Successive Refinement of Side Information using Adaptive Search Area for Long Duration GOPs in Distributed Video Coding
Résumé
In distributed video coding, the reference frames are used to generate a side information at the decoder in order to
decode the Wyner-Ziv frame. The quality of side information has a strong impact on the coding efficiency of distributed video
coding. The estimation of the side information becomes less accurate when the temporal distance between the neighboring
reference frames increases or when the sequence contains fast motion. In this paper, we propose a new method based on
successive refinement of the side information by adapting the motion search area after decoding the first DCT band. More
specifically, different search areas are initially set, according to the temporal distance between the neighboring reference frames.
Furthermore, the size of the search area is adapted to the motion content after decoding the first DCT band by using the
partially decoded Wyner-Ziv frame. This adapted search area is used in order to refine the side information after decoding
each remaining DCT band. The experimental results show that the proposed technique allows an improvement in rate distortion
performance that can reach 0.7 dB for a GOP size of 8, compared to the method where a constant search area is used, and a
significant gain up to 3.23 dB, with respect to DISCOVER codec.
The improvement in the quality of the final side information
reaches 5.6 dB for some frames, compared to constant search
area. Moreover, the proposed method reduces significantly the
decoding time for all test sequences.