Commit 5715c39c authored by Christian Duvivier's avatar Christian Duvivier Committed by John Koleszar
Browse files

Inline Intrinsic optimized Denoiser

Faster version of denoiser, cut cost by 1.7x for C path, by 3.3x for
SSE2 path.

Change-Id: I154786308550763bc0e3497e5fa5bfd1ce651beb
parent 48908530
......@@ -501,6 +501,12 @@ fi
prototype void vp8_yv12_copy_partial_frame "struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc"
specialize vp8_yv12_copy_partial_frame neon
#
# Denoiser filter
#
prototype void vp8_denoiser_filter "struct yv12_buffer_config* mc_running_avg, struct yv12_buffer_config* running_avg, struct macroblock* signal, unsigned int motion_magnitude2, int y_offset, int uv_offset"
specialize vp8_denoiser_filter sse2
# End of encoder only functions
fi
......
......@@ -15,23 +15,16 @@
#include "vpx_mem/vpx_mem.h"
#include "vpx_rtcd.h"
static const unsigned int NOISE_MOTION_THRESHOLD = 20*20;
static const unsigned int NOISE_DIFF2_THRESHOLD = 75;
static const unsigned int NOISE_MOTION_THRESHOLD = 20 * 20;
// SSE_DIFF_THRESHOLD is selected as ~95% confidence assuming var(noise) ~= 100.
static const unsigned int SSE_DIFF_THRESHOLD = 16*16*20;
static const unsigned int SSE_THRESHOLD = 16*16*40;
static const unsigned int SSE_DIFF_THRESHOLD = 16 * 16 * 20;
static const unsigned int SSE_THRESHOLD = 16 * 16 * 40;
static uint8_t blend(uint8_t state, uint8_t sample, uint8_t factor_q8)
{
return (uint8_t)(
(((uint16_t)factor_q8 * ((uint16_t)state) + // Q8
(uint16_t)(256 - factor_q8) * ((uint16_t)sample)) + 128) // Q8
>> 8);
}
static unsigned int denoiser_motion_compensate(YV12_BUFFER_CONFIG* src,
YV12_BUFFER_CONFIG* dst,
MACROBLOCK* x,
static unsigned int denoiser_motion_compensate(YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst,
MACROBLOCK *x,
unsigned int best_sse,
unsigned int zero_mv_sse,
int recon_yoffset,
......@@ -57,14 +50,15 @@ static unsigned int denoiser_motion_compensate(YV12_BUFFER_CONFIG* src,
filter_xd.need_to_clamp_best_mvs;
mv_col = filter_xd.best_sse_mv.as_mv.col;
mv_row = filter_xd.best_sse_mv.as_mv.row;
if (filter_xd.mode_info_context->mbmi.mode <= B_PRED ||
(mv_row*mv_row + mv_col*mv_col <= NOISE_MOTION_THRESHOLD &&
(mv_row *mv_row + mv_col *mv_col <= NOISE_MOTION_THRESHOLD &&
sse_diff < SSE_DIFF_THRESHOLD))
{
// Handle intra blocks as referring to last frame with zero motion and
// let the absolute pixel difference affect the filter factor.
// Also consider small amount of motion as being random walk due to noise,
// if it doesn't mean that we get a much bigger error.
// Also consider small amount of motion as being random walk due to
// noise, if it doesn't mean that we get a much bigger error.
// Note that any changes to the mode info only affects the denoising.
filter_xd.mode_info_context->mbmi.ref_frame = LAST_FRAME;
filter_xd.mode_info_context->mbmi.mode = ZEROMV;
......@@ -73,6 +67,7 @@ static unsigned int denoiser_motion_compensate(YV12_BUFFER_CONFIG* src,
x->e_mbd.best_sse_mv.as_int = 0;
best_sse = zero_mv_sse;
}
if (!x->skip)
{
vp8_build_inter_predictors_mb(&filter_xd);
......@@ -86,77 +81,160 @@ static unsigned int denoiser_motion_compensate(YV12_BUFFER_CONFIG* src,
filter_xd.dst.y_stride,
filter_xd.dst.uv_stride);
}
return best_sse;
}
static void denoiser_filter(YV12_BUFFER_CONFIG* mc_running_avg,
YV12_BUFFER_CONFIG* running_avg,
MACROBLOCK* signal,
unsigned int motion_magnitude2,
int y_offset,
// The filtering coefficients used for denoizing are adjusted for static
// blocks, or blocks with very small motion vectors. This is done through
// the motion magnitude parameter.
//
// There are currently 2048 possible mapping from absolute difference to
// filter coefficient depending on the motion magnitude. Each mapping is
// in a LUT table. All these tables are staticly allocated but they are only
// filled on their first use.
//
// Each entry is a pair of 16b values, the coefficient and its complement
// to 256. Each of these value should only be 8b but they are 16b wide to
// avoid slow partial register manipulations.
enum {num_motion_magnitude_adjustments = 2048};
static union coeff_pair filter_coeff_LUT[num_motion_magnitude_adjustments][256];
static uint8_t filter_coeff_LUT_initialized[num_motion_magnitude_adjustments] =
{ 0 };
union coeff_pair *vp8_get_filter_coeff_LUT(unsigned int motion_magnitude)
{
union coeff_pair *LUT;
unsigned int motion_magnitude_adjustment = motion_magnitude >> 3;
if (motion_magnitude_adjustment >= num_motion_magnitude_adjustments)
{
motion_magnitude_adjustment = num_motion_magnitude_adjustments - 1;
}
LUT = filter_coeff_LUT[motion_magnitude_adjustment];
if (!filter_coeff_LUT_initialized[motion_magnitude_adjustment])
{
int absdiff;
for (absdiff = 0; absdiff < 256; ++absdiff)
{
unsigned int filter_coefficient;
filter_coefficient = (255 << 8) / (256 + ((absdiff * 330) >> 3));
filter_coefficient += filter_coefficient /
(3 + motion_magnitude_adjustment);
if (filter_coefficient > 255)
{
filter_coefficient = 255;
}
LUT[absdiff].as_short[0] = filter_coefficient ;
LUT[absdiff].as_short[1] = 256 - filter_coefficient;
}
filter_coeff_LUT_initialized[motion_magnitude_adjustment] = 1;
}
return LUT;
}
void vp8_denoiser_filter_c(YV12_BUFFER_CONFIG *mc_running_avg,
YV12_BUFFER_CONFIG *running_avg, MACROBLOCK *signal,
unsigned int motion_magnitude, int y_offset,
int uv_offset)
{
unsigned char* sig = signal->thismb;
unsigned char *sig = signal->thismb;
int sig_stride = 16;
unsigned char* mc_running_avg_y = mc_running_avg->y_buffer + y_offset;
unsigned char *mc_running_avg_y = mc_running_avg->y_buffer + y_offset;
int mc_avg_y_stride = mc_running_avg->y_stride;
unsigned char* running_avg_y = running_avg->y_buffer + y_offset;
unsigned char *running_avg_y = running_avg->y_buffer + y_offset;
int avg_y_stride = running_avg->y_stride;
const union coeff_pair *LUT = vp8_get_filter_coeff_LUT(motion_magnitude);
int r, c;
for (r = 0; r < 16; r++)
for (r = 0; r < 16; ++r)
{
for (c = 0; c < 16; c++)
// Calculate absolute differences
unsigned char abs_diff[16];
union coeff_pair filter_coefficient[16];
for (c = 0; c < 16; ++c)
{
int diff;
int absdiff = 0;
unsigned int filter_coefficient;
absdiff = sig[c] - mc_running_avg_y[c];
int absdiff = sig[c] - mc_running_avg_y[c];
absdiff = absdiff > 0 ? absdiff : -absdiff;
assert(absdiff >= 0 && absdiff < 256);
filter_coefficient = (255 << 8) / (256 + ((absdiff * 330) >> 3));
// Allow some additional filtering of static blocks, or blocks with very
// small motion vectors.
filter_coefficient += filter_coefficient / (3 + (motion_magnitude2 >> 3));
filter_coefficient = filter_coefficient > 255 ? 255 : filter_coefficient;
abs_diff[c] = absdiff;
}
// Use LUT to get filter coefficients (two 16b value; f and 256-f)
for (c = 0; c < 16; ++c)
{
filter_coefficient[c] = LUT[abs_diff[c]];
}
// Filtering...
for (c = 0; c < 16; ++c)
{
const uint16_t state = (uint16_t)(mc_running_avg_y[c]);
const uint16_t sample = (uint16_t)(sig[c]);
running_avg_y[c] = (filter_coefficient[c].as_short[0] * state +
filter_coefficient[c].as_short[1] * sample + 128) >> 8;
}
running_avg_y[c] = blend(mc_running_avg_y[c], sig[c], filter_coefficient);
diff = sig[c] - running_avg_y[c];
// Depending on the magnitude of the difference between the signal and
// filtered version, either replace the signal by the filtered one or
// update the filter state with the signal when the change in a pixel
// isn't classified as noise.
for (c = 0; c < 16; ++c)
{
const int diff = sig[c] - running_avg_y[c];
if (diff * diff < NOISE_DIFF2_THRESHOLD)
{
// Replace with mean to suppress the noise.
sig[c] = running_avg_y[c];
}
else
{
// Replace the filter state with the signal since the change in this
// pixel isn't classified as noise.
running_avg_y[c] = sig[c];
}
}
// Update pointers for next iteration.
sig += sig_stride;
mc_running_avg_y += mc_avg_y_stride;
running_avg_y += avg_y_stride;
}
}
int vp8_denoiser_allocate(VP8_DENOISER *denoiser, int width, int height)
{
assert(denoiser);
denoiser->yv12_running_avg.flags = 0;
if (vp8_yv12_alloc_frame_buffer(&(denoiser->yv12_running_avg), width,
height, VP8BORDERINPIXELS) < 0)
{
vp8_denoiser_free(denoiser);
return 1;
}
denoiser->yv12_mc_running_avg.flags = 0;
if (vp8_yv12_alloc_frame_buffer(&(denoiser->yv12_mc_running_avg), width,
height, VP8BORDERINPIXELS) < 0)
{
vp8_denoiser_free(denoiser);
return 1;
}
vpx_memset(denoiser->yv12_running_avg.buffer_alloc, 0,
denoiser->yv12_running_avg.frame_size);
vpx_memset(denoiser->yv12_mc_running_avg.buffer_alloc, 0,
......@@ -176,7 +254,8 @@ void vp8_denoiser_denoise_mb(VP8_DENOISER *denoiser,
unsigned int best_sse,
unsigned int zero_mv_sse,
int recon_yoffset,
int recon_uvoffset) {
int recon_uvoffset)
{
int mv_row;
int mv_col;
unsigned int motion_magnitude2;
......@@ -191,22 +270,21 @@ void vp8_denoiser_denoise_mb(VP8_DENOISER *denoiser,
mv_row = x->e_mbd.best_sse_mv.as_mv.row;
mv_col = x->e_mbd.best_sse_mv.as_mv.col;
motion_magnitude2 = mv_row*mv_row + mv_col*mv_col;
motion_magnitude2 = mv_row * mv_row + mv_col * mv_col;
if (best_sse > SSE_THRESHOLD ||
motion_magnitude2 > 8 * NOISE_MOTION_THRESHOLD)
{
// No filtering of this block since it differs too much from the predictor,
// or the motion vector magnitude is considered too big.
// No filtering of this block since it differs too much from the
// predictor, or the motion vector magnitude is considered too big.
vp8_copy_mem16x16(x->thismb, 16,
denoiser->yv12_running_avg.y_buffer + recon_yoffset,
denoiser->yv12_running_avg.y_stride);
return;
}
// Filter.
denoiser_filter(&denoiser->yv12_mc_running_avg,
&denoiser->yv12_running_avg,
x,
motion_magnitude2,
recon_yoffset,
recon_uvoffset);
vp8_denoiser_filter(&denoiser->yv12_mc_running_avg,
&denoiser->yv12_running_avg, x, motion_magnitude2,
recon_yoffset, recon_uvoffset);
}
......@@ -13,6 +13,8 @@
#include "block.h"
#define NOISE_DIFF2_THRESHOLD (75)
typedef struct vp8_denoiser
{
YV12_BUFFER_CONFIG yv12_running_avg;
......@@ -30,4 +32,12 @@ void vp8_denoiser_denoise_mb(VP8_DENOISER *denoiser,
int recon_yoffset,
int recon_uvoffset);
union coeff_pair
{
uint32_t as_int;
uint16_t as_short[2];
};
union coeff_pair *vp8_get_filter_coeff_LUT(unsigned int motion_magnitude);
#endif // VP8_ENCODER_DENOISING_H_
/*
* Copyright (c) 2012 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vp8/encoder/denoising.h"
#include "vp8/common/reconinter.h"
#include "vpx/vpx_integer.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_rtcd.h"
#include <emmintrin.h>
void vp8_denoiser_filter_sse2(YV12_BUFFER_CONFIG *mc_running_avg,
YV12_BUFFER_CONFIG *running_avg,
MACROBLOCK *signal, unsigned int motion_magnitude,
int y_offset, int uv_offset)
{
unsigned char *sig = signal->thismb;
int sig_stride = 16;
unsigned char *mc_running_avg_y = mc_running_avg->y_buffer + y_offset;
int mc_avg_y_stride = mc_running_avg->y_stride;
unsigned char *running_avg_y = running_avg->y_buffer + y_offset;
int avg_y_stride = running_avg->y_stride;
const union coeff_pair *LUT = vp8_get_filter_coeff_LUT(motion_magnitude);
int r, c;
for (r = 0; r < 16; ++r)
{
__m128i filter_coefficient_00, filter_coefficient_04;
__m128i filter_coefficient_08, filter_coefficient_12;
__m128i v_sig0, v_sig1;
__m128i v_mc_running_avg_y0, v_mc_running_avg_y1;
__m128i state0, state1, state2, state3;
__m128i res0, res1, res2, res3;
__m128i v_running_avg_y;
__m128i diff0, diff1, diff0sq, diff1sq, diff_sq;
const __m128i kNOISE_DIFF2_THRESHOLD =
_mm_set1_epi8(NOISE_DIFF2_THRESHOLD);
__m128i take_running, p0, p1, p2;
const __m128i k_zero = _mm_set1_epi16(0);
const __m128i k_128 = _mm_set1_epi32(128);
// Calculate absolute differences
DECLARE_ALIGNED_ARRAY(16,unsigned char,abs_diff,16);
DECLARE_ALIGNED_ARRAY(16,uint32_t,filter_coefficient,16);
__m128i v_sig = _mm_loadu_si128((__m128i *)(&sig[0]));
__m128i v_mc_running_avg_y = _mm_loadu_si128(
(__m128i *)(&mc_running_avg_y[0]));
__m128i a_minus_b = _mm_subs_epu8(v_sig, v_mc_running_avg_y);
__m128i b_minus_a = _mm_subs_epu8(v_mc_running_avg_y, v_sig);
__m128i v_abs_diff = _mm_adds_epu8(a_minus_b, b_minus_a);
_mm_store_si128((__m128i *)(&abs_diff[0]), v_abs_diff);
// Use LUT to get filter coefficients (two 16b value; f and 256-f)
for (c = 0; c < 16; ++c)
{
filter_coefficient[c] = LUT[abs_diff[c]].as_int;
}
// Filtering...
// load filter coefficients (two 16b value; f and 256-f)
filter_coefficient_00 = _mm_load_si128(
(__m128i *)(&filter_coefficient[ 0]));
filter_coefficient_04 = _mm_load_si128(
(__m128i *)(&filter_coefficient[ 4]));
filter_coefficient_08 = _mm_load_si128(
(__m128i *)(&filter_coefficient[ 8]));
filter_coefficient_12 = _mm_load_si128(
(__m128i *)(&filter_coefficient[12]));
// expand sig from 8b to 16b
v_sig0 = _mm_unpacklo_epi8(v_sig, k_zero);
v_sig1 = _mm_unpackhi_epi8(v_sig, k_zero);
// expand mc_running_avg_y from 8b to 16b
v_mc_running_avg_y0 = _mm_unpacklo_epi8(v_mc_running_avg_y, k_zero);
v_mc_running_avg_y1 = _mm_unpackhi_epi8(v_mc_running_avg_y, k_zero);
// interleave sig and mc_running_avg_y for upcoming multiply-add
state0 = _mm_unpacklo_epi16(v_mc_running_avg_y0, v_sig0);
state1 = _mm_unpackhi_epi16(v_mc_running_avg_y0, v_sig0);
state2 = _mm_unpacklo_epi16(v_mc_running_avg_y1, v_sig1);
state3 = _mm_unpackhi_epi16(v_mc_running_avg_y1, v_sig1);
// blend values
res0 = _mm_madd_epi16(filter_coefficient_00, state0);
res1 = _mm_madd_epi16(filter_coefficient_04, state1);
res2 = _mm_madd_epi16(filter_coefficient_08, state2);
res3 = _mm_madd_epi16(filter_coefficient_12, state3);
res0 = _mm_add_epi32(res0, k_128);
res1 = _mm_add_epi32(res1, k_128);
res2 = _mm_add_epi32(res2, k_128);
res3 = _mm_add_epi32(res3, k_128);
res0 = _mm_srai_epi32(res0, 8);
res1 = _mm_srai_epi32(res1, 8);
res2 = _mm_srai_epi32(res2, 8);
res3 = _mm_srai_epi32(res3, 8);
// combine the 32b results into a single 8b vector
res0 = _mm_packs_epi32(res0, res1);
res2 = _mm_packs_epi32(res2, res3);
v_running_avg_y = _mm_packus_epi16(res0, res2);
// Depending on the magnitude of the difference between the signal and
// filtered version, either replace the signal by the filtered one or
// update the filter state with the signal when the change in a pixel
// isn't classified as noise.
diff0 = _mm_sub_epi16(v_sig0, res0);
diff1 = _mm_sub_epi16(v_sig1, res2);
diff0sq = _mm_mullo_epi16(diff0, diff0);
diff1sq = _mm_mullo_epi16(diff1, diff1);
diff_sq = _mm_packus_epi16(diff0sq, diff1sq);
take_running = _mm_cmplt_epi8(diff_sq, kNOISE_DIFF2_THRESHOLD);
p0 = _mm_and_si128(take_running, v_running_avg_y);
p1 = _mm_andnot_si128(take_running, v_sig);
p2 = _mm_or_si128(p0, p1);
_mm_storeu_si128((__m128i *)(&running_avg_y[0]), p2);
_mm_storeu_si128((__m128i *)(&sig[0]), p2);
// Update pointers for next iteration.
sig += sig_stride;
mc_running_avg_y += mc_avg_y_stride;
running_avg_y += avg_y_stride;
}
}
......@@ -99,6 +99,14 @@ VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/vp8_enc_stubs_mmx.c
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/dct_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/fwalsh_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/quantize_sse2.asm
ifeq ($(CONFIG_TEMPORAL_DENOISING),yes)
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/denoising_sse2.c
ifeq ($(HAVE_SSE2),yes)
vp8/encoder/x86/denoising_sse2.c.o: CFLAGS += -msse2
endif
endif
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/subtract_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/temporal_filter_apply_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp8_enc_stubs_sse2.c
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment