From e4f3b970ad10718fc3276eb3e514dbdf149e37bf Mon Sep 17 00:00:00 2001
From: Guillaume Martres <gmartres@google.com>
Date: Thu, 14 Nov 2013 19:23:57 +0100
Subject: [PATCH] vpxenc: add --aq-mode flag to control adaptive quantization
Change-Id: I2a08c00e8576099abc84b6ef05cb3567426e29cf
---
vp9/common/vp9_onyx.h | 7 +++++++
vp9/encoder/vp9_encodeframe.c | 10 +++++-----
vp9/encoder/vp9_firstpass.c | 10 +++++-----
vp9/encoder/vp9_onyx_if.c | 6 ++----
vp9/encoder/vp9_onyx_int.h | 1 -
vp9/vp9_cx_iface.c | 7 +++++++
vpx/vp8cx.h | 3 +++
vpxenc.c | 7 +++++--
8 files changed, 34 insertions(+), 17 deletions(-)
diff --git a/vp9/common/vp9_onyx.h b/vp9/common/vp9_onyx.h
index 452dd6b89c..cda68a2851 100644
--- a/vp9/common/vp9_onyx.h
+++ b/vp9/common/vp9_onyx.h
@@ -64,6 +64,12 @@ extern "C"
FRAMEFLAGS_ALTREF = 4,
} FRAMETYPE_FLAGS;
+ typedef enum {
+ NO_AQ = 0,
+ VARIANCE_AQ = 1,
+ AQ_MODES_COUNT // This should always be the last member of the enum
+ } AQ_MODES;
+
typedef struct {
int version; // 4 versions of bitstream defined:
// 0 - best quality/slowest decode,
@@ -128,6 +134,7 @@ extern "C"
int best_allowed_q;
int cq_level;
int lossless;
+ int aq_mode; // Adaptive Quantization mode
// two pass datarate control
int two_pass_vbrbias; // two pass datarate control tweaks
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index 3e75f3b285..8e71d6b574 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -409,7 +409,7 @@ static void update_state(VP9_COMP *cpi, PICK_MODE_CONTEXT *ctx,
&& (xd->mb_to_bottom_edge >> (3 + MI_SIZE_LOG2)) + mi_height > y)
xd->mi_8x8[x_idx + y * mis] = mi_addr;
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_mb_init_quantizer(cpi, x);
}
@@ -558,7 +558,7 @@ static void set_offsets(VP9_COMP *cpi, const TileInfo *const tile,
/* segment ID */
if (seg->enabled) {
- if (!cpi->sf.variance_adaptive_quantization) {
+ if (!cpi->oxcf.aq_mode == VARIANCE_AQ) {
uint8_t *map = seg->update_map ? cpi->segmentation_map
: cm->last_frame_seg_map;
mbmi->segment_id = vp9_get_segment_id(cm, map, bsize, mi_row, mi_col);
@@ -635,7 +635,7 @@ static void pick_sb_modes(VP9_COMP *cpi, const TileInfo *const tile,
x->source_variance = get_sby_perpixel_variance(cpi, x, bsize);
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
int energy;
if (bsize <= BLOCK_16X16) {
energy = x->mb_energy;
@@ -651,7 +651,7 @@ static void pick_sb_modes(VP9_COMP *cpi, const TileInfo *const tile,
if (cpi->oxcf.tuning == VP8_TUNE_SSIM)
vp9_activity_masking(cpi, x);
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_clear_system_state(); // __asm emms;
x->rdmult = round(x->rdmult * rdmult_ratio);
}
@@ -670,7 +670,7 @@ static void pick_sb_modes(VP9_COMP *cpi, const TileInfo *const tile,
totaldist, bsize, ctx, best_rd);
}
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
x->rdmult = orig_rdmult;
if (*totalrate != INT_MAX) {
vp9_clear_system_state(); // __asm emms;
diff --git a/vp9/encoder/vp9_firstpass.c b/vp9/encoder/vp9_firstpass.c
index 974c300e6f..791bc9ef77 100644
--- a/vp9/encoder/vp9_firstpass.c
+++ b/vp9/encoder/vp9_firstpass.c
@@ -602,14 +602,14 @@ void vp9_first_pass(VP9_COMP *cpi) {
num_8x8_blocks_wide_lookup[xd->mi_8x8[0]->mbmi.sb_type],
cm->mi_rows, cm->mi_cols);
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
int energy = vp9_block_energy(cpi, x, xd->mi_8x8[0]->mbmi.sb_type);
error_weight = vp9_vaq_inv_q_ratio(energy);
}
// do intra 16x16 prediction
this_error = vp9_encode_intra(x, use_dc_pred);
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_clear_system_state(); // __asm emms;
this_error *= error_weight;
}
@@ -647,7 +647,7 @@ void vp9_first_pass(VP9_COMP *cpi) {
first_pass_motion_search(cpi, x, &best_ref_mv,
&mv.as_mv, lst_yv12,
&motion_error, recon_yoffset);
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_clear_system_state(); // __asm emms;
motion_error *= error_weight;
}
@@ -658,7 +658,7 @@ void vp9_first_pass(VP9_COMP *cpi) {
tmp_err = INT_MAX;
first_pass_motion_search(cpi, x, &zero_ref_mv, &tmp_mv.as_mv,
lst_yv12, &tmp_err, recon_yoffset);
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_clear_system_state(); // __asm emms;
tmp_err *= error_weight;
}
@@ -678,7 +678,7 @@ void vp9_first_pass(VP9_COMP *cpi) {
first_pass_motion_search(cpi, x, &zero_ref_mv,
&tmp_mv.as_mv, gld_yv12,
&gf_motion_error, recon_yoffset);
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_clear_system_state(); // __asm emms;
gf_motion_error *= error_weight;
}
diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index dd4705da05..b7874d5152 100644
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -764,8 +764,6 @@ void vp9_set_speed_features(VP9_COMP *cpi) {
sf->static_segmentation = 0;
#endif
- sf->variance_adaptive_quantization = 0;
-
switch (mode) {
case 0: // This is the best quality mode.
break;
@@ -3188,7 +3186,7 @@ static void encode_frame_to_data_rate(VP9_COMP *cpi,
}
}
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_vaq_frame_setup(cpi);
}
@@ -3975,7 +3973,7 @@ int vp9_get_compressed_data(VP9_PTR ptr, unsigned int *frame_flags,
vp9_setup_interp_filters(&cpi->mb.e_mbd, DEFAULT_INTERP_FILTER, cm);
- if (cpi->sf.variance_adaptive_quantization) {
+ if (cpi->oxcf.aq_mode == VARIANCE_AQ) {
vp9_vaq_init();
}
diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h
index 9e802123c7..c098b5ed84 100644
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -248,7 +248,6 @@ typedef struct {
int auto_mv_step_size;
int optimize_coefficients;
int static_segmentation;
- int variance_adaptive_quantization;
int comp_inter_joint_search_thresh;
int adaptive_rd_thresh;
int skip_encode_sb;
diff --git a/vp9/vp9_cx_iface.c b/vp9/vp9_cx_iface.c
index 1942039679..9a23ebd539 100644
--- a/vp9/vp9_cx_iface.c
+++ b/vp9/vp9_cx_iface.c
@@ -38,6 +38,7 @@ struct vp9_extracfg {
unsigned int rc_max_intra_bitrate_pct;
unsigned int lossless;
unsigned int frame_parallel_decoding_mode;
+ unsigned int aq_mode;
};
struct extraconfig_map {
@@ -66,6 +67,7 @@ static const struct extraconfig_map extracfg_map[] = {
0, /* rc_max_intra_bitrate_pct */
0, /* lossless */
0, /* frame_parallel_decoding_mode */
+ 0, /* aq_mode */
}
}
};
@@ -157,6 +159,7 @@ static vpx_codec_err_t validate_config(vpx_codec_alg_priv_t *ctx,
RANGE_CHECK_HI(cfg, rc_max_quantizer, 0);
RANGE_CHECK_HI(cfg, rc_min_quantizer, 0);
}
+ RANGE_CHECK(vp8_cfg, aq_mode, 0, AQ_MODES_COUNT - 1);
RANGE_CHECK_HI(cfg, g_threads, 64);
RANGE_CHECK_HI(cfg, g_lag_in_frames, MAX_LAG_BUFFERS);
@@ -335,6 +338,8 @@ static vpx_codec_err_t set_vp9e_config(VP9_CONFIG *oxcf,
oxcf->error_resilient_mode = cfg.g_error_resilient;
oxcf->frame_parallel_decoding_mode = vp8_cfg.frame_parallel_decoding_mode;
+ oxcf->aq_mode = vp8_cfg.aq_mode;
+
oxcf->ss_number_layers = cfg.ss_number_layers;
/*
printf("Current VP9 Settings: \n");
@@ -445,6 +450,7 @@ static vpx_codec_err_t set_param(vpx_codec_alg_priv_t *ctx,
MAP(VP8E_SET_MAX_INTRA_BITRATE_PCT, xcfg.rc_max_intra_bitrate_pct);
MAP(VP9E_SET_LOSSLESS, xcfg.lossless);
MAP(VP9E_SET_FRAME_PARALLEL_DECODING, xcfg.frame_parallel_decoding_mode);
+ MAP(VP9E_SET_AQ_MODE, xcfg.aq_mode);
}
res = validate_config(ctx, &ctx->cfg, &xcfg);
@@ -1071,6 +1077,7 @@ static vpx_codec_ctrl_fn_map_t vp9e_ctf_maps[] = {
{VP8E_SET_MAX_INTRA_BITRATE_PCT, set_param},
{VP9E_SET_LOSSLESS, set_param},
{VP9E_SET_FRAME_PARALLEL_DECODING, set_param},
+ {VP9E_SET_AQ_MODE, set_param},
{VP9_GET_REFERENCE, get_reference},
{VP9E_SET_SVC, vp9e_set_svc},
{VP9E_SET_SVC_PARAMETERS, vp9e_set_svc_parameters},
diff --git a/vpx/vp8cx.h b/vpx/vp8cx.h
index 433cc0d8ac..c0424f1466 100644
--- a/vpx/vp8cx.h
+++ b/vpx/vp8cx.h
@@ -193,6 +193,7 @@ enum vp8e_enc_control_id {
VP9E_SET_TILE_COLUMNS,
VP9E_SET_TILE_ROWS,
VP9E_SET_FRAME_PARALLEL_DECODING,
+ VP9E_SET_AQ_MODE,
VP9E_SET_SVC,
VP9E_SET_SVC_PARAMETERS
@@ -343,6 +344,8 @@ VPX_CTRL_USE_TYPE(VP9E_SET_LOSSLESS, unsigned int)
VPX_CTRL_USE_TYPE(VP9E_SET_FRAME_PARALLEL_DECODING, unsigned int)
+VPX_CTRL_USE_TYPE(VP9E_SET_AQ_MODE, unsigned int)
+
/*! @} - end defgroup vp8_encoder */
#ifdef __cplusplus
} // extern "C"
diff --git a/vpxenc.c b/vpxenc.c
index b7897dbf3f..674da14900 100644
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -561,6 +561,9 @@ static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
#if CONFIG_VP9_ENCODER
static const arg_def_t frame_parallel_decoding = ARG_DEF(
NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
+static const arg_def_t aq_mode = ARG_DEF(
+ NULL, "aq-mode", 1,
+ "Adaptive quantization mode (0: disabled (by default), 1: variance based)");
#endif
#if CONFIG_VP8_ENCODER
@@ -585,7 +588,7 @@ static const arg_def_t *vp9_args[] = {
&cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
&tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
&tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
- &frame_parallel_decoding,
+ &frame_parallel_decoding, &aq_mode,
NULL
};
static const int vp9_arg_ctrl_map[] = {
@@ -594,7 +597,7 @@ static const int vp9_arg_ctrl_map[] = {
VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
- VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING,
+ VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
0
};
#endif
--
GitLab