A Population Adaptive Differential Evolution Strategy to Light Configuration Optimization of Photometric Stereo

摘要

Differential Evolution is an optimization technique that has been successfully employed in various applications. In this paper we propose a novel Population Adaptive Differential Evolution strategy to the problem of generating an optimal light configuration for photometric stereo. For 'n' lights, any 2π of orthogonal light directions minimizes the uncertainty in scaled normal computation. The assumption is that the camera noise is additive and normally distributed. Uncertainty is defined as the expectation of squared distance of scaled normal to the ground truth. This metric is optimized with respect to the illumination angles at constant slant angle. Superiority of the new method is demonstrated by comparing it with sensitivity analysis and classical DE.