Improved spatial prediction for 3D holoscopic image and video coding

Abstract

Holoscopic imaging, also known as integral imaging, is promising to change the market for 3D television since it provides a solution for glassless 3D. This paper starts by making a brief presentation of the general concepts behind holoscopic imaging, with a special emphasis on the spatial correlations that are present in this type of content, which are caused by the micro-lens array used both for acquisition and display. Behind this micro-lens array, many micro-images with a high cross-correlation between them are formed; in these micro-images only one pixel is viewable from a given observation point. This high cross-correlation can be seen as a form of self-similarity within the holoscopic image and it can effectively be exploited for coding. Therefore, in order to explore this self-similarity of holoscopic images, a novel scheme for spatial prediction is proposed in this paper. The proposed scheme can be used for coding both still images and intra-frames in video. Experimental results based on a modified H.264/AVC video codec that can handle 3D holoscopic images and video are presented and clearly show the advantages of using this approach.