Scalable Single and Multiple Description Scalar Quantization

Scalable representation of a source (e.g., image/video/3D mesh) enables decoding of the encoded bit-stream on a variety of end-user terminals with varying display, storage and processing capabilities. Furthermore, it allows for source communication via channels with different transmission bandwidths, as the source rate can be easily adapted to match the available channel bandwidth. From a different perspective, error-resilience against channel losses is also very important when transmitting scalable source streams over lossy transmission channels. Driven by aforementioned requirements of scalable representation and error-resilience, this thesis focuses on the analysis and design of scalable single and multiple description scalar quantizers. The first part of this dissertation deals with the design of scalable wavelet-based semi-regular 3D mesh compression systems, which provide superior compression performance when compared to the existing mesh coding methods. We point out that, in general, existing methods employ coding techniques which were previously used for wavelet-based scalable coding of images. In principle, image and mesh data exhibit different statistical properties. In this sense, our design methodology thoroughly analyzes different modules of the coding system in order to develop appropriate design choices for efficient compression of semi-regular meshes. In particular, a Laplacian mixture (LM) model is proposed to closely approximate the distribution of the mesh wavelet coefficients. The distortion-rate (D-R) function of the LM model is analytically computed in order to identify the optimal embedded dead-zone quantizer to be used in wavelet-based coding of semi-regular meshes. Following an information-theoretic analysis of the statistical dependencies between wavelet coefficients we conclude that, for meshes, intraband and composite dependencies are far stronger than the commonly employed interband dependencies. Based on our analysis, we propose intraband and composite mesh codecs which give state-of-the-art compression performance. The proposed codecs provide both resolution and quality scalability. This lies in contrast to the existing zero-tree based interband mesh coding techniques, which only support quality scalable decoding of the compressed mesh. The second part of the dissertation relates to the design of scalable multiple description quantizers, in order to provide source scalability and error-resilience in a single coding framework. In the literature, such a joint framework is referred to as the scalable multiple description coding (SMDC). In this context, a generic symmetric scalable multiple description quantizer (SSMDSQ) is proposed which generates perfectly balanced source descriptions. Compared to existing designs, it is shown that the proposed quantizer constructions exhibit superior D-R performance in both high and medium-to-low redundancy regimes. Moreover, an innovative extension of the Lloyd-Max algorithm is introduced in order to optimize scalable multiple description quantizers. Anchored in the designed SSMDSQs, an SMDC framework is established to realize packet-based transmission over erasure channels. In this framework, transmission strategies are determined for scenarios wherein the average packet loss rate over the transmission link is (a) unknown and (b) can be estimated at the encoder. Experimental results for generalized Gaussian (GG) and image sources confirm that, compared to contemporary schemes, the designed quantizer constructions (with or without optimization) account for a significant average gain in the signal-to-noise-ratio (SNR) for a wide range of packet loss rates.