Research On Extrapolating Structure Based Distributed Video Coding

Distributed Video Coding(DVC) is a new paradigm for video compression. Unlike the traditional video coding with complex encoder and simple decoder, DVC introduces independent encoding and the channel coding techniques and exploits the source statistics in decoder. Thus, DVC has the characteristic of simple encoder, complex decoder and good error-resilient performance. DVC is more suitable for wireless video applications where power consumption and computing power are limited.Firstly the fundamental theory and system framework of DVC is introduced. Then the researches are focused on the low delay DVC system under long group of pictures(GOP). The main work and achievements are as follows.1. This paper systematically introduces the system framework and key technologies of distributed video coding. Then interpolation-based side information(SI) generation methods and extrapolation-based side information generation methods are analyzed. The obtained performance results show that extrapolation structure is important on improving the compression performance of low delay DVC with long GOP.2. Based on the different characteristic of interpolation and extrapolation, a novel hybrid side information(HSI) scheme has been proposed. Firstly, the proposed algorithm generates side information by extrapolation method at the decoder; Then the side information is refined utilizing the decoded Wyner-Ziv(WZ) frame information by multi-hypothesis prediction algorithm. Finally, the WZ frame is reconstructed again based on the refined HSI and the decoded frame. Experimental results show that the proposed scheme improves the Rate-Distortion(RD) performance. Meanwhile it maintains low encoding complexity.3. Feedback-free DVC system is suitable for applications without feedback channels, this paper focus on low delay feedback-free DVC with large GOP size. The compression performance of the feedback-free DVC depends not only on the rate allocation accuracy at the encoder, but also on the decoding ability under the condition of rate underestimation at decoder. This paper proposed an unequal protection encoder rate control algorithm(UP_ERC) based on significance of bitplane at encoder, which can estimate the rate accurately. The HSI refinement and iterative WZ decoding procedure are proposed to improve the side information accuracy and reduce the impairment of the rate underestimation. Experimental results show that the proposed algorithms improve the system rate distortion performance and WZ frame quality, while maintaining the advantage of low encoder complexity and low delay.