1_ENCODER_RATECTRL_H_