Research On Key Technologies Of Distributed Coding

Some emerging applications, e.g. sensor network, multiview video surveillance system and hyperspectral remote sensing, require low complexity encoding because the computational power and the memory are limited at the encoder. The conventional image and video coding methods can not meet such demand.The distributed source coding technique is a fire-new coding paradigm with promising performance. The theoretical basis is Slepian-Wolf theory and Wyner-Ziv theory, which state that with two correlated sources encoded separately but decoded jointly one can achieve the same performance as with the two sources encoded and decoded jointly. This technique has many advantages, including the ability to move the computational burden from the encoder to the decoder, and becomes a very hot research area. Therefore, research on practical distributed source coding schemes is of both theoretical significance and practical value.This thesis experimentally analyzes the efficiency of the distributed source coding of ideal sources. And on this basis, the thesis investigates a number of key techniques of the distributed source coding of some pratical sources such as images and videos. The main contributions of this thesis are as follows:1. This thesis analyzes the efficiency of the distributed source coding of ideal sources.In this thesis, Slepian-Wolf coding and Wyner-Ziv coding are performed to compress the ideal sources including the binary source, the Gaussian source and the Laplacian source. Moreover, the compression performance of the distributed source coding and the entropy coding is compared to the theoretical bound. Also the thesis studies the joint source-channel coding scheme based on distributed source coding. Via lots of experiment results, the efficiency of the distributed source coding of ideal sources is studied in quantity.2. This thesis proposes a new lossless distributed hyperspectral image coding scheme.The proposed scheme moves the correlation exploitation operations to the decoder in order to design a lightweight encoder. At the encoder, sub-sampled images/bands are successively encoded and transmitted. At the decoder, side information is generated with the knowledge of decoded sub-sampled images and other previously decoded bands. Reference bands are adaptively selected, and sliding window prediction or K nearest neighbor prediction is performed to exploit the spatial and spectral correlation. With the characteristics of low-complexity encoding and high compression efficiency, the proposed scheme is very suitable for onboard applications.3. This thesis proposes a distributed video coding scheme based on progressive encoding and decoding.At the encoder, each Wyner-Ziv frame is sub-sampled using a block-based sub-sampling algorithm, and resulting sub-samples are successively encoded and transmitted. For the latter samples, adaptive intra prediction is employed with no intra mode bits transmitted. At the decoder, the current Wyner-Ziv frame is progressively decoded, and thus the motion estimation accuracy is iteratively improved. The decoder has partial access to the current Wyner-Ziv frame without receiving any extra information from the encoder, and thus the quality of the side information is improved to achieve better compression performance.A number of simulations have been carried out to demonstrate that the proposed two practical distributed source coding schemes are able to achieve high compression efficiency while providing a very low-complexity encoder.