转载自:http://blog.csdn.net/chenhuaijin/article/details/9292811
Color Transfer between images 是一片比较经典的色调映射的文章,应用率较高,算法也比较简单。适合做较为简单的色调处理。
算法主要步骤;
1. RGB转化为Lab
2.计算lab空间每个channel的 均值和方差。
3.根据公式
实验结果:
- #include <opencv2/opencv.hpp>
- #include <opencv2/core/core.hpp>
- #include <opencv2/imgproc/imgproc.hpp>
- using namespace std;
- using namespace cv;
- class ColorTranfer
- {
- private:
- Mat srcImg;
- Mat targetImg;
- Mat srcImg_lab;//CV_32FC3
- Mat targetImg_lab;//CV_32FC3
- //Mat result_lab;
- vector<float> srcMeans;
- vector<float> srcVariances;
- vector<float> targetMeans;
- vector<float> targetVariances;
- void initialize()
- {
- Mat srcImg_32F, targetImg_32F;
- srcImg.convertTo(srcImg_32F, CV_32FC3,1.0f/255.f);
- targetImg.convertTo(targetImg_32F, CV_32FC3, 1.0f/255.0f);
- cvtColor(srcImg_32F, srcImg_lab, CV_BGR2Lab );
- cvtColor(targetImg_32F, targetImg_lab, CV_BGR2Lab);
- //if(srcImg_lab.depth() == CV_32FC3) cout << "Yes" << endl;
- cout << srcImg_lab.depth() << endl;
- cout << srcImg.depth() << endl;
- computeMeans();
- computeVariances();
- }
- void computeMeans(){
- int srcPixels = srcImg.rows * srcImg.cols;
- int targetPixels = targetImg.rows * targetImg.cols;
- //computing the mean of source image in lab space
- float sum[3] ={ 0 };
- for (int row = 0; row < srcImg.rows; row++)
- {
- for (int col = 0; col < srcImg.cols; col++)
- {
- //Point pt(col, row);
- Vec3f color = srcImg_lab.at<Vec3f>(row, col);
- for(int layer = 0; layer < 3; layer++)
- sum[layer] += color[layer];
- }
- }
- for (int i = 0; i < 3; i++)
- srcMeans[i] = sum[i] / srcPixels;
- //computing the mean of target image int lab color space
- for(int i = 0; i < 3; i++)
- sum[i] = 0;
- for (int row = 0; row < targetImg.rows; row++)
- {
- for (int col = 0; col < targetImg.cols; col++)
- {
- //Point pt(col, row);
- Vec3f color = targetImg_lab.at<Vec3f>(row, col);
- for(int layer = 0; layer < 3; layer++)
- sum[layer] += color[layer];
- }
- }
- for (int i = 0; i < 3; i++)
- targetMeans[i] = sum[i] / targetPixels;
- }//end function compuetMeans
- void computeVariances(){
- int srcPixels = srcImg_lab.cols * srcImg_lab.rows;
- int targetPixels = targetImg_lab.cols * targetImg_lab.rows;
- //computing the variance of source image
- float sum_variance[3] ={0.f};
- for (int y = 0; y < srcImg_lab.rows; y++)
- {
- for (int x = 0; x < srcImg_lab.cols; x++)
- {
- Vec3f color = srcImg_lab.at<Vec3f>(y,x);
- sum_variance[0] += (color[0] - srcMeans[0])*(color[0] - srcMeans[0]);
- sum_variance[1] += (color[1] - srcMeans[1])*(color[1] - srcMeans[1]);
- sum_variance[2] += (color[2] - srcMeans[2])*(color[2] - srcMeans[2]);
- }
- }
- srcVariances[0] = sqrt( sum_variance[0] / srcPixels );
- srcVariances[1] = sqrt( sum_variance[1] / srcPixels );
- srcVariances[2] = sqrt( sum_variance[2] / srcPixels );
- //computing the variance of target image
- for(int i = 0; i < 3; i++)
- sum_variance[i] = 0.f;
- for (int y = 0; y < targetImg_lab.rows; y++)
- {
- for (int x = 0; x < targetImg_lab.cols; x++)
- {
- Vec3f color = targetImg_lab.at<Vec3f>(y,x);
- sum_variance[0] += (color[0] - targetMeans[0])*(color[0] - targetMeans[0]);
- sum_variance[1] += (color[1] - targetMeans[1])*(color[1] - targetMeans[1]);
- sum_variance[2] += (color[2] - targetMeans[2])*(color[2] - targetMeans[2]);
- }
- }
- targetVariances[0] = sqrt( sum_variance[0] / targetPixels );
- targetVariances[1] = sqrt( sum_variance[1] / targetPixels );
- targetVariances[2] = sqrt( sum_variance[2] / targetPixels );
- }
- public:
- Mat result;
- void solve()
- {
- initialize();
- //constructing the final value in very pixel and store the value in result matrix
- int width = srcImg.cols;
- int height = srcImg.rows;
- //Mat result(height, width, CV_32FC3);
- Mat result_lab(height, width, CV_32FC3);
- float deta_rate[3];
- for (int k = 0; k < 3; k++)
- {
- deta_rate[k] = targetVariances[k] / srcVariances[k];
- }
- for (int y = 0; y < height; y++ )
- {
- for (int x = 0; x < width; x++)
- {
- Vec3f color = srcImg_lab.at<Vec3f>(y,x);
- Vec3f value;
- for(int channel = 0; channel < 3; channel++)
- value[channel] = deta_rate[channel]*(color[channel] - srcMeans[channel]) + targetMeans[channel];
- result_lab.at<Vec3f>(y,x) = value;
- }
- }
- //construct the final image
- cvtColor(result_lab, result, CV_Lab2BGR);
- }
- ColorTranfer(Mat src, Mat target):srcImg(src), targetImg(target)
- {
- srcMeans.resize(3, 0.f);
- srcVariances.resize(3, 0.f);
- targetMeans.resize(3, 0.f);
- targetVariances.resize(3, 0.f);
- solve();
- }
- };
- int main()
- {
- Mat src = imread("11.jpg");
- Mat target = imread("12.jpg");
- namedWindow("src");
- imshow("src", src);
- namedWindow("target");
- imshow("target", target);
- ColorTranfer clt(src, target);
- namedWindow("result");
- imshow("result", clt.result);
- waitKey(0);
- return 0;
- }