Towards Practical Wyner-Ziv Video Coding Solutions.
Martínez Martínez, José Luis
MetadataShow full item record
Multimedia is increasingly becoming a utility rather than mere entertainment. Its influence has diversified in context and deepened to a level of indispensability. Low cost video communications is one area that has attracted significant amount of consumer focus and also benefited from considerable research effort for development. Low cost video encoders enable the use of video encoding in low cost devices such as toys, low-power surveillance, wireless cameras, mobile phones and disposable video cameras; all these devices can be formed by a low cost video camera which can be supported with reduced complexity encoding algorithms. However, it is noted that there is still a notable technological gap between an ideal platform and the available solutions, mainly in the form of the encoder and decoder complexity balance and the associated design costs. Distributed Video Coding (DVC) is a new coding paradigm based on two major Information Theory results: the Slepian-Wolf (1973) and Wyner-Ziv Theorems (1976). This new video coding paradigm allows exploiting the source statistic, partially or totally, at the decoder only. A particular case of DVC is the Wyner-Ziv video coding. In this scenario, two correlated sources are independently encoded using separated encoders and the bit streams associated to each source are jointly decoded exploiting the correlation between them. Although the distributed coding study dates back to the 1970s, efforts towards developing practical solutions of Wyner-Ziv video coding are more recent. Emerging applications depicted above with encoding requirements quite different from those targeted by the MPEG-x and H.26x video coding standards have stimulated such efforts. In the MPEG-x and H.26x standards, the correlation between temporally adjacent frames is exploited through a complex motion estimation task which leads to a high complexity encoder. Since the correlation between temporally adjacent frames in Wyner-Ziv video coding is performed only at the decoder, the encoder can typically present a low complexity. This thesis proposes a solution that is well suited for this low cost video communications scenario, and fits many other potential applications too, using Wyner-Ziv video coding paradigm. The research scope of this thesis covers a number of design aspects of Wyner-Ziv coding, making significant contributions towards building a practical video codec in view of commercial developments and standardization. This includes designing novel coding frameworks, eliminating the requirements that deal Wyner-Ziv video coding impractical, enhancing the existing structures, and designing enhanced algorithms for individual functional elements within the codec framework, affecting both encoder and decoder. More specifically, design of Transform Domain Turbo Trellis Coded Modulation (TD TTCM) based Wyner-Ziv codec, feedback free Wyner-Ziv architecture and a framework where the new Wyner-Ziv and H.264 can operate together supporting the advantages of both paradigms labeled as lower encoding algorithm (Wyner-Ziv encoders) and lower complexity decoding algorithm (H.264 decoders) TD TTCM architecture is proposed as an alternative to the more common turbo coding based codec implementation to improve the coding efficiency. Moreover, the proposed architecture works in transform domain which improves the performance of the pixel based by exploiting the human visual system. The combination of transform domain jointly with TTCM codes have never been proposed in the literature before and it outperforms some of the related architectures of the state of the art. The feedback free architecture, one of the pioneering unidirectional frameworks is presented in this thesis, which enables the suppression of practically hindering reverse feedback channel. This kind of architectures brings us practical and development Wyner-Ziv video coding solutions. Finally, a novel approach for Wyner-Ziv to H.264 video transcoder is also proposed; at this point, it is proposed a framework where traditional video codecs such as H.264 and Wyner-Ziv video codec can operate together. The improvements done over the reference video transcoder make that the time delay and complexity constraints of the end user devices are not any restrictions. The efficiency and effectiveness of all proposed algorithms are verified through simulations and comparing the results with the state of the art. Moreover, this thesis have also as objectives i) perform a revision of the state-of-the-art on distributed video coding as well as identify the more prominent approaches regarding to this thesis approaches and, ii) develop and implement a owner Wyner-Ziv video coding solution as reference software where can implement the different approaches due to the does no exist any reference software or standardization.