Image Deconvolution using Fast Fourier Transforms
This is a project I worked on in Graduate Level Mathematical Modeling Case Studies. Due to overlap with our competition, we only had one week to work on this project.
In this project we were asked to deblur two images that were blurred with unknown bandwidth, then create a method of automatically choosing bandwidth and noise parameters. We used fast Fourier transforms to deblur the images then we visually compared them to determine the ”best” deblurred image. We used two algorithms to automatically choose bandwidth and noise parameters: contrast and entropy. The contrast algorithm worked poorly and often chose more distorted images than the original blurred image. The entropy algorithm always returned the same values for parameters; however, it usually returns an improved image. There are numerous things we can do to improve algorithms if given more time and computing power.