From b52db6b7e8ca2b49d09500cdf599a2f15b8baedd Mon Sep 17 00:00:00 2001 From: Christian Duvivier Date: Wed, 1 Jan 2014 18:46:47 -0800 Subject: [PATCH] ARM NEON version of denoiser. Change-Id: I951abd4ad0078f78949f3cb79453ac334fb82a7e --- vp8/common/rtcd_defs.sh | 2 +- vp8/encoder/arm/neon/denoising_neon.c | 165 ++++++++++++++++++++++++++ vp8/vp8cx_arm.mk | 1 + 3 files changed, 167 insertions(+), 1 deletion(-) create mode 100644 vp8/encoder/arm/neon/denoising_neon.c diff --git a/vp8/common/rtcd_defs.sh b/vp8/common/rtcd_defs.sh index 9ebf389d8..28e675477 100644 --- a/vp8/common/rtcd_defs.sh +++ b/vp8/common/rtcd_defs.sh @@ -535,7 +535,7 @@ specialize vp8_yv12_copy_partial_frame neon # if [ "$CONFIG_TEMPORAL_DENOISING" = "yes" ]; then prototype int 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 + specialize vp8_denoiser_filter sse2 neon fi # End of encoder only functions diff --git a/vp8/encoder/arm/neon/denoising_neon.c b/vp8/encoder/arm/neon/denoising_neon.c new file mode 100644 index 000000000..d517dfa37 --- /dev/null +++ b/vp8/encoder/arm/neon/denoising_neon.c @@ -0,0 +1,165 @@ +/* + * 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 + +#include "vp8/encoder/denoising.h" +#include "vpx_mem/vpx_mem.h" +#include "./vp8_rtcd.h" + +/* + * The filter function was modified to reduce the computational complexity. + * + * Step 1: + * Instead of applying tap coefficients for each pixel, we calculated the + * pixel adjustments vs. pixel diff value ahead of time. + * adjustment = filtered_value - current_raw + * = (filter_coefficient * diff + 128) >> 8 + * where + * filter_coefficient = (255 << 8) / (256 + ((abs_diff * 330) >> 3)); + * filter_coefficient += filter_coefficient / + * (3 + motion_magnitude_adjustment); + * filter_coefficient is clamped to 0 ~ 255. + * + * Step 2: + * The adjustment vs. diff curve becomes flat very quick when diff increases. + * This allowed us to use only several levels to approximate the curve without + * changing the filtering algorithm too much. + * The adjustments were further corrected by checking the motion magnitude. + * The levels used are: + * diff level adjustment w/o adjustment w/ + * motion correction motion correction + * [-255, -16] 3 -6 -7 + * [-15, -8] 2 -4 -5 + * [-7, -4] 1 -3 -4 + * [-3, 3] 0 diff diff + * [4, 7] 1 3 4 + * [8, 15] 2 4 5 + * [16, 255] 3 6 7 + */ + +int vp8_denoiser_filter_neon(YV12_BUFFER_CONFIG *mc_running_avg, + YV12_BUFFER_CONFIG *running_avg, + MACROBLOCK *signal, unsigned int motion_magnitude, + int y_offset, int uv_offset) { + /* If motion_magnitude is small, making the denoiser more aggressive by + * increasing the adjustment for each level, level1 adjustment is + * increased, the deltas stay the same. + */ + const uint8x16_t v_level1_adjustment = vdupq_n_u8( + (motion_magnitude <= MOTION_MAGNITUDE_THRESHOLD) ? 4 : 3); + const uint8x16_t v_delta_level_1_and_2 = vdupq_n_u8(1); + const uint8x16_t v_delta_level_2_and_3 = vdupq_n_u8(2); + const uint8x16_t v_level1_threshold = vdupq_n_u8(4); + const uint8x16_t v_level2_threshold = vdupq_n_u8(8); + const uint8x16_t v_level3_threshold = vdupq_n_u8(16); + + /* Local variables for array pointers and strides. */ + unsigned char *sig = signal->thismb; + int sig_stride = 16; + unsigned char *mc_running_avg_y = mc_running_avg->y_buffer + y_offset; + int mc_running_avg_y_stride = mc_running_avg->y_stride; + unsigned char *running_avg_y = running_avg->y_buffer + y_offset; + int running_avg_y_stride = running_avg->y_stride; + + /* Go over lines. */ + int i; + int sum_diff = 0; + for (i = 0; i < 16; ++i) { + int8x16_t v_sum_diff = vdupq_n_s8(0); + uint8x16_t v_running_avg_y; + + /* Load inputs. */ + const uint8x16_t v_sig = vld1q_u8(sig); + const uint8x16_t v_mc_running_avg_y = vld1q_u8(mc_running_avg_y); + + /* Calculate absolute difference and sign masks. */ + const uint8x16_t v_abs_diff = vabdq_u8(v_sig, v_mc_running_avg_y); + const uint8x16_t v_diff_pos_mask = vcltq_u8(v_sig, v_mc_running_avg_y); + const uint8x16_t v_diff_neg_mask = vcgtq_u8(v_sig, v_mc_running_avg_y); + + /* Figure out which level that put us in. */ + const uint8x16_t v_level1_mask = vcleq_u8(v_level1_threshold, + v_abs_diff); + const uint8x16_t v_level2_mask = vcleq_u8(v_level2_threshold, + v_abs_diff); + const uint8x16_t v_level3_mask = vcleq_u8(v_level3_threshold, + v_abs_diff); + + /* Calculate absolute adjustments for level 1, 2 and 3. */ + const uint8x16_t v_level2_adjustment = vandq_u8(v_level2_mask, + v_delta_level_1_and_2); + const uint8x16_t v_level3_adjustment = vandq_u8(v_level3_mask, + v_delta_level_2_and_3); + const uint8x16_t v_level1and2_adjustment = vaddq_u8(v_level1_adjustment, + v_level2_adjustment); + const uint8x16_t v_level1and2and3_adjustment = vaddq_u8( + v_level1and2_adjustment, v_level3_adjustment); + + /* Figure adjustment absolute value by selecting between the absolute + * difference if in level0 or the value for level 1, 2 and 3. + */ + const uint8x16_t v_abs_adjustment = vbslq_u8(v_level1_mask, + v_level1and2and3_adjustment, v_abs_diff); + + /* Calculate positive and negative adjustments. Apply them to the signal + * and accumulate them. Adjustments are less than eight and the maximum + * sum of them (7 * 16) can fit in a signed char. + */ + const uint8x16_t v_pos_adjustment = vandq_u8(v_diff_pos_mask, + v_abs_adjustment); + const uint8x16_t v_neg_adjustment = vandq_u8(v_diff_neg_mask, + v_abs_adjustment); + v_running_avg_y = vqaddq_u8(v_sig, v_pos_adjustment); + v_running_avg_y = vqsubq_u8(v_running_avg_y, v_neg_adjustment); + v_sum_diff = vqaddq_s8(v_sum_diff, (int8x16_t)v_pos_adjustment); + v_sum_diff = vqsubq_s8(v_sum_diff, (int8x16_t)v_neg_adjustment); + + /* Store results. */ + vst1q_u8(running_avg_y, v_running_avg_y); + + /* Sum all the accumulators to have the sum of all pixel differences + * for this macroblock. + */ + { + int s0 = vgetq_lane_s8(v_sum_diff, 0) + + vgetq_lane_s8(v_sum_diff, 1) + + vgetq_lane_s8(v_sum_diff, 2) + + vgetq_lane_s8(v_sum_diff, 3); + int s1 = vgetq_lane_s8(v_sum_diff, 4) + + vgetq_lane_s8(v_sum_diff, 5) + + vgetq_lane_s8(v_sum_diff, 6) + + vgetq_lane_s8(v_sum_diff, 7); + int s2 = vgetq_lane_s8(v_sum_diff, 8) + + vgetq_lane_s8(v_sum_diff, 9) + + vgetq_lane_s8(v_sum_diff, 10) + + vgetq_lane_s8(v_sum_diff, 11); + int s3 = vgetq_lane_s8(v_sum_diff, 12) + + vgetq_lane_s8(v_sum_diff, 13) + + vgetq_lane_s8(v_sum_diff, 14) + + vgetq_lane_s8(v_sum_diff, 15); + sum_diff += s0 + s1+ s2 + s3; + } + + /* Update pointers for next iteration. */ + sig += sig_stride; + mc_running_avg_y += mc_running_avg_y_stride; + running_avg_y += running_avg_y_stride; + } + + /* Too much adjustments => copy block. */ + if (abs(sum_diff) > SUM_DIFF_THRESHOLD) + return COPY_BLOCK; + + /* Tell above level that block was filtered. */ + vp8_copy_mem16x16(running_avg->y_buffer + y_offset, running_avg_y_stride, + signal->thismb, sig_stride); + return FILTER_BLOCK; +} diff --git a/vp8/vp8cx_arm.mk b/vp8/vp8cx_arm.mk index b030ee57e..398172a2b 100644 --- a/vp8/vp8cx_arm.mk +++ b/vp8/vp8cx_arm.mk @@ -37,6 +37,7 @@ VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/walsh_v6$(ASM) # encoder VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/fastquantizeb_neon$(ASM) VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/picklpf_arm.c +VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/denoising_neon.c VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/shortfdct_neon$(ASM) VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/subtract_neon$(ASM) VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/vp8_mse16x16_neon$(ASM)