Texture-adaptive hole-filling algorithm in raster-order for three-dimensional video applications

The technique of image rendering plays a vital role in many 3D applications, such as free view-point television, 2D-to-3D video conversion, etc. Depth image based rendering (DIBR) is considered as an effective way for image rendering. One challenge of DIBR is how to deal with the holes appeared in the procedure of rendering. In this paper, we propose a texture-adaptive hole-filling algorithm for post-processing of rendered image on 3D video applications. Especially, the raster-order is fully considered for hardware design and flexible interaction. The algorithm first determines the type of holes, and then fills the missed pixels in raster-order. The filling method depends on the hole types and texture gradient of neighbors with simple data operation, which benefits for fast processing. Experimental results show that the proposed algorithm outperforms other benchmarks on perceptual quality with considerable savings of computational cost.