I have a set of image pairs, i.e., {A, A’}, {B, B’}, {C,C’} etc.
Given an image A, for instance, 256*256, are there any sampling mechanisms to generate a small image, i.e., 128*128, that can capture the global contexts of original images.
At the same time, I would like to keep track information of those sampled pixels. For instance, if pixel[I,J] in A is selected, I need to track it because I need to select the same pixel of A’.
In other words, I need to generate a set of smaller image pairs based on these large image pairs. The sampling mechanism, however, is based on A, B, C, etc.