This MATLAB function is used to reconstruct high-resolution images from a set of low-resolution images from the same scene by iteratively back projecting and updating HR image with a kernel consisting of operators of global pixel displacements, uniform gaussian blurring and under-sampling HR image.
This program assumes all LR images have identical size and there are same numbers of LR images along horizontal and vertical directions resulting in same resolution enhancement along either orientation.
input 1: a cell array that contains all LR images each of which is represented by grayscale values of all pixels in it as a matrix.
input 2: a m-by-2 matrix where m is the number of total LR images, containing sub-pixel global ranslations for all LR images in which the first column represents horizontal displacement while second column represents vertical displacement. The values for all elements must be less than 1 denoting sub-pixel translation.
input 3: the sigma value, or standard deviation for the gaussian filter applied to all pixels in HR image. Only the surrounding pixels closest to the one applied with this filter are included in calculation. This parameter should be a positive scalar.
input 4: initial guess of HR image in the form of a matrix. It must exactly match the required resolution of HR image in specific applications.