diff options
author | Matt A. Tobin <email@mattatobin.com> | 2020-04-07 23:30:51 -0400 |
---|---|---|
committer | wolfbeast <mcwerewolf@wolfbeast.com> | 2020-04-14 13:26:42 +0200 |
commit | 277f2116b6660e9bbe7f5d67524be57eceb49b8b (patch) | |
tree | 4595f7cc71418f71b9a97dfaeb03a30aa60f336a /media/libaom/src/aom_dsp/psnr.c | |
parent | d270404436f6e84ffa3b92af537ac721bf10d66e (diff) | |
download | UXP-277f2116b6660e9bbe7f5d67524be57eceb49b8b.tar UXP-277f2116b6660e9bbe7f5d67524be57eceb49b8b.tar.gz UXP-277f2116b6660e9bbe7f5d67524be57eceb49b8b.tar.lz UXP-277f2116b6660e9bbe7f5d67524be57eceb49b8b.tar.xz UXP-277f2116b6660e9bbe7f5d67524be57eceb49b8b.zip |
Move aom source to a sub-directory under media/libaom
There is no damned reason to treat this differently than any other media lib given its license and there never was.
Diffstat (limited to 'media/libaom/src/aom_dsp/psnr.c')
-rw-r--r-- | media/libaom/src/aom_dsp/psnr.c | 381 |
1 files changed, 381 insertions, 0 deletions
diff --git a/media/libaom/src/aom_dsp/psnr.c b/media/libaom/src/aom_dsp/psnr.c new file mode 100644 index 000000000..50f376a4a --- /dev/null +++ b/media/libaom/src/aom_dsp/psnr.c @@ -0,0 +1,381 @@ +/* + * 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. + */ + +#include <assert.h> +#include <math.h> + +#include "config/aom_dsp_rtcd.h" + +#include "aom_dsp/psnr.h" +#include "aom_scale/yv12config.h" + +double aom_sse_to_psnr(double samples, double peak, double sse) { + if (sse > 0.0) { + const double psnr = 10.0 * log10(samples * peak * peak / sse); + return psnr > MAX_PSNR ? MAX_PSNR : psnr; + } else { + return MAX_PSNR; + } +} + +/* TODO(yaowu): The block_variance calls the unoptimized versions of variance() + * and highbd_8_variance(). It should not. + */ +static void encoder_variance(const uint8_t *a, int a_stride, const uint8_t *b, + int b_stride, int w, int h, unsigned int *sse, + int *sum) { + int i, j; + + *sum = 0; + *sse = 0; + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const int diff = a[j] - b[j]; + *sum += diff; + *sse += diff * diff; + } + + a += a_stride; + b += b_stride; + } +} + +static void encoder_highbd_variance64(const uint8_t *a8, int a_stride, + const uint8_t *b8, int b_stride, int w, + int h, uint64_t *sse, int64_t *sum) { + const uint16_t *a = CONVERT_TO_SHORTPTR(a8); + const uint16_t *b = CONVERT_TO_SHORTPTR(b8); + int64_t tsum = 0; + uint64_t tsse = 0; + for (int i = 0; i < h; ++i) { + int32_t lsum = 0; + for (int j = 0; j < w; ++j) { + const int diff = a[j] - b[j]; + lsum += diff; + tsse += (uint32_t)(diff * diff); + } + tsum += lsum; + a += a_stride; + b += b_stride; + } + *sum = tsum; + *sse = tsse; +} + +static void encoder_highbd_8_variance(const uint8_t *a8, int a_stride, + const uint8_t *b8, int b_stride, int w, + int h, unsigned int *sse, int *sum) { + uint64_t sse_long = 0; + int64_t sum_long = 0; + encoder_highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, + &sum_long); + *sse = (unsigned int)sse_long; + *sum = (int)sum_long; +} + +static int64_t get_sse(const uint8_t *a, int a_stride, const uint8_t *b, + int b_stride, int width, int height) { + const int dw = width % 16; + const int dh = height % 16; + int64_t total_sse = 0; + unsigned int sse = 0; + int sum = 0; + int x, y; + + if (dw > 0) { + encoder_variance(&a[width - dw], a_stride, &b[width - dw], b_stride, dw, + height, &sse, &sum); + total_sse += sse; + } + + if (dh > 0) { + encoder_variance(&a[(height - dh) * a_stride], a_stride, + &b[(height - dh) * b_stride], b_stride, width - dw, dh, + &sse, &sum); + total_sse += sse; + } + + for (y = 0; y < height / 16; ++y) { + const uint8_t *pa = a; + const uint8_t *pb = b; + for (x = 0; x < width / 16; ++x) { + aom_mse16x16(pa, a_stride, pb, b_stride, &sse); + total_sse += sse; + + pa += 16; + pb += 16; + } + + a += 16 * a_stride; + b += 16 * b_stride; + } + + return total_sse; +} + +static int64_t highbd_get_sse_shift(const uint8_t *a8, int a_stride, + const uint8_t *b8, int b_stride, int width, + int height, unsigned int input_shift) { + const uint16_t *a = CONVERT_TO_SHORTPTR(a8); + const uint16_t *b = CONVERT_TO_SHORTPTR(b8); + int64_t total_sse = 0; + int x, y; + for (y = 0; y < height; ++y) { + for (x = 0; x < width; ++x) { + int64_t diff; + diff = (a[x] >> input_shift) - (b[x] >> input_shift); + total_sse += diff * diff; + } + a += a_stride; + b += b_stride; + } + return total_sse; +} + +static int64_t highbd_get_sse(const uint8_t *a, int a_stride, const uint8_t *b, + int b_stride, int width, int height) { + int64_t total_sse = 0; + int x, y; + const int dw = width % 16; + const int dh = height % 16; + unsigned int sse = 0; + int sum = 0; + if (dw > 0) { + encoder_highbd_8_variance(&a[width - dw], a_stride, &b[width - dw], + b_stride, dw, height, &sse, &sum); + total_sse += sse; + } + if (dh > 0) { + encoder_highbd_8_variance(&a[(height - dh) * a_stride], a_stride, + &b[(height - dh) * b_stride], b_stride, + width - dw, dh, &sse, &sum); + total_sse += sse; + } + for (y = 0; y < height / 16; ++y) { + const uint8_t *pa = a; + const uint8_t *pb = b; + for (x = 0; x < width / 16; ++x) { + aom_highbd_8_mse16x16(pa, a_stride, pb, b_stride, &sse); + total_sse += sse; + pa += 16; + pb += 16; + } + a += 16 * a_stride; + b += 16 * b_stride; + } + return total_sse; +} + +int64_t aom_get_y_sse_part(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b, int hstart, int width, + int vstart, int height) { + return get_sse(a->y_buffer + vstart * a->y_stride + hstart, a->y_stride, + b->y_buffer + vstart * b->y_stride + hstart, b->y_stride, + width, height); +} + +int64_t aom_get_y_sse(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b) { + assert(a->y_crop_width == b->y_crop_width); + assert(a->y_crop_height == b->y_crop_height); + + return get_sse(a->y_buffer, a->y_stride, b->y_buffer, b->y_stride, + a->y_crop_width, a->y_crop_height); +} + +int64_t aom_get_u_sse_part(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b, int hstart, int width, + int vstart, int height) { + return get_sse(a->u_buffer + vstart * a->uv_stride + hstart, a->uv_stride, + b->u_buffer + vstart * b->uv_stride + hstart, b->uv_stride, + width, height); +} + +int64_t aom_get_u_sse(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b) { + assert(a->uv_crop_width == b->uv_crop_width); + assert(a->uv_crop_height == b->uv_crop_height); + + return get_sse(a->u_buffer, a->uv_stride, b->u_buffer, b->uv_stride, + a->uv_crop_width, a->uv_crop_height); +} + +int64_t aom_get_v_sse_part(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b, int hstart, int width, + int vstart, int height) { + return get_sse(a->v_buffer + vstart * a->uv_stride + hstart, a->uv_stride, + b->v_buffer + vstart * b->uv_stride + hstart, b->uv_stride, + width, height); +} + +int64_t aom_get_v_sse(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b) { + assert(a->uv_crop_width == b->uv_crop_width); + assert(a->uv_crop_height == b->uv_crop_height); + + return get_sse(a->v_buffer, a->uv_stride, b->v_buffer, b->uv_stride, + a->uv_crop_width, a->uv_crop_height); +} + +int64_t aom_highbd_get_y_sse_part(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b, int hstart, + int width, int vstart, int height) { + return highbd_get_sse( + a->y_buffer + vstart * a->y_stride + hstart, a->y_stride, + b->y_buffer + vstart * b->y_stride + hstart, b->y_stride, width, height); +} + +int64_t aom_highbd_get_y_sse(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b) { + assert(a->y_crop_width == b->y_crop_width); + assert(a->y_crop_height == b->y_crop_height); + assert((a->flags & YV12_FLAG_HIGHBITDEPTH) != 0); + assert((b->flags & YV12_FLAG_HIGHBITDEPTH) != 0); + + return highbd_get_sse(a->y_buffer, a->y_stride, b->y_buffer, b->y_stride, + a->y_crop_width, a->y_crop_height); +} + +int64_t aom_highbd_get_u_sse_part(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b, int hstart, + int width, int vstart, int height) { + return highbd_get_sse(a->u_buffer + vstart * a->uv_stride + hstart, + a->uv_stride, + b->u_buffer + vstart * b->uv_stride + hstart, + b->uv_stride, width, height); +} + +int64_t aom_highbd_get_u_sse(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b) { + assert(a->uv_crop_width == b->uv_crop_width); + assert(a->uv_crop_height == b->uv_crop_height); + assert((a->flags & YV12_FLAG_HIGHBITDEPTH) != 0); + assert((b->flags & YV12_FLAG_HIGHBITDEPTH) != 0); + + return highbd_get_sse(a->u_buffer, a->uv_stride, b->u_buffer, b->uv_stride, + a->uv_crop_width, a->uv_crop_height); +} + +int64_t aom_highbd_get_v_sse_part(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b, int hstart, + int width, int vstart, int height) { + return highbd_get_sse(a->v_buffer + vstart * a->uv_stride + hstart, + a->uv_stride, + b->v_buffer + vstart * b->uv_stride + hstart, + b->uv_stride, width, height); +} + +int64_t aom_highbd_get_v_sse(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b) { + assert(a->uv_crop_width == b->uv_crop_width); + assert(a->uv_crop_height == b->uv_crop_height); + assert((a->flags & YV12_FLAG_HIGHBITDEPTH) != 0); + assert((b->flags & YV12_FLAG_HIGHBITDEPTH) != 0); + + return highbd_get_sse(a->v_buffer, a->uv_stride, b->v_buffer, b->uv_stride, + a->uv_crop_width, a->uv_crop_height); +} + +int64_t aom_get_sse_plane(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b, int plane, int highbd) { + if (highbd) { + switch (plane) { + case 0: return aom_highbd_get_y_sse(a, b); + case 1: return aom_highbd_get_u_sse(a, b); + case 2: return aom_highbd_get_v_sse(a, b); + default: assert(plane >= 0 && plane <= 2); return 0; + } + } + switch (plane) { + case 0: return aom_get_y_sse(a, b); + case 1: return aom_get_u_sse(a, b); + case 2: return aom_get_v_sse(a, b); + default: assert(plane >= 0 && plane <= 2); return 0; + } +} + +void aom_calc_highbd_psnr(const YV12_BUFFER_CONFIG *a, + const YV12_BUFFER_CONFIG *b, PSNR_STATS *psnr, + uint32_t bit_depth, uint32_t in_bit_depth) { + const int widths[3] = { a->y_crop_width, a->uv_crop_width, a->uv_crop_width }; + const int heights[3] = { a->y_crop_height, a->uv_crop_height, + a->uv_crop_height }; + const int a_strides[3] = { a->y_stride, a->uv_stride, a->uv_stride }; + const int b_strides[3] = { b->y_stride, b->uv_stride, b->uv_stride }; + int i; + uint64_t total_sse = 0; + uint32_t total_samples = 0; + const double peak = (double)((1 << in_bit_depth) - 1); + const unsigned int input_shift = bit_depth - in_bit_depth; + + for (i = 0; i < 3; ++i) { + const int w = widths[i]; + const int h = heights[i]; + const uint32_t samples = w * h; + uint64_t sse; + if (a->flags & YV12_FLAG_HIGHBITDEPTH) { + if (input_shift) { + sse = highbd_get_sse_shift(a->buffers[i], a_strides[i], b->buffers[i], + b_strides[i], w, h, input_shift); + } else { + sse = highbd_get_sse(a->buffers[i], a_strides[i], b->buffers[i], + b_strides[i], w, h); + } + } else { + sse = get_sse(a->buffers[i], a_strides[i], b->buffers[i], b_strides[i], w, + h); + } + psnr->sse[1 + i] = sse; + psnr->samples[1 + i] = samples; + psnr->psnr[1 + i] = aom_sse_to_psnr(samples, peak, (double)sse); + + total_sse += sse; + total_samples += samples; + } + + psnr->sse[0] = total_sse; + psnr->samples[0] = total_samples; + psnr->psnr[0] = + aom_sse_to_psnr((double)total_samples, peak, (double)total_sse); +} + +void aom_calc_psnr(const YV12_BUFFER_CONFIG *a, const YV12_BUFFER_CONFIG *b, + PSNR_STATS *psnr) { + static const double peak = 255.0; + const int widths[3] = { a->y_crop_width, a->uv_crop_width, a->uv_crop_width }; + const int heights[3] = { a->y_crop_height, a->uv_crop_height, + a->uv_crop_height }; + const int a_strides[3] = { a->y_stride, a->uv_stride, a->uv_stride }; + const int b_strides[3] = { b->y_stride, b->uv_stride, b->uv_stride }; + int i; + uint64_t total_sse = 0; + uint32_t total_samples = 0; + + for (i = 0; i < 3; ++i) { + const int w = widths[i]; + const int h = heights[i]; + const uint32_t samples = w * h; + const uint64_t sse = + get_sse(a->buffers[i], a_strides[i], b->buffers[i], b_strides[i], w, h); + psnr->sse[1 + i] = sse; + psnr->samples[1 + i] = samples; + psnr->psnr[1 + i] = aom_sse_to_psnr(samples, peak, (double)sse); + + total_sse += sse; + total_samples += samples; + } + + psnr->sse[0] = total_sse; + psnr->samples[0] = total_samples; + psnr->psnr[0] = + aom_sse_to_psnr((double)total_samples, peak, (double)total_sse); +} |