webrtc/test/testsupport/metrics/video_metrics.cc

353 lines
12 KiB
C++
Raw Normal View History

/*
* Copyright (c) 2011 The WebRTC 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 "video_metrics.h"
#include <algorithm> // min_element, max_element
#include <cmath>
#include <fstream>
#include "system_wrappers/interface/cpu_features_wrapper.h"
// Calculates PSNR from MSE
static inline double CalcPsnr(double mse) {
// Formula: PSNR = 10 log (255^2 / MSE) = 20 log(255) - 10 log(MSE)
return 20.0 * std::log10(255.0) - 10.0 * std::log10(mse);
}
// Used for calculating min and max values
static bool lessForFrameResultValue (const FrameResult& s1, const FrameResult& s2) {
return s1.value < s2.value;
}
WebRtc_Word32
PsnrFromFiles(const WebRtc_Word8 *refFileName, const WebRtc_Word8 *testFileName,
WebRtc_Word32 width, WebRtc_Word32 height, QualityMetricsResult *result)
{
FILE *refFp = fopen(refFileName, "rb");
if (refFp == NULL )
{
// cannot open reference file
fprintf(stderr, "Cannot open file %s\n", refFileName);
return -1;
}
FILE *testFp = fopen(testFileName, "rb");
if (testFp == NULL )
{
// cannot open test file
fprintf(stderr, "Cannot open file %s\n", testFileName);
return -2;
}
double mse = 0.0;
double mseSum = 0.0;
WebRtc_Word32 frames = 0;
// Allocating size for one I420 frame.
WebRtc_Word32 frameBytes = 3 * width * height >> 1;
WebRtc_UWord8 *ref = new WebRtc_UWord8[frameBytes];
WebRtc_UWord8 *test = new WebRtc_UWord8[frameBytes];
WebRtc_Word32 refBytes = (WebRtc_Word32) fread(ref, 1, frameBytes, refFp);
WebRtc_Word32 testBytes = (WebRtc_Word32) fread(test, 1, frameBytes, testFp);
while (refBytes == frameBytes && testBytes == frameBytes)
{
mse = 0.0;
WebRtc_Word32 sh = 8; //boundary offset
for (WebRtc_Word32 k2 = sh; k2 < height - sh; k2++)
for (WebRtc_Word32 k = sh; k < width - sh; k++)
{
WebRtc_Word32 kk = k2 * width + k;
mse += (test[kk] - ref[kk]) * (test[kk] - ref[kk]);
}
// divide by number of pixels
mse /= (double) (width * height);
// Save statistics for this specific frame
FrameResult frame_result;
frame_result.value = CalcPsnr(mse);
frame_result.frame_number = frames;
result->frames.push_back(frame_result);
// accumulate for total average
mseSum += mse;
frames++;
refBytes = (WebRtc_Word32) fread(ref, 1, frameBytes, refFp);
testBytes = (WebRtc_Word32) fread(test, 1, frameBytes, testFp);
}
if (mse == 0)
{
// The PSNR value is undefined in this case.
// This value effectively means that the files are equal.
result->average = std::numeric_limits<double>::max();
}
else
{
result->average = CalcPsnr(mseSum / frames);
}
// Calculate min/max statistics
std::vector<FrameResult>::iterator element;
element = min_element(result->frames.begin(),
result->frames.end(), lessForFrameResultValue);
result->min = element->value;
result->min_frame_number = element->frame_number;
element = max_element(result->frames.begin(),
result->frames.end(), lessForFrameResultValue);
result->max = element->value;
result->max_frame_number = element->frame_number;
delete [] ref;
delete [] test;
fclose(refFp);
fclose(testFp);
return 0;
}
static double
Similarity(WebRtc_UWord64 sum_s, WebRtc_UWord64 sum_r, WebRtc_UWord64 sum_sq_s,
WebRtc_UWord64 sum_sq_r, WebRtc_UWord64 sum_sxr, WebRtc_Word32 count)
{
WebRtc_Word64 ssim_n, ssim_d;
WebRtc_Word64 c1, c2;
const WebRtc_Word64 cc1 = 26634; // (64^2*(.01*255)^2
const WebRtc_Word64 cc2 = 239708; // (64^2*(.03*255)^2
// Scale the constants by number of pixels
c1 = (cc1 * count * count) >> 12;
c2 = (cc2 * count * count) >> 12;
ssim_n = (2 * sum_s * sum_r + c1) * ((WebRtc_Word64) 2 * count * sum_sxr-
(WebRtc_Word64) 2 * sum_s * sum_r + c2);
ssim_d = (sum_s * sum_s + sum_r * sum_r + c1)*
((WebRtc_Word64)count * sum_sq_s - (WebRtc_Word64)sum_s * sum_s +
(WebRtc_Word64)count * sum_sq_r - (WebRtc_Word64) sum_r * sum_r + c2);
return ssim_n * 1.0 / ssim_d;
}
static double
Ssim8x8C(WebRtc_UWord8 *s, WebRtc_Word32 sp,
WebRtc_UWord8 *r, WebRtc_Word32 rp)
{
WebRtc_UWord64 sum_s = 0;
WebRtc_UWord64 sum_r = 0;
WebRtc_UWord64 sum_sq_s = 0;
WebRtc_UWord64 sum_sq_r = 0;
WebRtc_UWord64 sum_sxr = 0;
WebRtc_Word32 i, j;
for (i = 0; i < 8; i++, s += sp,r += rp)
{
for (j = 0; j < 8; j++)
{
sum_s += s[j];
sum_r += r[j];
sum_sq_s += s[j] * s[j];
sum_sq_r += r[j] * r[j];
sum_sxr += s[j] * r[j];
}
}
return Similarity(sum_s, sum_r, sum_sq_s, sum_sq_r, sum_sxr, 64);
}
#if defined(WEBRTC_USE_SSE2)
#include <emmintrin.h>
#include <xmmintrin.h>
static double
Ssim8x8Sse2(WebRtc_UWord8 *s, WebRtc_Word32 sp,
WebRtc_UWord8 *r, WebRtc_Word32 rp)
{
WebRtc_Word32 i;
const __m128i z = _mm_setzero_si128();
__m128i sum_s_16 = _mm_setzero_si128();
__m128i sum_r_16 = _mm_setzero_si128();
__m128i sum_sq_s_32 = _mm_setzero_si128();
__m128i sum_sq_r_32 = _mm_setzero_si128();
__m128i sum_sxr_32 = _mm_setzero_si128();
for (i = 0; i < 8; i++, s += sp,r += rp)
{
const __m128i s_8 = _mm_loadl_epi64((__m128i*)(s));
const __m128i r_8 = _mm_loadl_epi64((__m128i*)(r));
const __m128i s_16 = _mm_unpacklo_epi8(s_8,z);
const __m128i r_16 = _mm_unpacklo_epi8(r_8,z);
sum_s_16 = _mm_adds_epu16(sum_s_16, s_16);
sum_r_16 = _mm_adds_epu16(sum_r_16, r_16);
const __m128i sq_s_32 = _mm_madd_epi16(s_16, s_16);
sum_sq_s_32 = _mm_add_epi32(sum_sq_s_32, sq_s_32);
const __m128i sq_r_32 = _mm_madd_epi16(r_16, r_16);
sum_sq_r_32 = _mm_add_epi32(sum_sq_r_32, sq_r_32);
const __m128i sxr_32 = _mm_madd_epi16(s_16, r_16);
sum_sxr_32 = _mm_add_epi32(sum_sxr_32, sxr_32);
}
const __m128i sum_s_32 = _mm_add_epi32(_mm_unpackhi_epi16(sum_s_16, z),
_mm_unpacklo_epi16(sum_s_16, z));
const __m128i sum_r_32 = _mm_add_epi32(_mm_unpackhi_epi16(sum_r_16, z),
_mm_unpacklo_epi16(sum_r_16, z));
__m128i sum_s_128;
__m128i sum_r_128;
__m128i sum_sq_s_128;
__m128i sum_sq_r_128;
__m128i sum_sxr_128;
_mm_store_si128 (&sum_s_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_s_32, z),
_mm_unpacklo_epi32(sum_s_32, z)));
_mm_store_si128 (&sum_r_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_r_32, z),
_mm_unpacklo_epi32(sum_r_32, z)));
_mm_store_si128 (&sum_sq_s_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_sq_s_32, z),
_mm_unpacklo_epi32(sum_sq_s_32, z)));
_mm_store_si128 (&sum_sq_r_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_sq_r_32, z),
_mm_unpacklo_epi32(sum_sq_r_32, z)));
_mm_store_si128 (&sum_sxr_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_sxr_32, z),
_mm_unpacklo_epi32(sum_sxr_32, z)));
const WebRtc_UWord64 *sum_s_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_s_128);
const WebRtc_UWord64 *sum_r_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_r_128);
const WebRtc_UWord64 *sum_sq_s_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_sq_s_128);
const WebRtc_UWord64 *sum_sq_r_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_sq_r_128);
const WebRtc_UWord64 *sum_sxr_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_sxr_128);
const WebRtc_UWord64 sum_s = sum_s_64[0] + sum_s_64[1];
const WebRtc_UWord64 sum_r = sum_r_64[0] + sum_r_64[1];
const WebRtc_UWord64 sum_sq_s = sum_sq_s_64[0] + sum_sq_s_64[1];
const WebRtc_UWord64 sum_sq_r = sum_sq_r_64[0] + sum_sq_r_64[1];
const WebRtc_UWord64 sum_sxr = sum_sxr_64[0] + sum_sxr_64[1];
return Similarity(sum_s, sum_r, sum_sq_s, sum_sq_r, sum_sxr, 64);
}
#endif
double
SsimFrame(WebRtc_UWord8 *img1, WebRtc_UWord8 *img2, WebRtc_Word32 stride_img1,
WebRtc_Word32 stride_img2, WebRtc_Word32 width, WebRtc_Word32 height)
{
WebRtc_Word32 i,j;
WebRtc_UWord32 samples = 0;
double ssim_total = 0;
double (*ssim_8x8)(WebRtc_UWord8*, WebRtc_Word32,
WebRtc_UWord8*, WebRtc_Word32 rp);
ssim_8x8 = Ssim8x8C;
if (WebRtc_GetCPUInfo(kSSE2))
{
#if defined(WEBRTC_USE_SSE2)
ssim_8x8 = Ssim8x8Sse2;
#endif
}
// Sample point start with each 4x4 location
for (i = 0; i < height - 8; i += 4, img1 += stride_img1 * 4,
img2 += stride_img2 * 4)
{
for (j = 0; j < width - 8; j += 4 )
{
double v = ssim_8x8(img1 + j, stride_img1, img2 + j, stride_img2);
ssim_total += v;
samples++;
}
}
ssim_total /= samples;
return ssim_total;
}
WebRtc_Word32
SsimFromFiles(const WebRtc_Word8 *refFileName, const WebRtc_Word8 *testFileName,
WebRtc_Word32 width, WebRtc_Word32 height, QualityMetricsResult *result)
{
FILE *refFp = fopen(refFileName, "rb");
if (refFp == NULL)
{
// cannot open reference file
fprintf(stderr, "Cannot open file %s\n", refFileName);
return -1;
}
FILE *testFp = fopen(testFileName, "rb");
if (testFp == NULL)
{
// cannot open test file
fprintf(stderr, "Cannot open file %s\n", testFileName);
return -2;
}
WebRtc_Word32 frames = 0;
// Bytes in one frame I420
const WebRtc_Word32 frameBytes = 3 * width * height / 2;
WebRtc_UWord8 *ref = new WebRtc_UWord8[frameBytes];
WebRtc_UWord8 *test = new WebRtc_UWord8[frameBytes];
WebRtc_Word32 refBytes = (WebRtc_Word32) fread(ref, 1, frameBytes, refFp);
WebRtc_Word32 testBytes = (WebRtc_Word32) fread(test, 1, frameBytes, testFp);
double ssimScene = 0.0; //average SSIM for sequence
while (refBytes == frameBytes && testBytes == frameBytes )
{
double ssimFrameValue = SsimFrame(ref, test, width, width, width, height);
// Save statistics for this specific frame
FrameResult frame_result;
frame_result.value = ssimFrameValue;
frame_result.frame_number = frames;
result->frames.push_back(frame_result);
ssimScene += ssimFrameValue;
frames++;
refBytes = (WebRtc_Word32) fread(ref, 1, frameBytes, refFp);
testBytes = (WebRtc_Word32) fread(test, 1, frameBytes, testFp);
}
// SSIM: normalize/average for sequence
ssimScene = ssimScene / frames;
result->average = ssimScene;
// Calculate min/max statistics
std::vector<FrameResult>::iterator element;
element = min_element(result->frames.begin(),
result->frames.end(), lessForFrameResultValue);
result->min = element->value;
result->min_frame_number = element->frame_number;
element = max_element(result->frames.begin(),
result->frames.end(), lessForFrameResultValue);
result->max = element->value;
result->max_frame_number = element->frame_number;
delete [] ref;
delete [] test;
fclose(refFp);
fclose(testFp);
return 0;
}