Use of OpenCV 6.4 median filter

Nonlinear filter

The mean filter introduced in the previous sections 、 Gaussian filters are linear filters , That is, replace the pixel value with the weighted accumulated value of the neighborhood pixel .

This section mainly introduces nonlinear filters , It has no kernel , And convolution operation is not allowed , However, the output pixel value is also determined by identifying the neighborhood pixel value .

Commonly used nonlinear filters are Median filter , Median filter Set the current pixel and its neighborhood , Calculate the middle value of the set as the current pixel value , It is very effective in eliminating salt and pepper noise .

cv::Mat image = cv::imread("salted.bmp");
cv::Mat median;
cv::medianBlur(image, median, 5);
cv::imshow("image", image);
cv::imshow("median blur", median);

It's obvious , Median filter is very efficient in eliminating noise , It can well identify abnormal black or white pixels and then replace them , The mean filter will be greatly affected by noise .

In addition, median filter It helps to preserve the sharpness of the edge , But it will wash away the texture in the homogeneous area , It will have a good visual effect , Similar to ink style .

cv::medianBlur Use of functions

Function signature

CV_EXPORTS_W void medianBlur( InputArray src, OutputArray dst, int ksize );

• The parameters are ： The input image , Output image , Filter size .
• Different effects can be achieved by modifying different filter sizes . The following figure shows changing the filter size to 9 The effect of the .