A new framework for layered depth video coding

Abstract

Layered Depth Video (LDV) is recognized as a promising 3D video data representation format in the form of multiple layer texture-plus-depth video from a reference viewpoint. Due to the uncertainty of effective pixel locations in the occlusion layers and the number of total layers, existing video coding standards cannot be directly used for LDV coding. A new framework for LDV coding is presented in this paper. A novel polynomial approximation based prediction tool is proposed for coding of border macroblocks, which contain both effective and blank pixels in the occlusion layers. Precise borders are reconstructed on the decoder side based on analysis of depth discontinuity, which is well preserved by allocating more bits in encoding. Comparing with the state-of-the-art, the proposed method demonstrates improved rate-distortion performance. © 2013 Springer-Verlag.