Research On Key Technologies Of Distributed Video Coding

During the past few years, video coding technologies have achieved great progress. The advanced video compression standards, for example H.264/MPEG4 AVC, make it possible to store and transmit huge amount of video data efficiently. Effective video transmission facilitates real-time communications of multimedia information and promotes video telephone, video conference, video on demand and some other interactive real-time multimedia applications. However, the coding performance is affected by the network conditions. On the one hand, the diversity of media and protocols leads to the variation of transmission rate. Besides, the burden of the server or the router sometimes limits the available bandwidth. On the other hand, due to flexible configuration and low cost, wireless network is widely used. As we know, the weather, the pollution and other environmental factors inevitably influence the wireless network and introduce some noises.Due to the complicate, dynamic and changeable network, the current video coding technologies are quite difficult to satisfy the requirements of high definition, stability and fluency. How to overcome the channel noises and the unstable transmission rate so that the promising video content can be provided at the client side becomes the main challenge of video coding technologies. This dissertation focuses on distributed video coding with the purpose of developing the efficient, robust and multiple bit-rate video coding technologies under various network situations. Inspired by that, some particular issues on the design of distributed video coding scheme, the correlation modeling, the syndrome based distributed coding, the optimization of coding modes and bit rates, the bit-stream switching and the rate allocation for switching streams are discussed and lucubrated. The detailed descriptions for these techniques are given as follows.1. This dissertation designs and develops an error-resilient distributed video coding scheme, which is founded on the correlation model of video signal, especially for wireless video transmission. Mode decision in H.264/AVC is first utilized to classify the video blocks into Inter,Intra,Skipped modes based on the temporal correlation. Inter coding, which is the most sensitive to channel noises, is replaced by the syndrome-based distributed coding. The blocks belonging to the distributed coding mode are further classified according to the temporal correlation and the optimized coding parameters are then assigned to each class. Different from the entropy coding of predictive residue in Inter coding, the distributed coding encodes the original video data separately. The distributed decoder generates the prediction (namely side information) with the previously decoded frame and jointly decodes the bit-stream together with the side information in order to exploite the mutual correlation. Due to the independence of the bit-stream on previous frame, better error robustness is achieved in the proposed distributed video coding scheme. Extensive simulations of video transmission over wireless networks show that the proposed video coding scheme is much more robust against channel errors than H.264/AVC.2. An optimized mode decision between syndrome coding and entropy coding is proposed based on the correlation model of various elements to be encoded in distributed coding mode. For each element to be encoded, a selection between syndrome coding and entropy coding should be made according to the correlation so that the optimal rate-distortion performance can be achieved. Thanks to the proposed algorithm, the proposed distributed coding scheme is designed with sufficient consideration of the correlation inherent in video data and the coding performance is enhanced significantly.3. This dissertation proposes a bit-stream switching scheme based on distributed coding for multiple bit-rate (MBR) video streaming over networks. Identical video content is encoded into a set of normal streams, which are generated by conventional hybrid video coding with multiple bit rates, so that the streaming can dynamically switch among these normal streams according to the available bandwidth. At encoder side, the distributed codec generates a switching stream by compressing the reconstructed frames of a certain normal stream that will be switched to, no matter which normal stream it switches from. At decoder side, for switching to the same frame, the same switching stream is used to reconstruct the switching-to frame by taking the switching-from frame as the side information. The number of required bits depends on the inherent mutual correlation between two frames switching to and from. Since the distributed switching streams are generated independently of the normal switching-from streams, given N normal streams that can switch from any one to another, the proposed scheme reduces the number of switching streams from N(N-1) to N. Thus, the complexity of generating and storing switching streams is reduced significantly. Furthermore, switching streams do not deteriorate the coding efficiency of normal streams when no switching occurs.4. A rate-allocation algorithm is proposed for distributed switching stream. A big problem here, similar to requesting bits in distributed video coding, is how many bits should be transmitted when a switching happens since the streaming scenario does not tolerate too much extra delay caused by the requests back and forth. A Laplacian model, which is proved in the simplified case, is proposed to characterize the correlation between switching-to and switching-from frames. It can be used to accurately estimate the bit rate of switching stream at the server side. Both the feedback channel and the delay are removed effectively.