| Distributed video coding (DVC) is a new video coding architecture. The theoretical basis of DVC is the distributed source coding theory which includes:(1) the lossless coding of two correlated information sequences from a given bivariate distribution, established by Slepian and Wolf, and (2) the loss coding with side information at the decoder, established by Wyner and Ziv. The combination of the distributed source coding and the video coding yields the distributed video coding and provides many potential benefits, compared with the traditional hybrid video coding: more flexible allocation of computational complexity between the encoder and decoder, simpler encoder with acceptable compression efficiency (compared with the same-level encoders of the hybrid video coding), high robustness against the channel noises.As a result, the DVC has a great development during the past decade.There is a rate allocation challenge of DVC:the bitrate required by each coding unit varies due to the time-varying content of video, however, it is difficult for the encoder of DVC estimates a proper bitrate for each unit since there is no accurate temporal predictor of the coding unit, namely, the side information, at the encoder. Most of the recent practical DVC systems utilize the rate-adaptive coding tools and the feedback channel to provide the solution of the challenge:the encoder transmits a subset of the bits to decoder, and the decoder tries to decode the unit using the bits received; if the decoding fails, the decoder will send a request signal for more coding bits of the current unit via the feedback channel.The feedback-based DVC systems achieve more accurate bit allocation and higher compression efficiency than the same-level feedback-free DVC systems.However, some of drawbacks limit the applications of feedback-based DVC systems:(1)the cost of feedback channel; (2) the delay on the feedback channel will affect both the encoder and decoder; (3) the interaction and synchronization between the encoder and decoder.This dissertation addresses the architecture, the computational complexity of encoder and the compression efficiency of the feedback-free DVC systems. The system is divieded into two related parts:the key frames coding, and the Wyner-Ziv coding. The major contributions of this work are listed below:1,For the part of key frames coding, a fast algorithm with adaptive mode selection for H.264/AVC I slices coding is proposed. Utilizing the Sobel operator to calculate the edge direction and magnitude of the source pictures, the fast algorithm exploits the relationship between the image edge and the optimal prediction mode to reduce the encoder's complexity with negligible loss of compression efficiency.2, For the part of Wyner-Ziv coding, the modeling of the correlation noises between the source information and the side information is presented. Based on this modeling, the coset code is designed to encode the source information. Furthermore, the rate-distortion optimization for the coset code is proposed.3,The statistics of hash collision is analyzed to research the relationship between the optimal hash function and the content of source information. The count of nonzero coefficients of source information is proposed to be one of the statistical conditions. Based on the analysis, two practical algorithms, namely the Adaptive Selection of Hash Functions and the Nonzero Prefix of Coefficients, are proposed to improve the compression efficiency of Wyner-Ziv coding.4,A common case during the practical encoding and decoding of natural video is studied:the co-located block in reference frames can be used as the side information to yield the correct reconstructed information at the decoder. Given the natural video sequences, the probability of this case is analyzed, as well as the sufficient and necessary condition of this case. Then a practical algorithm, namely the Predictor's Pre-Matching in Encoder, is proposed to save the bitrate and improve the compression efficiency.
|
PDF Full Text Request