The previous image convolution
, this one explains the pyramid
1. Image Pyramid
An image pyramid is a collection of images - all derived from the same original image - obtained by successive downsampling until a certain termination condition is reached.
There are two types of image pyramids that frequently appear in the literature and applications:
Gaussian pyramid: used for downsampling
Laplacian pyramid: used to reconstruct upper-level unsampled images from pyramidal lower-level images
In this document we will use a Gaussian pyramid.
2. Gaussian Pyramid
Think of an image where the pyramid is layer by layer. The higher the layer, the smaller the image.
Each level is numbered from bottom to top, level (i+1) (denoted as G_{i+1} size is smaller than level i (G_{i})).
To obtain a pyramid image with level (i+1), we use the following method:
Convolve G_{i} with a Gaussian kernel:
put all even rows and columns