Capturing light field textures for video coding

There is a significant amount of redundancy between video frames or images that can be explained by considering these observations as samples of a light field function. By using a compact depth- augmented representation for such a light field function, it may even be possible to tie-together inter-frame dependencies in a more meaningful way than conventional 2-D intensity based motion compensation methods. We propose a depth-augmented layered orthographic light field representation and show how it may be constructed from actual data at a basic level as the solution to an over-determined linear inverse problem. We finally demonstrate the potential utility of such information in video coding with a compression example when this light field side information is given as a simple texture map.