Distributed Video Coding: Principles and Evaluation of Wavelet-Based Schemes

Abstract

3. a presentation of some original work by the authors on the design and evaluation of wavelet-based video coding;4. an overview of applications of the DVC paradigm to other aspects besides video coding. In particular, we will present results relative to robust transmission of video using an auxiliary DVC stream;5. conclusions, advantages and limitations of the DVC paradigm.In particular, we will introduce an original distributed video coder based on processing the wavelet transform with a modulo-reduction function. The reduced wavelet coefficients are compressed with a wavelet coder. At the receiver side, the statistical properties between similar frames are used to recover the compressed frame. A second contribution is the analysis and the comparison of DVC schemes in two different scenarios: in the first scenario the information frames are separated from the other frames, and they are compressed following the original framework considered for Wyner-Ziv coding. In the second scenario, all the frames are available at the encoder making this an interesting proposal for the design of a low-complexity video coder, with no motion compensation, where the information frames are coded using DSC techniques. The whole set of experiments show that the proposed schemes -that do not use any feedback channel -have good performance when compared to similar asymmetric video compression schemes considered in the literature. Finally, we will consider an original error-resilient scheme that employs distributed video coding tools. A bitstream, produced by any standard motion-compensated predictive codec, is sent over an error-prone channel. A Wyner-Ziv encoded auxiliary bitstream is sent as redundant information to serve as a forward error correction code. We propose the use of an extended version of the Recursive Optimal per-Pixel Estimate (ROPE) algorithm to establish how many parity bits should be sent to the decoder in order to correct the decoded and concealed frames. At the decoder side, error concealed reconstructed frames and parity bits are used by the Wyner-Ziv decoder, and each corrected frame is used as a reference by future frames, thus reducing drift. Tests with video sequences and realistic loss patterns are reported. Experimental results show that the proposed scheme performs well when compared to other schemes that use Forward Error Correcting (FEC) codes or the H.264 standard intra-macroblock refresh procedure. Foundation of Distributed Source CodingIn this section we will introduce in detail two major results provided by Information Theory that prove that, under the DSC paradigm, it is still possible to achieve or to approach, in total generality, the optimal performance of a joint coder: the Slepian-Wolf theorem and the Wyner-Ziv theorem. We first introduce the main ideas behind distributed source coding by means of examples. A glimpse at Distributed Source CodingConsider the case of N distributed sensors that communicate with a data collection center using a radio link. Data recorded by each sensor, at each t...