A Fundamental Study on Coded Aperture Design for Depth from Projector Defocus Based on Evolutionary Multi-objective Optimization

Abstract

It is a reasonable idea to solve a problem whose property has not been clarified yet using evolutionary multiobjective optimization (EMO) because EMO does not require domain specific knowledge and analysis for obtained pareto solution set may reveal relationships between design variables and objective functions. This paper focuses on designing a coded aperture (CA), one of the hot topics in computer photography; CA is used in a camera or a projector for deblurring, constructing all-focus image, depth estimation and so on. Since little knowledge has been accumulated for designing CA, this paper proposes a CA design method using EMO. Experiments on designing projector aperture have shown relationships between the performance of depth estimation, aperture ratio, and noise robustness.