Changes to the RD and Rate control code
These changes are based on experiments in VP9.
The changes give solid metrics gains for SSIM but are mixed
for average and overall psnr (down a little for lower resolutions
but better for HD).
Figures are % BDRATE changes for average psnr,
overall psnr and SSIM with -ve being better.
lowres2 -0.069 0.051 -0.568
ugc360p 0.068 0.248 -0.187
midres2 -0.038 0.037 -0.581
hdres -0.550 -0.483 -1.140
hdres2 -0.220 -0.125 -0.676
STATS_CHANGED
Change-Id: I828ed8a826f9bdc60996350f01fb97e8d3f30834
diff --git a/av1/encoder/encoder.c b/av1/encoder/encoder.c
index ba03642..eca2ff4 100644
--- a/av1/encoder/encoder.c
+++ b/av1/encoder/encoder.c
@@ -2933,8 +2933,8 @@
if (err != AOM_CODEC_OK) return err;
// Note: Both use common rdmult based on base qindex of fullres.
- const int64_t rdmult = av1_compute_rd_mult_based_on_qindex(
- cm->seq_params->bit_depth, cm->quant_params.base_qindex);
+ const int64_t rdmult =
+ av1_compute_rd_mult_based_on_qindex(cpi, cm->quant_params.base_qindex);
// Find the best rdcost among all superres denoms.
int best_denom = -1;
@@ -2997,8 +2997,8 @@
if (err != AOM_CODEC_OK) return err;
// Note: Both use common rdmult based on base qindex of fullres.
- const int64_t rdmult = av1_compute_rd_mult_based_on_qindex(
- cm->seq_params->bit_depth, cm->quant_params.base_qindex);
+ const int64_t rdmult =
+ av1_compute_rd_mult_based_on_qindex(cpi, cm->quant_params.base_qindex);
proj_rdcost1 = RDCOST_DBL_WITH_NATIVE_BD_DIST(rdmult, rate1, sse1,
cm->seq_params->bit_depth);
const double proj_rdcost2 = RDCOST_DBL_WITH_NATIVE_BD_DIST(
diff --git a/av1/encoder/pass2_strategy.c b/av1/encoder/pass2_strategy.c
index 2937034..f7e03db 100644
--- a/av1/encoder/pass2_strategy.c
+++ b/av1/encoder/pass2_strategy.c
@@ -312,57 +312,63 @@
}
}
-#define SR_DIFF_PART 0.0015
-#define MOTION_AMP_PART 0.003
#define INTRA_PART 0.005
#define DEFAULT_DECAY_LIMIT 0.75
#define LOW_SR_DIFF_TRHESH 0.1
-#define SR_DIFF_MAX 128.0
#define NCOUNT_FRAME_II_THRESH 5.0
+#define LOW_CODED_ERR_PER_MB 10.0
-static double get_sr_decay_rate(const FRAME_INFO *frame_info,
- const FIRSTPASS_STATS *frame) {
- const int num_mbs = frame_info->num_mbs;
- double sr_diff = (frame->sr_coded_error - frame->coded_error) / num_mbs;
+/* This function considers how the quality of prediction may be deteriorating
+ * with distance. It comapres the coded error for the last frame and the
+ * second reference frame (usually two frames old) and also applies a factor
+ * based on the extent of INTRA coding.
+ *
+ * The decay factor is then used to reduce the contribution of frames further
+ * from the alt-ref or golden frame, to the bitframe boost calculation for that
+ * alt-ref or golden frame.
+ */
+static double get_sr_decay_rate(const FIRSTPASS_STATS *frame) {
+ double sr_diff = (frame->sr_coded_error - frame->coded_error);
double sr_decay = 1.0;
double modified_pct_inter;
double modified_pcnt_intra;
- const double motion_amplitude_factor =
- frame->pcnt_motion * ((frame->mvc_abs + frame->mvr_abs) / 2);
modified_pct_inter = frame->pcnt_inter;
- if ((frame->intra_error / DOUBLE_DIVIDE_CHECK(frame->coded_error)) <
- (double)NCOUNT_FRAME_II_THRESH) {
+ if ((frame->coded_error > LOW_CODED_ERR_PER_MB) &&
+ ((frame->intra_error / DOUBLE_DIVIDE_CHECK(frame->coded_error)) <
+ (double)NCOUNT_FRAME_II_THRESH)) {
modified_pct_inter = frame->pcnt_inter - frame->pcnt_neutral;
}
modified_pcnt_intra = 100 * (1.0 - modified_pct_inter);
if ((sr_diff > LOW_SR_DIFF_TRHESH)) {
- sr_diff = AOMMIN(sr_diff, SR_DIFF_MAX);
- sr_decay = 1.0 - (SR_DIFF_PART * sr_diff) -
- (MOTION_AMP_PART * motion_amplitude_factor) -
- (INTRA_PART * modified_pcnt_intra);
+ double sr_diff_part = ((sr_diff * 0.25) / frame->intra_error);
+ sr_decay = 1.0 - sr_diff_part - (INTRA_PART * modified_pcnt_intra);
}
- return AOMMAX(sr_decay, AOMMIN(DEFAULT_DECAY_LIMIT, modified_pct_inter));
+ return AOMMAX(sr_decay, DEFAULT_DECAY_LIMIT);
}
// This function gives an estimate of how badly we believe the prediction
// quality is decaying from frame to frame.
-static double get_zero_motion_factor(const FRAME_INFO *frame_info,
- const FIRSTPASS_STATS *frame) {
+static double get_zero_motion_factor(const FIRSTPASS_STATS *frame) {
const double zero_motion_pct = frame->pcnt_inter - frame->pcnt_motion;
- double sr_decay = get_sr_decay_rate(frame_info, frame);
+ double sr_decay = get_sr_decay_rate(frame);
return AOMMIN(sr_decay, zero_motion_pct);
}
-#define ZM_POWER_FACTOR 0.75
+#define DEFAULT_ZM_FACTOR 0.5
+static double get_prediction_decay_rate(const FIRSTPASS_STATS *frame_stats) {
+ const double sr_decay_rate = get_sr_decay_rate(frame_stats);
+ double zero_motion_factor =
+ DEFAULT_ZM_FACTOR * (frame_stats->pcnt_inter - frame_stats->pcnt_motion);
-static double get_prediction_decay_rate(const FRAME_INFO *frame_info,
- const FIRSTPASS_STATS *next_frame) {
- const double sr_decay_rate = get_sr_decay_rate(frame_info, next_frame);
- const double zero_motion_factor =
- (0.95 * pow((next_frame->pcnt_inter - next_frame->pcnt_motion),
- ZM_POWER_FACTOR));
+ // Clamp value to range 0.0 to 1.0
+ // This should happen anyway if input values are sensibly clamped but checked
+ // here just in case.
+ if (zero_motion_factor > 1.0)
+ zero_motion_factor = 1.0;
+ else if (zero_motion_factor < 0.0)
+ zero_motion_factor = 0.0;
return AOMMAX(zero_motion_factor,
(sr_decay_rate + ((1.0 - sr_decay_rate) * zero_motion_factor)));
@@ -449,7 +455,6 @@
}
static void accumulate_next_frame_stats(const FIRSTPASS_STATS *stats,
- const FRAME_INFO *frame_info,
const int flash_detected,
const int frames_since_key,
const int cur_idx,
@@ -470,16 +475,15 @@
// Accumulate the effect of prediction quality decay
if (!flash_detected) {
gf_stats->last_loop_decay_rate = gf_stats->loop_decay_rate;
- gf_stats->loop_decay_rate = get_prediction_decay_rate(frame_info, stats);
+ gf_stats->loop_decay_rate = get_prediction_decay_rate(stats);
gf_stats->decay_accumulator =
gf_stats->decay_accumulator * gf_stats->loop_decay_rate;
// Monitor for static sections.
if ((frames_since_key + cur_idx - 1) > 1) {
- gf_stats->zero_motion_accumulator =
- AOMMIN(gf_stats->zero_motion_accumulator,
- get_zero_motion_factor(frame_info, stats));
+ gf_stats->zero_motion_accumulator = AOMMIN(
+ gf_stats->zero_motion_accumulator, get_zero_motion_factor(stats));
}
}
}
@@ -672,8 +676,7 @@
// Accumulate the effect of prediction quality decay.
if (!flash_detected) {
- gf_stats.decay_accumulator *=
- get_prediction_decay_rate(frame_info, this_frame);
+ gf_stats.decay_accumulator *= get_prediction_decay_rate(this_frame);
gf_stats.decay_accumulator = gf_stats.decay_accumulator < MIN_DECAY_FACTOR
? MIN_DECAY_FACTOR
: gf_stats.decay_accumulator;
@@ -706,8 +709,7 @@
// Cumulative effect of prediction quality decay.
if (!flash_detected) {
- gf_stats.decay_accumulator *=
- get_prediction_decay_rate(frame_info, this_frame);
+ gf_stats.decay_accumulator *= get_prediction_decay_rate(this_frame);
gf_stats.decay_accumulator = gf_stats.decay_accumulator < MIN_DECAY_FACTOR
? MIN_DECAY_FACTOR
: gf_stats.decay_accumulator;
@@ -1870,7 +1872,6 @@
FIRSTPASS_STATS next_frame;
const FIRSTPASS_STATS *const start_pos = twopass->stats_in;
const FIRSTPASS_STATS *const stats = start_pos - (rc->frames_since_key == 0);
- FRAME_INFO *frame_info = &cpi->frame_info;
int i;
int flash_detected;
@@ -1919,7 +1920,7 @@
flash_detected = detect_flash(twopass, 0);
// TODO(bohanli): remove redundant accumulations here, or unify
// this and the ones in define_gf_group
- accumulate_next_frame_stats(&next_frame, frame_info, flash_detected,
+ accumulate_next_frame_stats(&next_frame, flash_detected,
rc->frames_since_key, i, &gf_stats);
cut_here = detect_gf_cut(cpi, i, cur_start, flash_detected,
@@ -2375,7 +2376,7 @@
flash_detected = detect_flash(twopass, 0);
// accumulate stats for next frame
- accumulate_next_frame_stats(&next_frame, frame_info, flash_detected,
+ accumulate_next_frame_stats(&next_frame, flash_detected,
rc->frames_since_key, i, &gf_stats);
++i;
@@ -2896,8 +2897,7 @@
}
// How fast is the prediction quality decaying?
- loop_decay_rate =
- get_prediction_decay_rate(frame_info, twopass->stats_in);
+ loop_decay_rate = get_prediction_decay_rate(twopass->stats_in);
// We want to know something about the recent past... rather than
// as used elsewhere where we are concerned with decay in prediction
@@ -3069,8 +3069,7 @@
// For the first frame in kf group, the second ref indicator is invalid.
if (i > 0) {
*zero_motion_accumulator =
- AOMMIN(*zero_motion_accumulator,
- get_zero_motion_factor(frame_info, &frame_stat));
+ AOMMIN(*zero_motion_accumulator, get_zero_motion_factor(&frame_stat));
} else {
*zero_motion_accumulator = frame_stat.pcnt_inter - frame_stat.pcnt_motion;
}
diff --git a/av1/encoder/rd.c b/av1/encoder/rd.c
index 33810d0..f6a5d93 100644
--- a/av1/encoder/rd.c
+++ b/av1/encoder/rd.c
@@ -354,11 +354,45 @@
160, 160, 160, 160, 192, 208, 224
};
-int av1_compute_rd_mult_based_on_qindex(aom_bit_depth_t bit_depth, int qindex) {
- const int q = av1_dc_quant_QTX(qindex, 0, bit_depth);
- int rdmult = (int)(((int64_t)88 * q * q) / 24);
+// Returns the default rd multiplier for inter frames for a given qindex.
+// The function here is a first pass estimate based on data from
+// a previous Vizer run
+static double def_inter_rd_multiplier(int qindex) {
+ return 3.2 + (0.0035 * (double)qindex);
+}
- switch (bit_depth) {
+// Returns the default rd multiplier for ARF/Golden Frames for a given qindex.
+// The function here is a first pass estimate based on data from
+// a previous Vizer run
+static double def_arf_rd_multiplier(int qindex) {
+ return 3.25 + (0.0035 * (double)qindex);
+}
+
+// Returns the default rd multiplier for key frames for a given qindex.
+// The function here is a first pass estimate based on data from
+// a previous Vizer run
+static double def_kf_rd_multiplier(int qindex) {
+ return 3.3 + (0.0035 * (double)qindex);
+}
+
+int av1_compute_rd_mult_based_on_qindex(const AV1_COMP *cpi, int qindex) {
+ const int q = av1_dc_quant_QTX(qindex, 0, cpi->common.seq_params->bit_depth);
+ const FRAME_UPDATE_TYPE update_type =
+ cpi->ppi->gf_group.update_type[cpi->gf_frame_index];
+ int rdmult = q * q;
+
+ if (update_type == KF_UPDATE) {
+ double def_rd_q_mult = def_kf_rd_multiplier(qindex);
+ rdmult = (int)((double)rdmult * def_rd_q_mult);
+ } else if ((update_type == GF_UPDATE) || (update_type == ARF_UPDATE)) {
+ double def_rd_q_mult = def_arf_rd_multiplier(qindex);
+ rdmult = (int)((double)rdmult * def_rd_q_mult);
+ } else {
+ double def_rd_q_mult = def_inter_rd_multiplier(qindex);
+ rdmult = (int)((double)rdmult * def_rd_q_mult);
+ }
+
+ switch (cpi->common.seq_params->bit_depth) {
case AOM_BITS_8: break;
case AOM_BITS_10: rdmult = ROUND_POWER_OF_TWO(rdmult, 4); break;
case AOM_BITS_12: rdmult = ROUND_POWER_OF_TWO(rdmult, 8); break;
@@ -370,8 +404,7 @@
}
int av1_compute_rd_mult(const AV1_COMP *cpi, int qindex) {
- int64_t rdmult = av1_compute_rd_mult_based_on_qindex(
- cpi->common.seq_params->bit_depth, qindex);
+ int64_t rdmult = av1_compute_rd_mult_based_on_qindex(cpi, qindex);
if (is_stat_consumption_stage(cpi) &&
(cpi->common.current_frame.frame_type != KEY_FRAME)) {
const GF_GROUP *const gf_group = &cpi->ppi->gf_group;
diff --git a/av1/encoder/rd.h b/av1/encoder/rd.h
index 0b13a6a..c1ba819 100644
--- a/av1/encoder/rd.h
+++ b/av1/encoder/rd.h
@@ -191,7 +191,7 @@
struct AV1_COMP;
struct macroblock;
-int av1_compute_rd_mult_based_on_qindex(aom_bit_depth_t bit_depth, int qindex);
+int av1_compute_rd_mult_based_on_qindex(const struct AV1_COMP *cpi, int qindex);
int av1_compute_rd_mult(const struct AV1_COMP *cpi, int qindex);
diff --git a/av1/encoder/tpl_model.c b/av1/encoder/tpl_model.c
index cde3b47..64318cd 100644
--- a/av1/encoder/tpl_model.c
+++ b/av1/encoder/tpl_model.c
@@ -1141,9 +1141,8 @@
cm->quant_params.base_qindex = base_qindex;
av1_frame_init_quantizer(cpi);
- tpl_frame->base_rdmult = av1_compute_rd_mult_based_on_qindex(
- cm->seq_params->bit_depth, pframe_qindex) /
- 6;
+ tpl_frame->base_rdmult =
+ av1_compute_rd_mult_based_on_qindex(cpi, pframe_qindex) / 6;
av1_init_tpl_txfm_stats(tpl_txfm_stats);
}
