OpenCV——LCC（Local Color Correction）的Python复现

最新推荐文章于 2024-09-13 18:11:35 发布

KingOfCinders

最新推荐文章于 2024-09-13 18:11:35 发布

阅读量3.6k

点赞数

分类专栏：毕设notes 文章标签： python 图像处理

本文链接：https://blog.csdn.net/GongYing2024/article/details/123983516

版权

毕设notes 专栏收录该内容

1 篇文章 0 订阅

订阅专栏

本文解析了《LocalColorCorrection》论文，介绍了使用C++和Python实现的局部色彩校正方法，通过计算和调整图像像素值，实现色彩增强和校准。涉及步骤包括灰度转换、高斯模糊和指数变换，适用于图像处理和视觉效果应用。

摘要由CSDN通过智能技术生成

前言

《Local Color Correction 》论文地址：LCC
参考知乎文章：LCC论文阅读和C++复现

Python代码

def LCC_ColorImg(src):
    rows = np.shape(src)[0]
    cols = np.shape(src)[1]
    
    I = np.arange(rows * cols).reshape((rows, cols))
    inv_I = np.arange(rows * cols).reshape((rows, cols))
    Mast = np.empty([rows, cols, 1], dtype=np.uint8)
    for i in range(0, rows):
        data1 = Mast[i]
        for j in range(0, cols):
            I[i][j] = (src[i, j][0] + src[i, j][1] + src[i, j][2]) / 3.0
            inv_I[i][j] = 255
            data1[j] = inv_I[i][j] - I[i][j]
  
    Mast = cv2.GaussianBlur(Mast, (41, 41), 0)
    dst = np.empty([rows, cols, 3], dtype=np.uint8)
    for i in range(0, rows):
        data2 = Mast[i]
        for j in range(0, cols):
            for k in range(0, 3):
                Exp = math.pow(2, (128 - data2[j]) / 128.0)
                value = int(255 * math.pow(src[i, j][k] / 255.0, Exp))
                dst[i, j][k] = value
                
    return dst

参考的C++代码

Mat LCC(const Mat &src){
    int rows = src.rows;
    int cols = src.cols;
    int **I;
    I = new int *[rows];
    for(int i = 0; i < rows; i++){
        I[i] = new int [cols];
    }
    int **inv_I;
    inv_I = new int *[rows];
    for(int i = 0; i < rows; i++){
        inv_I[i] = new int [cols];
    }
    Mat Mast(rows, cols, CV_8UC1);
    for(int i = 0; i < rows; i++){
        uchar *data = Mast.ptr<uchar>(i);
        for(int j = 0; j < cols; j++){
            I[i][j] = (src.at<Vec3b>(i, j)[0] + src.at<Vec3b>(i, j)[1] + src.at<Vec3b>(i, j)[1]) / 3.0;
            inv_I[i][j] = 255;
            *data = inv_I[i][j] - I[i][j];
            data++;
        }
    }
    GaussianBlur(Mast, Mast, Size(41, 41), BORDER_DEFAULT);
    Mat dst(rows, cols, CV_8UC3);
    for(int i = 0; i < rows; i++){
        uchar *data = Mast.ptr<uchar>(i);
        for(int j = 0; j < cols; j++){
            for(int k = 0; k < 3; k++){
                float Exp = pow(2, (128 - data[j]) / 128.0);
                int value = int(255 * pow(src.at<Vec3b>(i, j)[k] / 255.0, Exp));
                dst.at<Vec3b>(i, j)[k] = value;
            }
        }
    }
    return dst;
}