1. Convolution
Convolution is an old friend of the signal processing field. The original definition is as follows
In machine learning terms:
I(...) is usually called the input
K(…) as the kernel, and
F(...) as the feature map of I ( x) for a given K.
Considering a multidimensional discrete domain, we can transform the integral into the following summation
For 2D digital images, we can rewrite this as:
<