Compression of residual layers of layered depth video using hierarchical block truncation coding

Layered depth video (LDV) is a method of compressing and rendering multi-view video sequences. The LDV format consists of multiple layers of information of which the first layer is made up of the reference camera's color and depth components and the remaining layers consist of residual color and depth information. The residual layers of the LDV consist of sparse and irregularly shaped objects which pose a problem during compression. In this paper, we propose a scheme for compressing the residual color and depth layers of the LDV using hierarchical block truncation coding (HBTC). In this paper, we first propose a two-mode BTC scheme that can take advantage of uniform blocks of pixels and effectively reduce the overall bitrate. Next, we combine the proposed two-mode BTC scheme with traditional quadtree decomposition to create a hierarchical block truncation coding scheme which we prove is suitable for encoding sparse irregularly shaped objects in an image. In order to show the performance of the proposed compression scheme, we compare the rate-distortion performance with that of H.264/AVC in the Intra mode. The results show that the proposed compression scheme can give upto 0.8-1.0 dB improvement in the synthesized view quality at comparable bitrates.