The CEV camera system is a multi-image system using a planar array of micro-lenses inspired from the eyes of insects. This is a next-generation camera system that can overcome the optical limitations of a conventional single-lens camera. CEV images consist of an array of low-resolution images containing different information under certain conditions, and several problems may occur in the process of converting into a single high-resolution image.
In this thesis, we cover not only improving the performance of CEV super-resolution, but also the distance issue according to the geometrical relation of the imaging system. We investigate the numerical instability of the CEV image restoration problem and how the CEV problem varies depending on the geometric relations between objects, lenses, and sensors.
To solve the CEV problems, we introduce effective regularized methods based on Forward-Backward splitting, and perform experiments with the proposed methods. In particular, we suggest a novel method called Deformable Deconvolution to solve the indoor CEV problem, which is difficult to represent with conventional convolution.