opencv 進行圖像的花屏檢測（模糊檢測）

參考:

https://www.pyimagesearch.com/2015/09/07/blur-detection-with-opencv/

https://www.cnblogs.com/arkenstone/p/7900978.html

先對圖像用拉普拉斯算子進行濾波，然后求取得到的結果圖像的方差，如果方差小於一定值則圖片視為模糊。利用python很好實現：

img2gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)  # 將圖片壓縮為單通道的灰度圖
score = cv2.Laplacian(img2gray, cv2.CV_64F).var()

C++實現如下：

bool isImageBlurry(cv::Mat& img)
{
    cv::Mat matImageGray;
    // converting image's color space (RGB) to grayscale
    cv::cvtColor(img, matImageGray, CV_BGR2GRAY);
    cv::Mat dst, abs_dst;
        cv::Laplacian(matImageGray, dst, CV_16S, 3);
        cv::convertScaleAbs( dst, abs_dst );
    cv::Mat tmp_m, tmp_sd;  
    double m = 0, sd = 0;  
    int threshold = 1000;
    cv::meanStdDev(dst, tmp_m, tmp_sd);  
    m = tmp_m.at<double>(0,0);  
    sd = tmp_sd.at<double>(0,0);  
    std::cout << "StdDev: " << sd * sd << std::endl; 
    return ((sd * sd) <= threshold);
}