Light field registration: A multi-view geometry approach

Abstract

Through capturing spatial and angular radiance distribution, light field cameras bring new capabilities that are not possible with conventional imaging systems. So far in the light field imaging literature, the focus has been on the theory and applications of single light field capture. The natural extension of single light field processing is the processing of light field sequences, which is likely to bring in new applications, capabilities, and enhancements. One of the first challenges of light field video processing is the registration of light field data. In this paper, we investigate light field registration through multi-view projective geometry. We utilize the regularity of the spatial and angular sampling in the light field data, and extend the methods developed for stereo vision systems to such a regular data. By merging multiple light fields, we demonstrate the extension of synthetic aperture, which results in more accurate depth estimation compared to that obtained by a single light field.