Computational Imaging with a DiffuserCam

A team of the UC Berkely has developed a device similar to a camera called DiffuserCam. The peculiarity of this device is that it does not have a lens. The researchers had replaced the lens with a diffuser, a piece of tape. In this project, we analyse this device and the mathematical models behind this system: computational imaging, compressed sensing, the forward model, the inverse problem, ... In the laboratory, we have built this camera. Here, we detail the calibrating process and the setup that optimizes the results. We implement the recovery using two algorithms: Matching Pursuit and Iterative Hard Thresholding. We discuss booth algorithms step by step and demonstrating the operations involved. Finally, we take some captures to obtain the desired scene and we not only evaluate the results but also we compare them with simulations.