OpenCV 4 with Python Blueprints - Second Edition

Implementing a Gaussian blur with two-dimensional convolution

A Gaussian blur is implemented by convolving the image with a kernel of Gaussian values. Two- dimensional convolution is something that is used very widely in image processing. Usually, we have a big picture (let's look at a 5 x 5 subsection of that particular image), and we have a kernel (or filter) that is another matrix of a smaller size (in our example, 3 x 3).

In order to get the convolution values, let's suppose that we want to get the value at location (2, 3). We place the kernel centered at the location (2, 3), and we calculate the pointwise product of the overlay matrix (highlighted area, in the following image (red color)) with the kernel and take the overall sum. The ...