2012-06-21 04:34:58 -07:00
|
|
|
/*
|
2012-10-04 09:09:31 -07:00
|
|
|
* Copyright 2011 The LibYuv Project Authors. All rights reserved.
|
2012-06-21 04:34:58 -07:00
|
|
|
*
|
|
|
|
* 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
|
2014-01-30 16:53:28 -08:00
|
|
|
* in the file PATENTS. All contributing project authors may
|
2012-06-21 04:34:58 -07:00
|
|
|
* be found in the AUTHORS file in the root of the source tree.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "libyuv/compare.h"
|
|
|
|
|
|
|
|
#include <float.h>
|
|
|
|
#include <math.h>
|
2012-10-04 09:09:31 -07:00
|
|
|
#ifdef _OPENMP
|
|
|
|
#include <omp.h>
|
|
|
|
#endif
|
2012-06-21 04:34:58 -07:00
|
|
|
|
|
|
|
#include "libyuv/basic_types.h"
|
|
|
|
#include "libyuv/cpu_id.h"
|
2013-02-09 20:16:09 -08:00
|
|
|
#include "libyuv/row.h"
|
2012-06-21 04:34:58 -07:00
|
|
|
|
|
|
|
#ifdef __cplusplus
|
|
|
|
namespace libyuv {
|
|
|
|
extern "C" {
|
|
|
|
#endif
|
|
|
|
|
2012-10-04 09:09:31 -07:00
|
|
|
// hash seed of 5381 recommended.
|
|
|
|
// Internal C version of HashDjb2 with int sized count for efficiency.
|
2014-01-30 16:53:28 -08:00
|
|
|
uint32 HashDjb2_C(const uint8* src, int count, uint32 seed);
|
2012-10-04 09:09:31 -07:00
|
|
|
|
|
|
|
// This module is for Visual C x86
|
2014-01-30 16:53:28 -08:00
|
|
|
#if !defined(LIBYUV_DISABLE_X86) && \
|
|
|
|
(defined(_M_IX86) || \
|
|
|
|
(defined(__x86_64__) || (defined(__i386__) && !defined(__pic__))))
|
2012-10-04 09:09:31 -07:00
|
|
|
#define HAS_HASHDJB2_SSE41
|
2014-01-30 16:53:28 -08:00
|
|
|
uint32 HashDjb2_SSE41(const uint8* src, int count, uint32 seed);
|
2012-10-04 09:09:31 -07:00
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
#if _MSC_VER >= 1700
|
|
|
|
#define HAS_HASHDJB2_AVX2
|
|
|
|
uint32 HashDjb2_AVX2(const uint8* src, int count, uint32 seed);
|
2012-10-04 09:09:31 -07:00
|
|
|
#endif
|
2014-01-30 16:53:28 -08:00
|
|
|
|
2012-10-04 09:09:31 -07:00
|
|
|
#endif // HAS_HASHDJB2_SSE41
|
|
|
|
|
|
|
|
// hash seed of 5381 recommended.
|
2013-02-09 20:16:09 -08:00
|
|
|
LIBYUV_API
|
2012-10-04 09:09:31 -07:00
|
|
|
uint32 HashDjb2(const uint8* src, uint64 count, uint32 seed) {
|
2014-01-30 16:53:28 -08:00
|
|
|
const int kBlockSize = 1 << 15; // 32768;
|
|
|
|
int remainder;
|
2012-10-04 09:09:31 -07:00
|
|
|
uint32 (*HashDjb2_SSE)(const uint8* src, int count, uint32 seed) = HashDjb2_C;
|
|
|
|
#if defined(HAS_HASHDJB2_SSE41)
|
|
|
|
if (TestCpuFlag(kCpuHasSSE41)) {
|
|
|
|
HashDjb2_SSE = HashDjb2_SSE41;
|
|
|
|
}
|
|
|
|
#endif
|
2014-01-30 16:53:28 -08:00
|
|
|
#if defined(HAS_HASHDJB2_AVX2)
|
|
|
|
if (TestCpuFlag(kCpuHasAVX2)) {
|
|
|
|
HashDjb2_SSE = HashDjb2_AVX2;
|
|
|
|
}
|
|
|
|
#endif
|
2012-10-04 09:09:31 -07:00
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
while (count >= (uint64)(kBlockSize)) {
|
2012-10-04 09:09:31 -07:00
|
|
|
seed = HashDjb2_SSE(src, kBlockSize, seed);
|
|
|
|
src += kBlockSize;
|
|
|
|
count -= kBlockSize;
|
|
|
|
}
|
2014-01-30 16:53:28 -08:00
|
|
|
remainder = (int)(count) & ~15;
|
2012-10-04 09:09:31 -07:00
|
|
|
if (remainder) {
|
|
|
|
seed = HashDjb2_SSE(src, remainder, seed);
|
|
|
|
src += remainder;
|
|
|
|
count -= remainder;
|
|
|
|
}
|
2014-01-30 16:53:28 -08:00
|
|
|
remainder = (int)(count) & 15;
|
2012-10-04 09:09:31 -07:00
|
|
|
if (remainder) {
|
|
|
|
seed = HashDjb2_C(src, remainder, seed);
|
|
|
|
}
|
|
|
|
return seed;
|
|
|
|
}
|
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
uint32 SumSquareError_C(const uint8* src_a, const uint8* src_b, int count);
|
|
|
|
#if !defined(LIBYUV_DISABLE_NEON) && \
|
|
|
|
(defined(__ARM_NEON__) || defined(LIBYUV_NEON))
|
2012-06-21 04:34:58 -07:00
|
|
|
#define HAS_SUMSQUAREERROR_NEON
|
2014-01-30 16:53:28 -08:00
|
|
|
uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, int count);
|
|
|
|
#endif
|
|
|
|
#if !defined(LIBYUV_DISABLE_X86) && \
|
|
|
|
(defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
|
2012-06-21 04:34:58 -07:00
|
|
|
#define HAS_SUMSQUAREERROR_SSE2
|
2014-01-30 16:53:28 -08:00
|
|
|
uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, int count);
|
2012-06-21 04:34:58 -07:00
|
|
|
#endif
|
2014-01-30 16:53:28 -08:00
|
|
|
// Visual C 2012 required for AVX2.
|
|
|
|
#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && _MSC_VER >= 1700
|
|
|
|
#define HAS_SUMSQUAREERROR_AVX2
|
|
|
|
uint32 SumSquareError_AVX2(const uint8* src_a, const uint8* src_b, int count);
|
2012-06-21 04:34:58 -07:00
|
|
|
#endif
|
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
// TODO(fbarchard): Refactor into row function.
|
2013-02-09 20:16:09 -08:00
|
|
|
LIBYUV_API
|
2012-10-04 09:09:31 -07:00
|
|
|
uint64 ComputeSumSquareError(const uint8* src_a, const uint8* src_b,
|
|
|
|
int count) {
|
2014-01-30 16:53:28 -08:00
|
|
|
// SumSquareError returns values 0 to 65535 for each squared difference.
|
|
|
|
// Up to 65536 of those can be summed and remain within a uint32.
|
|
|
|
// After each block of 65536 pixels, accumulate into a uint64.
|
|
|
|
const int kBlockSize = 65536;
|
|
|
|
int remainder = count & (kBlockSize - 1) & ~31;
|
|
|
|
uint64 sse = 0;
|
|
|
|
int i;
|
2012-10-04 09:09:31 -07:00
|
|
|
uint32 (*SumSquareError)(const uint8* src_a, const uint8* src_b, int count) =
|
|
|
|
SumSquareError_C;
|
2012-06-21 04:34:58 -07:00
|
|
|
#if defined(HAS_SUMSQUAREERROR_NEON)
|
|
|
|
if (TestCpuFlag(kCpuHasNEON)) {
|
|
|
|
SumSquareError = SumSquareError_NEON;
|
2012-10-04 09:09:31 -07:00
|
|
|
}
|
2014-01-30 16:53:28 -08:00
|
|
|
#endif
|
|
|
|
#if defined(HAS_SUMSQUAREERROR_SSE2)
|
2012-06-21 04:34:58 -07:00
|
|
|
if (TestCpuFlag(kCpuHasSSE2) &&
|
|
|
|
IS_ALIGNED(src_a, 16) && IS_ALIGNED(src_b, 16)) {
|
2012-10-04 09:09:31 -07:00
|
|
|
// Note only used for multiples of 16 so count is not checked.
|
2012-06-21 04:34:58 -07:00
|
|
|
SumSquareError = SumSquareError_SSE2;
|
|
|
|
}
|
2012-10-04 09:09:31 -07:00
|
|
|
#endif
|
2014-01-30 16:53:28 -08:00
|
|
|
#if defined(HAS_SUMSQUAREERROR_AVX2)
|
|
|
|
if (TestCpuFlag(kCpuHasAVX2)) {
|
|
|
|
// Note only used for multiples of 32 so count is not checked.
|
|
|
|
SumSquareError = SumSquareError_AVX2;
|
|
|
|
}
|
|
|
|
#endif
|
2012-10-04 09:09:31 -07:00
|
|
|
#ifdef _OPENMP
|
|
|
|
#pragma omp parallel for reduction(+: sse)
|
|
|
|
#endif
|
2014-01-30 16:53:28 -08:00
|
|
|
for (i = 0; i < (count - (kBlockSize - 1)); i += kBlockSize) {
|
2012-10-04 09:09:31 -07:00
|
|
|
sse += SumSquareError(src_a + i, src_b + i, kBlockSize);
|
2012-06-21 04:34:58 -07:00
|
|
|
}
|
2012-10-04 09:09:31 -07:00
|
|
|
src_a += count & ~(kBlockSize - 1);
|
|
|
|
src_b += count & ~(kBlockSize - 1);
|
2012-06-21 04:34:58 -07:00
|
|
|
if (remainder) {
|
|
|
|
sse += SumSquareError(src_a, src_b, remainder);
|
|
|
|
src_a += remainder;
|
|
|
|
src_b += remainder;
|
|
|
|
}
|
2014-01-30 16:53:28 -08:00
|
|
|
remainder = count & 31;
|
2012-10-04 09:09:31 -07:00
|
|
|
if (remainder) {
|
|
|
|
sse += SumSquareError_C(src_a, src_b, remainder);
|
2012-06-21 04:34:58 -07:00
|
|
|
}
|
|
|
|
return sse;
|
|
|
|
}
|
|
|
|
|
2013-02-09 20:16:09 -08:00
|
|
|
LIBYUV_API
|
2012-06-21 04:34:58 -07:00
|
|
|
uint64 ComputeSumSquareErrorPlane(const uint8* src_a, int stride_a,
|
|
|
|
const uint8* src_b, int stride_b,
|
|
|
|
int width, int height) {
|
2014-01-31 06:02:48 -08:00
|
|
|
uint64 sse = 0;
|
2014-01-30 16:53:28 -08:00
|
|
|
int h;
|
|
|
|
// Coalesce rows.
|
|
|
|
if (stride_a == width &&
|
|
|
|
stride_b == width) {
|
|
|
|
width *= height;
|
|
|
|
height = 1;
|
|
|
|
stride_a = stride_b = 0;
|
|
|
|
}
|
|
|
|
for (h = 0; h < height; ++h) {
|
|
|
|
sse += ComputeSumSquareError(src_a, src_b, width);
|
2012-06-21 04:34:58 -07:00
|
|
|
src_a += stride_a;
|
|
|
|
src_b += stride_b;
|
|
|
|
}
|
|
|
|
return sse;
|
|
|
|
}
|
|
|
|
|
2013-02-09 20:16:09 -08:00
|
|
|
LIBYUV_API
|
2012-06-21 04:34:58 -07:00
|
|
|
double SumSquareErrorToPsnr(uint64 sse, uint64 count) {
|
|
|
|
double psnr;
|
|
|
|
if (sse > 0) {
|
2014-01-30 16:53:28 -08:00
|
|
|
double mse = (double)(count) / (double)(sse);
|
2012-06-21 04:34:58 -07:00
|
|
|
psnr = 10.0 * log10(255.0 * 255.0 * mse);
|
|
|
|
} else {
|
|
|
|
psnr = kMaxPsnr; // Limit to prevent divide by 0
|
|
|
|
}
|
|
|
|
|
|
|
|
if (psnr > kMaxPsnr)
|
|
|
|
psnr = kMaxPsnr;
|
|
|
|
|
|
|
|
return psnr;
|
|
|
|
}
|
|
|
|
|
2013-02-09 20:16:09 -08:00
|
|
|
LIBYUV_API
|
2012-06-21 04:34:58 -07:00
|
|
|
double CalcFramePsnr(const uint8* src_a, int stride_a,
|
|
|
|
const uint8* src_b, int stride_b,
|
|
|
|
int width, int height) {
|
|
|
|
const uint64 samples = width * height;
|
|
|
|
const uint64 sse = ComputeSumSquareErrorPlane(src_a, stride_a,
|
|
|
|
src_b, stride_b,
|
|
|
|
width, height);
|
|
|
|
return SumSquareErrorToPsnr(sse, samples);
|
|
|
|
}
|
|
|
|
|
2013-02-09 20:16:09 -08:00
|
|
|
LIBYUV_API
|
2012-06-21 04:34:58 -07:00
|
|
|
double I420Psnr(const uint8* src_y_a, int stride_y_a,
|
|
|
|
const uint8* src_u_a, int stride_u_a,
|
|
|
|
const uint8* src_v_a, int stride_v_a,
|
|
|
|
const uint8* src_y_b, int stride_y_b,
|
|
|
|
const uint8* src_u_b, int stride_u_b,
|
|
|
|
const uint8* src_v_b, int stride_v_b,
|
|
|
|
int width, int height) {
|
|
|
|
const uint64 sse_y = ComputeSumSquareErrorPlane(src_y_a, stride_y_a,
|
|
|
|
src_y_b, stride_y_b,
|
|
|
|
width, height);
|
|
|
|
const int width_uv = (width + 1) >> 1;
|
|
|
|
const int height_uv = (height + 1) >> 1;
|
|
|
|
const uint64 sse_u = ComputeSumSquareErrorPlane(src_u_a, stride_u_a,
|
|
|
|
src_u_b, stride_u_b,
|
|
|
|
width_uv, height_uv);
|
|
|
|
const uint64 sse_v = ComputeSumSquareErrorPlane(src_v_a, stride_v_a,
|
|
|
|
src_v_b, stride_v_b,
|
|
|
|
width_uv, height_uv);
|
|
|
|
const uint64 samples = width * height + 2 * (width_uv * height_uv);
|
|
|
|
const uint64 sse = sse_y + sse_u + sse_v;
|
|
|
|
return SumSquareErrorToPsnr(sse, samples);
|
|
|
|
}
|
|
|
|
|
|
|
|
static const int64 cc1 = 26634; // (64^2*(.01*255)^2
|
|
|
|
static const int64 cc2 = 239708; // (64^2*(.03*255)^2
|
|
|
|
|
|
|
|
static double Ssim8x8_C(const uint8* src_a, int stride_a,
|
|
|
|
const uint8* src_b, int stride_b) {
|
|
|
|
int64 sum_a = 0;
|
|
|
|
int64 sum_b = 0;
|
|
|
|
int64 sum_sq_a = 0;
|
|
|
|
int64 sum_sq_b = 0;
|
|
|
|
int64 sum_axb = 0;
|
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
int i;
|
|
|
|
for (i = 0; i < 8; ++i) {
|
|
|
|
int j;
|
|
|
|
for (j = 0; j < 8; ++j) {
|
2012-06-21 04:34:58 -07:00
|
|
|
sum_a += src_a[j];
|
|
|
|
sum_b += src_b[j];
|
|
|
|
sum_sq_a += src_a[j] * src_a[j];
|
|
|
|
sum_sq_b += src_b[j] * src_b[j];
|
|
|
|
sum_axb += src_a[j] * src_b[j];
|
|
|
|
}
|
|
|
|
|
|
|
|
src_a += stride_a;
|
|
|
|
src_b += stride_b;
|
|
|
|
}
|
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
{
|
|
|
|
const int64 count = 64;
|
|
|
|
// scale the constants by number of pixels
|
|
|
|
const int64 c1 = (cc1 * count * count) >> 12;
|
|
|
|
const int64 c2 = (cc2 * count * count) >> 12;
|
2012-06-21 04:34:58 -07:00
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
const int64 sum_a_x_sum_b = sum_a * sum_b;
|
2012-06-21 04:34:58 -07:00
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
const int64 ssim_n = (2 * sum_a_x_sum_b + c1) *
|
|
|
|
(2 * count * sum_axb - 2 * sum_a_x_sum_b + c2);
|
2012-06-21 04:34:58 -07:00
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
const int64 sum_a_sq = sum_a*sum_a;
|
|
|
|
const int64 sum_b_sq = sum_b*sum_b;
|
2012-06-21 04:34:58 -07:00
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
const int64 ssim_d = (sum_a_sq + sum_b_sq + c1) *
|
|
|
|
(count * sum_sq_a - sum_a_sq +
|
|
|
|
count * sum_sq_b - sum_b_sq + c2);
|
2012-06-21 04:34:58 -07:00
|
|
|
|
2014-01-30 16:53:28 -08:00
|
|
|
if (ssim_d == 0.0) {
|
|
|
|
return DBL_MAX;
|
|
|
|
}
|
|
|
|
return ssim_n * 1.0 / ssim_d;
|
|
|
|
}
|
2012-06-21 04:34:58 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// We are using a 8x8 moving window with starting location of each 8x8 window
|
|
|
|
// on the 4x4 pixel grid. Such arrangement allows the windows to overlap
|
|
|
|
// block boundaries to penalize blocking artifacts.
|
2013-02-09 20:16:09 -08:00
|
|
|
LIBYUV_API
|
2012-06-21 04:34:58 -07:00
|
|
|
double CalcFrameSsim(const uint8* src_a, int stride_a,
|
|
|
|
const uint8* src_b, int stride_b,
|
|
|
|
int width, int height) {
|
|
|
|
int samples = 0;
|
|
|
|
double ssim_total = 0;
|
|
|
|
double (*Ssim8x8)(const uint8* src_a, int stride_a,
|
2014-01-30 16:53:28 -08:00
|
|
|
const uint8* src_b, int stride_b) = Ssim8x8_C;
|
2012-06-21 04:34:58 -07:00
|
|
|
|
|
|
|
// sample point start with each 4x4 location
|
2014-01-30 16:53:28 -08:00
|
|
|
int i;
|
|
|
|
for (i = 0; i < height - 8; i += 4) {
|
|
|
|
int j;
|
|
|
|
for (j = 0; j < width - 8; j += 4) {
|
2012-06-21 04:34:58 -07:00
|
|
|
ssim_total += Ssim8x8(src_a + j, stride_a, src_b + j, stride_b);
|
|
|
|
samples++;
|
|
|
|
}
|
|
|
|
|
|
|
|
src_a += stride_a * 4;
|
|
|
|
src_b += stride_b * 4;
|
|
|
|
}
|
|
|
|
|
|
|
|
ssim_total /= samples;
|
|
|
|
return ssim_total;
|
|
|
|
}
|
|
|
|
|
2013-02-09 20:16:09 -08:00
|
|
|
LIBYUV_API
|
2012-06-21 04:34:58 -07:00
|
|
|
double I420Ssim(const uint8* src_y_a, int stride_y_a,
|
|
|
|
const uint8* src_u_a, int stride_u_a,
|
|
|
|
const uint8* src_v_a, int stride_v_a,
|
|
|
|
const uint8* src_y_b, int stride_y_b,
|
|
|
|
const uint8* src_u_b, int stride_u_b,
|
|
|
|
const uint8* src_v_b, int stride_v_b,
|
|
|
|
int width, int height) {
|
|
|
|
const double ssim_y = CalcFrameSsim(src_y_a, stride_y_a,
|
|
|
|
src_y_b, stride_y_b, width, height);
|
|
|
|
const int width_uv = (width + 1) >> 1;
|
|
|
|
const int height_uv = (height + 1) >> 1;
|
|
|
|
const double ssim_u = CalcFrameSsim(src_u_a, stride_u_a,
|
|
|
|
src_u_b, stride_u_b,
|
|
|
|
width_uv, height_uv);
|
|
|
|
const double ssim_v = CalcFrameSsim(src_v_a, stride_v_a,
|
|
|
|
src_v_b, stride_v_b,
|
|
|
|
width_uv, height_uv);
|
|
|
|
return ssim_y * 0.8 + 0.1 * (ssim_u + ssim_v);
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef __cplusplus
|
|
|
|
} // extern "C"
|
|
|
|
} // namespace libyuv
|
|
|
|
#endif
|