/* FasTC * Copyright (c) 2012 University of North Carolina at Chapel Hill. All rights reserved. * * Permission to use, copy, modify, and distribute this software and its documentation for educational, * research, and non-profit purposes, without fee, and without a written agreement is hereby granted, * provided that the above copyright notice, this paragraph, and the following four paragraphs appear * in all copies. * * Permission to incorporate this software into commercial products may be obtained by contacting the * authors or the Office of Technology Development at the University of North Carolina at Chapel Hill . * * This software program and documentation are copyrighted by the University of North Carolina at Chapel Hill. * The software program and documentation are supplied "as is," without any accompanying services from the * University of North Carolina at Chapel Hill or the authors. The University of North Carolina at Chapel Hill * and the authors do not warrant that the operation of the program will be uninterrupted or error-free. The * end-user understands that the program was developed for research purposes and is advised not to rely * exclusively on the program for any reason. * * IN NO EVENT SHALL THE UNIVERSITY OF NORTH CAROLINA AT CHAPEL HILL OR THE AUTHORS BE LIABLE TO ANY PARTY FOR * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING LOST PROFITS, ARISING OUT OF THE * USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN IF THE UNIVERSITY OF NORTH CAROLINA AT CHAPEL HILL OR THE * AUTHORS HAVE BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * * THE UNIVERSITY OF NORTH CAROLINA AT CHAPEL HILL AND THE AUTHORS SPECIFICALLY DISCLAIM ANY WARRANTIES, INCLUDING, * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE AND ANY * STATUTORY WARRANTY OF NON-INFRINGEMENT. THE SOFTWARE PROVIDED HEREUNDER IS ON AN "AS IS" BASIS, AND THE UNIVERSITY * OF NORTH CAROLINA AT CHAPEL HILL AND THE AUTHORS HAVE NO OBLIGATIONS TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, * ENHANCEMENTS, OR MODIFICATIONS. * * Please send all BUG REPORTS to . * * The authors may be contacted via: * * Pavel Krajcevski * Dept of Computer Science * 201 S Columbia St * Frederick P. Brooks, Jr. Computer Science Bldg * Chapel Hill, NC 27599-3175 * USA * * */ #include "Image.h" #include #include #include #include #include #include #include "Color.h" #include "Pixel.h" #include "IPixel.h" template static inline T sad( const T &a, const T &b ) { return (a > b)? a - b : b - a; } // wtf #ifdef _MSC_VER template T log2(T x) { return static_cast(log((long double)x) / log(2.0)); } #endif namespace FasTC { template Image::Image(uint32 width, uint32 height) : m_Width(width) , m_Height(height) , m_Pixels(new PixelType[GetNumPixels()]) { } template Image::Image(uint32 width, uint32 height, const PixelType *pixels) : m_Width(width) , m_Height(height) { if(pixels) { m_Pixels = new PixelType[GetNumPixels()]; memcpy(m_Pixels, pixels, GetNumPixels() * sizeof(PixelType)); } else { m_Pixels = 0; } } template Image::Image(const Image &other) : m_Width(other.m_Width) , m_Height(other.m_Height) , m_Pixels(new PixelType[GetNumPixels()]) { memcpy(m_Pixels, other.m_Pixels, GetNumPixels() * sizeof(PixelType)); } template bool Image::ReadPixels(const uint32 *rgba) { assert(m_Pixels); for(uint32 i = 0; i < GetNumPixels(); i++) { m_Pixels[i].Unpack(rgba[i]); } return true; } template Image::Image(uint32 width, uint32 height, const uint32 *pixels) : m_Width(width) , m_Height(height) { if(pixels) { m_Pixels = new PixelType[GetNumPixels()]; ReadPixels(pixels); } else { m_Pixels = NULL; } } template Image::~Image() { if(m_Pixels) { delete [] m_Pixels; m_Pixels = 0; } } template Image &Image::operator=(const Image &other) { m_Width = other.m_Width; m_Height = other.m_Height; if(m_Pixels) { delete [] m_Pixels; } if(other.m_Pixels) { m_Pixels = new PixelType[GetNumPixels()]; if(m_Pixels) memcpy(m_Pixels, other.m_Pixels, GetNumPixels() * sizeof(PixelType)); else fprintf(stderr, "Out of memory!\n"); } else { m_Pixels = NULL; } return *this; } template PixelType & Image::operator()(uint32 i, uint32 j) { assert(i < GetWidth()); assert(j < GetHeight()); return m_Pixels[j * GetWidth() + i]; } template const PixelType & Image::operator()(uint32 i, uint32 j) const { assert(i < GetWidth()); assert(j < GetHeight()); return m_Pixels[j * GetWidth() + i]; } template double Image::ComputePSNR(Image *other) { if(!other) return -1.0; if(GetWidth() != other->GetWidth() || GetHeight() != other->GetHeight()) { return -1.0; } // Compute raw 8-bit RGBA data... ComputePixels(); other->ComputePixels(); const PixelType *ourPixels = GetPixels(); const PixelType *otherPixels = other->GetPixels(); // const double w[3] = { 0.2126, 0.7152, 0.0722 }; const double w[3] = { 1.0, 1.0, 1.0 }; double mse = 0.0; const uint32 imageSz = GetNumPixels(); for(uint32 i = 0; i < imageSz; i++) { uint32 ourPixel = ourPixels[i].Pack(); uint32 otherPixel = otherPixels[i].Pack(); double r[4], u[4]; for(uint32 c = 0; c < 4; c++) { uint32 shift = c * 8; if(c == 3) { r[c] = static_cast((ourPixel >> shift) & 0xFF) / 255.0; u[c] = static_cast((otherPixel >> shift) & 0xFF) / 255.0; } else { r[c] = static_cast((ourPixel >> shift) & 0xFF) * w[c]; u[c] = static_cast((otherPixel >> shift) & 0xFF) * w[c]; } } for(uint32 c = 0; c < 3; c++) { double diff = (r[3] * r[c] - u[3] * u[c]); mse += diff * diff; } } mse /= GetWidth() * GetHeight(); const double C = 255.0 * 255.0; double maxi = (w[0]*w[0] + w[1]*w[1] + w[2]*w[2]) * C; return 10 * log10(maxi/mse); } static Image FilterValid(const Image &img, uint32 size, double sigma) { assert(size % 2); Image gaussian(size, size); GenerateGaussianKernel(gaussian, size, static_cast(sigma)); double sum = 0.0; for(uint32 j = 0; j < size; j++) { for(uint32 i = 0; i < size; i++) { sum += static_cast(gaussian(i, j)); } } for(uint32 j = 0; j < size; j++) { for(uint32 i = 0; i < size; i++) { double v = static_cast(gaussian(i, j)); gaussian(i, j) = static_cast(v / sum); } } int32 h = static_cast(img.GetHeight()); int32 w = static_cast(img.GetWidth()); Image out(img.GetWidth() - size + 1, img.GetHeight() - size + 1); int32 halfSz = static_cast(size) >> 1; for(int32 j = halfSz; j < h-halfSz; j++) { for(int32 i = halfSz; i < w-halfSz; i++) { int32 xoffset = -halfSz; int32 yoffset = -halfSz; double result = 0; for(int32 y = 0; y < static_cast(size); y++) for(int32 x = 0; x < static_cast(size); x++) { double s = static_cast(gaussian(x, y)); result += s * static_cast(img(i+xoffset+x, j+yoffset+y)); } out(i+xoffset, j+yoffset) = static_cast(result); } } return out; } template double Image::ComputeSSIM(Image *other) { if(!other) { return -1.0; } if(GetWidth() != other->GetWidth() || GetHeight() != other->GetHeight()) { return -1.0; } ComputePixels(); other->ComputePixels(); double C1 = (0.01 * 255.0 * 0.01 * 255.0); double C2 = (0.03 * 255.0 * 0.03 * 255.0); Image img1(GetWidth(), GetHeight()); Image img2(GetWidth(), GetHeight()); ConvertTo(img1); other->ConvertTo(img2); for(uint32 j = 0; j < GetHeight(); j++) { for(uint32 i = 0; i < GetWidth(); i++) { img1(i, j) = 255.0f * static_cast(img1(i, j)); img2(i, j) = 255.0f * static_cast(img2(i, j)); } } /* Matlab code taken from http://www.cns.nyu.edu/lcv/ssim/ssim_index.m C1 = (K(1)*L)^2; C2 = (K(2)*L)^2; window = window/sum(sum(window)); img1 = double(img1); img2 = double(img2); mu1 = filter2(window, img1, 'valid'); mu2 = filter2(window, img2, 'valid'); mu1_sq = mu1.*mu1; mu2_sq = mu2.*mu2; mu1_mu2 = mu1.*mu2; sigma1_sq = filter2(window, img1.*img1, 'valid') - mu1_sq; sigma2_sq = filter2(window, img2.*img2, 'valid') - mu2_sq; sigma12 = filter2(window, img1.*img2, 'valid') - mu1_mu2; ssim_map = ((2*mu1_mu2 + C1).*(2*sigma12 + C2))./ ((mu1_sq + mu2_sq + C1).*(sigma1_sq + sigma2_sq + C2)); */ const uint32 filterSz = 11; const double filterSigma = 1.5; Image mu1 = FilterValid(img1, filterSz, filterSigma); Image mu2 = FilterValid(img2, filterSz, filterSigma); assert(mu1.GetHeight() == mu2.GetHeight()); assert(mu1.GetWidth() == mu2.GetWidth()); Image mu1_sq(mu1); Image mu2_sq(mu2); Image mu1_mu2(mu1); Image sigma1_sq(img1); Image sigma2_sq(img2); Image sigma12(img1); uint32 w = ::std::max(img1.GetWidth(), mu1.GetWidth()); uint32 h = ::std::max(img1.GetHeight(), mu1.GetHeight()); for(uint32 j = 0; j < h; j++) { for(uint32 i = 0; i < w; i++) { if(i < mu1.GetWidth() && j < mu1.GetHeight()) { double m1 = static_cast(mu1(i, j)); double m2 = static_cast(mu2(i, j)); mu1_sq(i, j) = static_cast(m1 * m1); mu2_sq(i, j) = static_cast(m2 * m2); mu1_mu2(i, j) = static_cast(m1 * m2); } if(i < img1.GetWidth() && j < img1.GetHeight()) { double i1 = static_cast(img1(i, j)); double i2 = static_cast(img2(i, j)); sigma1_sq(i, j) = static_cast(i1 * i1); sigma2_sq(i, j) = static_cast(i2 * i2); sigma12(i, j) = static_cast(i1 * i2); } } } sigma1_sq = FilterValid(sigma1_sq, filterSz, filterSigma); sigma2_sq = FilterValid(sigma2_sq, filterSz, filterSigma); sigma12 = FilterValid(sigma12, filterSz, filterSigma); assert(sigma1_sq.GetWidth() == mu1.GetWidth()); assert(sigma1_sq.GetHeight() == mu1.GetHeight()); assert(sigma2_sq.GetWidth() == mu1.GetWidth()); assert(sigma2_sq.GetHeight() == mu1.GetHeight()); assert(sigma12.GetWidth() == mu1.GetWidth()); assert(sigma12.GetHeight() == mu1.GetHeight()); w = mu1_sq.GetWidth(); h = mu2_sq.GetHeight(); for(uint32 j = 0; j < h; j++) { for(uint32 i = 0; i < w; i++) { double m1sq = static_cast(mu1_sq(i, j)); double m2sq = static_cast(mu2_sq(i, j)); double m1m2 = static_cast(mu1_mu2(i, j)); double s1sq = static_cast(sigma1_sq(i, j)); double s2sq = static_cast(sigma2_sq(i, j)); double s1s2 = static_cast(sigma12(i, j)); sigma1_sq(i, j) = static_cast(s1sq - m1sq); sigma2_sq(i, j) = static_cast(s2sq - m2sq); sigma12(i, j) = static_cast(s1s2 - m1m2); } } double mssim = 0.0; for(uint32 j = 0; j < h; j++) { for(uint32 i = 0; i < w; i++) { double m1sq = static_cast(mu1_sq(i, j)); double m2sq = static_cast(mu2_sq(i, j)); double m1m2 = static_cast(mu1_mu2(i, j)); double s1sq = static_cast(sigma1_sq(i, j)); double s2sq = static_cast(sigma2_sq(i, j)); double s1s2 = static_cast(sigma12(i, j)); double ssim = ((2.0 * m1m2 + C1) * (2.0 * s1s2 + C2)) / ((m1sq + m2sq + C1) * (s1sq + s2sq + C2)); mssim += ssim; } } return mssim / static_cast(w * h); } template double Image::ComputeMeanLocalEntropy() { const uint32 kKernelSz = 15; const uint32 kHalfKernelSz = kKernelSz / 2; Image entropyIdx(GetWidth() - kKernelSz + 1, GetHeight() - kKernelSz + 1); for(uint32 j = kHalfKernelSz; j < GetHeight() - kHalfKernelSz; j++) { for(uint32 i = kHalfKernelSz; i < GetWidth() - kHalfKernelSz; i++) { Image subImg(kKernelSz, kKernelSz); for(uint32 y = 0; y < kKernelSz; y++) for(uint32 x = 0; x < kKernelSz; x++) { subImg(x, y) = (*this)(i - kHalfKernelSz + x, j - kHalfKernelSz + y); } entropyIdx(i-kHalfKernelSz, j-kHalfKernelSz) = static_cast(subImg.ComputeEntropy()); } } double sum = 0; for(uint32 j = 0; j < entropyIdx.GetHeight(); j++) for(uint32 i = 0; i < entropyIdx.GetWidth(); i++) { sum += static_cast(entropyIdx(i, j)); } return sum / (entropyIdx.GetHeight() * entropyIdx.GetWidth()); } template double Image::ComputeEntropy() { uint32 hist[256]; memset(hist, 0, sizeof(hist)); ComputePixels(); Image intensity(GetWidth(), GetHeight()); ConvertTo(intensity); for(uint32 j = 0; j < GetHeight(); j++) { for(uint32 i = 0; i < GetWidth(); i++) { float iflt = static_cast(intensity(i, j)); uint32 iv = static_cast(iflt * 255.0f + 0.5f); assert(iv < 256); hist[iv]++; } } double ret = 0; for(uint32 i = 0; i < 256; i++) { if(hist[i] > 0) { float p = static_cast(hist[i]) / static_cast(GetHeight() * GetWidth()); ret += p * log2(p); } } return -ret; } template void Image::SetImageData(uint32 width, uint32 height, PixelType *data) { if(m_Pixels) { delete m_Pixels; } if(!data) { width = 0; height = 0; m_Pixels = NULL; } else { m_Width = width; m_Height = height; m_Pixels = data; } } template static inline T Clamp(const T &v, const T &a, const T &b) { return ::std::min(::std::max(a, v), b); } template void Image::Filter(const Image &kernel) { Image k(kernel); // Only odd sized filters make sense.... assert(k.GetWidth() % 2); assert(k.GetHeight() % 2); double sum = 0.0; for(uint32 j = 0; j < k.GetHeight(); j++) { for(uint32 i = 0; i < k.GetWidth(); i++) { sum += static_cast(k(i, j)); } } for(uint32 j = 0; j < k.GetHeight(); j++) { for(uint32 i = 0; i < k.GetWidth(); i++) { k(i, j) = static_cast(k(i, j) / sum); } } int32 ih = static_cast(GetHeight()); int32 iw = static_cast(GetWidth()); int32 kh = static_cast(k.GetHeight()); int32 kw = static_cast(k.GetWidth()); Image filtered(iw, ih); for(int32 j = 0; j < ih; j++) { for(int32 i = 0; i < iw; i++) { int32 yoffset = j - (k.GetHeight() / 2); int32 xoffset = i - (k.GetWidth() / 2); Color newPixel; for(int32 y = 0; y < kh; y++) { for(int32 x = 0; x < kw; x++) { PixelType pixel = ((*this)( Clamp(x + xoffset, 0, GetWidth() - 1), Clamp(y + yoffset, 0, GetHeight() - 1))); Color c; c.Unpack(pixel.Pack()); Color scaled = c * static_cast(k(x, y)); newPixel += scaled; } } filtered(i, j).Unpack(newPixel.Pack()); } } *this = filtered; } template class Image; template class Image; template class Image; void GenerateGaussianKernel(Image &out, uint32 size, float sigma) { assert(size % 2); out = Image(size, size); if(size == 1) { out(0, 0) = 1.0f; return; } int32 halfSz = static_cast(size) / 2; for(int32 j = -halfSz; j <= halfSz; j++) { for(int32 i = -halfSz; i <= halfSz; i++) { out(halfSz + i, halfSz + j) = exp(- (j*j + i*i) / (2*sigma*sigma)); } } } } // namespace FasTC