Contents I. Introduction
Recently, interest has been increasing in a new image capturing device called a Light Field Camera (LFC), because it provides attractive functionality such as changing the focus after the shot [1]. Since the LFC generally captures four dimensional light field information by a single image sensor with a micro-lens array and a color filter, requirement for resolution of the sensor is much higher than the conventional cameras. Accordingly, the LFC produces a large amount of raw data for every shot. This fact yields great demands for high efficient compression techniques to make the device more usable. In [2], lossy compression algorithms for a stack of two-dimensional images, which respectively correspond to different viewpoints and are synthesized from the raw data through several processing steps such as demosaicing, devignetting, brightness clipping and color space conversion, are discussed. However, such the processing steps introduce non-negligible round-off errors as well as resampling artifacts, and might deteriorate quality of the final images generated from the LFC data even if there is no degradation in a compression process.
