Hole filling for view synthesis

Depth-image-based depth rendering (DIBR) technique is recognized as a promising tool for supporting advanced 3D video services required in multi-view video (MVV) systems. However, an inherent problem with DIBR is to deal with the newly exposed areas that appear in synthesized views. This occurs when parts of the scene are not visible in every viewpoint, leaving blanks spots, called disocclusions. These disocclusions may grow larger as the distance between cameras increases. This chapter addresses the disocclusion problem in two manners: (1) the preprocessing of the depth data, and (2) the image inpainting of the synthesizing view. To deal with small disocclusions, a hole filling strategy is designed by preprocessing the depth video before DIBR, while for larger disocclusions an inpainting approach is proposed to retrieve the missing pixels by leveraging the given depth information.