INTRA-WZQUANTIZATION MISMATCH IN DISTRIBUTED VIDEO CODING J¨ urgen Slowack, Jozef ˇ Skorupa, Peter Lambert, and Rik Van de Walle Ghent University – IBBT ELIS Dept., Multimedia Lab Gaston Crommenlaan 8 bus 201 B-9050 Ledeberg-Ghent, Belgium Nikos Deligiannis, Adrian Munteanu Vrije Universiteit Brussel – IBBT ETRO Department, Pleinlaan 1 B-1050 Brussels, Belgium ABSTRACT During the past decade, Distributed Video Coding (DVC) has emerged as a new video coding paradigm, shifting the complexity from the encoder- to the decoder-side. This pa- per addresses a problem of current DVC architectures that has not been studied in the literature so far, that is, the mis- match between the intra and Wyner-Ziv (WZ) quantization processes. Due to this mismatch, WZ rate is spent even for spatial regions that are accurately approximated by the side-information. As a solution, this paper proposes side- information generation using selective unidirectional motion compensation from temporally adjacent WZ frames. Ex- perimental results show that the proposed approach yields promising WZ rate gains of up to 7% relative to the conven- tional method. Index Terms— Distributed Video Coding, Wyner-Ziv coding, quantization 1. INTRODUCTION Distributed Video Coding (DVC) has received a lot of re- search attention in the past decade, being a radically new way of performing video compression. Instead of exploiting the similarities between frames at the encoder, in DVC these sim- ilarities are exploited at the decoder. This leads to a reversed complexity distribution, with a simple encoder but complex decoder as opposed to a complex encoder but simple decoder in conventional systems, e.g., H.264/AVC. Most of the current DVC systems described in the lit- erature are based on the architecture initially proposed by Aaron et al. [1]. In such systems, the frame sequence is parti- tioned into key frames and Wyner-Ziv (WZ) frames. The key frames are coded without using other frames as references (e.g., through H.264/AVC intra coding). For each WZ frame, the decoder generates a prediction – called side information – using already decoded frames as references. Next, errors in the side information are corrected via error correcting infor- mation sent by the encoder (such as turbo or LDPC codes). Many researchers have made contributions focusing on different aspects of DVC. In particular, in the context of side information generation, contributions include spatial smooth- ing and subpixel refinement [2], mesh-based techniques [3], temporal non-uniform interpolation [4], and partial decoding and refinement [5]. Other research directions followed in the context of DVC focus on estimating the correlation between the side informa- tion (available at the decoder) and the original (available at the encoder) [6, 7], and on eliminating the feedback chan- nel [8, 9]. In this paper we identify an additional problem in DVC that has not been studied in the literature so far, i.e., the mis- match between intra and WZ quantization. A detailed de- scription of this problem is provided in Section 2. Next, in Section 3, we provide a solution to this problem through se- lective unidirectional motion compensation. The results in Section 4 indicate bit rate improvements up to 7% for the WZ frames. Finally, conclusions and a discussion of future work are provided in Section 5. 2. PROBLEM ANALYSIS The quantization process in the context of intra coding is quite different from its counterpart in the context of WZ coding. First of all, the nature and range of the values that are pro- vided as inputs to the quantizer are different. For example, in H.264/AVC intra coding the encoder first generates a spatial prediction of the current block of pixels, and subsequently transforms and quantizes the residual between the current block and its prediction. In WZ coding, the procedure is in general much simpler, i.e., no prediction is generated at the encoder and the pixel values are transformed and quantized directly. Apart from the fact that the nature of the input val- ues is different between the intra and the WZ quantizer, both quantizers typically have different bin widths, and different characteristics (e.g., having a deadzone or not). At the de- coder, the reverse quantization steps are different as well; for example, optimal minimum mean square error reconstruction