Vignetting Modeling and Correction for a Microlens-based Light Field Camera

Microlens-based light field cameras, which are capable of recording both angular and spatial information of light, are already commercially available as consumer commodities. Intrinsically, the large f-number and the use of a microlens array introduce a more severe vignetting effect than a conventional camera. Proper devignetting is required to reconstruct a high quality light field from the captured 2D raw image. In this paper, a 2D Gaussian kernel is proposed to model the microlens image vignetting and local parameters are estimated by solving a nonlinear optimization problem. We assume this kernel is smoothly varying across the whole image. The global parameters are estimated by polynomial fitting. Our results show that it accurately predicts the vignetting effect. We also demonstrate successful vignetting correction based on our modeling prediction.