Opencv basic operation 2 realizes label2rgb and converts gray-scale images into color images

All the images obtained by semantic segmentation are gray-scale images , It is not conducive to viewing the segmentation effect , Through to label The value changes linearly three times , You can get information about R、G、B Pseudo color of three channels , Then generate a pseudo color image .

The specific code is as follows ： For different class_num Of label Data sets , Just adjust the code class_num that will do

cv::Mat set_color(cv::Mat label) {
    int class_num = 8;// One has 8 Categories 
    int color_setp = int(255 / class_num);
    cv::Mat tmp,bg;
    label.copyTo(tmp);
    // Set the background to white 
    // Put more than 0 All the parts of the are assigned to 0, Less than or equal to 0 The part of is assigned as 255
    threshold(tmp, bg, 0, 255, THRESH_BINARY_INV);

    cv::Mat G = (class_num - label) * color_setp;
    cv::Mat R = label * color_setp;
    cv::Mat B = (R + G) / 2;
    B.convertTo(B, CV_8UC1);

    // Let the background be white 
    vector<cv::Mat> mv;
    mv.push_back(R + bg);
    mv.push_back(G + bg);
    mv.push_back(B + bg);

    cv::Mat RGB;
    cv::merge(mv, RGB);
    return RGB;
}

The effect of the generated pseudo color image is shown in the figure 1 Shown , The background is white .

chart 1 Generated pseudo color image