|
The video encoder with low-complexity has received a lot of attention in recent years, where distributed video coding (DVC) and compressive sensing (CS) are proposed as two enabling techniques. On one hand, by exploiting the correlation at the decoder rather than the encoder, DVC shifts the computation burden from the encoder to the decoder. On the other hand, the framework of CS has shown that sparse signals can be recovered from far less samples than that required by the classical Shannon-Nyquist Theorem. In this paper, to utilize advantages of both two techniques, we introduce a sub-sampling enhanced DVC (SSED) framework. Two layers are transmitted in SSED, where the base layer is generated by a standard video coder, while the sub-sampling enhancement layer is generated by CS operation to improve the signal quality in the base layer. When compared with the traditional fully-sampling framework, SSED saves huge sampling and storage costs due to the sub-sampling operation. Furthermore, without compromising the recovered video quality, SSED is shown to enjoy more favorable features, i.e., rate reduction, robustness to channel loss and lower operation complexity, which are verified by simulations. These results demonstrate that SSED is a promising framework in video streaming applications. |
|
Keywords:Compressive sensing; distributed video coding; Slepian-Wolf coding; sub-sampling; Wyner-Ziv coding |
|