Convolutions and Filters
Last updated
Last updated
To better understand all the sections which follow, Let's understand some of the jargon present within this space; A Convolution is a function which takes in an image and an Area let's call this N
N refers to what is known as a Kernel size in context of CV we perform modifications to the pixels across an area or the Kernel, think of it as a stencil -- you place it on a 2D list of pixels and do some calculation on whatever pixels are contained within this stencil, this would mean that a kernel of dimensions 3x3 would encapsulate the current pixel, and eight other neighboring pixels. Rinse and repeat for every pixel within the image and you've got yourself a Convolution. Here is a visualization of what a kernel is.
Note I will use pixels and array of integers interchangeably, when I say pixel it always implies it encapsulates three integers for RGB (for most cases)
The terms filter and mask can also be used interchangeably with convolution functions. You may see me do this in the modules that follow.
