Theoretical Analysis of Multi-View Camera Arrangement and Light-Field Super-Resolution

Abstract

We analyzed a light-field super-resolution problem in which, with agiven set of multi-view images with a low resolution, the 3-D scene isreconstructed with a higher resolution using super-resolution (SR)reconstruction. The arrangement of the multi-view cameras is importantbecause it determines the quality of the reconstruction. To simplify theanalysis, we considered a situation in which a plane is located at acertain depth and a texture on that plane is super-resolved. Weformulated the SR reconstruction process in the frequency domain, wherethe camera arrangement can be independently expressed as a matrix in theimage formation model. We then evaluated the condition number of thematrix to quantify the quality of the SR reconstruction. We clarifiedthat when the cameras are arranged in a regular grid, there existsingular depths in which the SR reconstruction becomes ill-posed. Wealso determined that this singularity can be avoided if the arrangementis randomly perturbed.