Artificial Intelligence for Robotics

Convolution

Every once in a while, you come across some mathematical construction that turns a complex task into just a bunch of adding, subtracting, multiplying, and dividing. Vectors in geometry work like that, and, in image processing, we have the convolution kernel. It transpires that most of the common image processing techniques – edge detection, corner detection, blurring, sharpening, enhancing, and so on, can be accomplished with a simple array construct.

It is pretty easy to understand that in an image, the neighbors of a pixel are just as important to what a pixel is as the pixel itself. If you were going to try and find all the edge pixels of a box, you would look for a pixel that has one type of color on one side, and another ...