Joint texture and depth coding using cuboid data compression

Abstract

The latest multiview video coding (MVC) standards such as 3D-HEVC and H.264/MVC normally encodes texture and depth videos separately. Significant amount of ratedistortion performance and computational performance are sacrificed due to separate encoding due to the lack of exploitation of joint information. Obviously, separate encoding also creates synchronization issue for 3D scene formation in the decoder. Moreover, the hierarchical frame referencing architecture in the MVC creates random access frame delay. In this paper we develop an encoder and decoder framework where we can encode texture and depth video jointly by forming and encoding 3D cuboid using high dimensional entropy coding. The results from our experiments show that our proposed framework outperforms the 3D-HEVC in rate-distortion performance and reduces the computational time significantly by reducing random access frame delay. Keywords-video coding; cuboid; McFIS; dynamic background; mutiview video I.