Multi-view codec with low-complexity encoding for Distributed Video Coding

摘要

The low-complexity encoding, as fundamental requirement of Distributed Video Coding, relies on performing the bulk of computation at decoder, including tasks as the generation of side information and particularly, inter-camera registration in the case of multi-view systems with complete-overlapped views and free motion of the cameras (e.g., video surveillance). In Ciobanu and Corte-Real (Multimedia Tools Appl 48(3):411-436, 2010) we introduced a codec-independent solution for such tasks at decoder. In this paper, we present a multi-view Wyner-Ziv codec (IWZ) designed for the architecture and scenarios from Ciobanu and Corte-Real (2010) (e.g., free motion of the cameras, no a priori knowledge of the instant camera positions, no feedback channel), based on transform domain (DCT), block-based coset coding. We aimed to achieve a compromise between the low encoder complexity and the rate-distortion performance. A detailed evaluation is presented for comparison with conventional coding (Intra 4×4 and Intra 16×16). Practical results show a better overall performance of the proposed codec at low bitrates.