Densely-sampled light field reconstruction

Abstract

In this chapter, we motivate the use of densely-sampled light fields as the representation which can bring the required density of light rays for the correct recreation of 3D visual cues such as focus and continuous parallax and can serve as an intermediary between light field sensing and light field display. We consider the problem of reconstructing such a representation from few camera views and approach it in a sparsification framework. More specifically, we demonstrate that the light field is well structured in the set of so-called epipolar images and can be sparsely represented by a dictionary of directional and multi-scale atoms called shearlets. We present the corresponding regularization method, along with its main algorithm and speed-accelerating modifications. Finally, we illustrate its applicability for the cases of holographic stereograms and light field compression.