Research On Side Information Generation And Evaluation Method In Distributed Coding

With the rapid development of wireless network and information technology, the application of mobile video terminal is becoming more and more popular in multimedia communication. The traditional coding standards, such as MPEG-x and H.26 x, need quantifying, coding, motion estimation, compensation and other operations in the coding end. As a result, the encoder needs a strong performance to take the amount of calculation. Thus, coding standard of wireless video equipment which has limited calculation capacity, such as video surveillance, mobile video telephone, is no longer suitable. However, distributed video coding shifts the complex operation from the encoder to the decoder. At the same time, because of the technique of channel coding in distributed video coding, the error resilience performance of the system is greatly improved. So it is very suitable for the field of wireless video communication.There are many factors that affect the performance of distributed coding system,such as the quantization error, reconstruction. One of the most important factor is the prediction accuracy for the current video frame. The more accurate generated by side information, the amount of data needed to be transmitted in encoding end is smaller.This will improve compression performance of the system. Therefore, how to predict the edge information quickly and accurately is the focus research on distributed video coding system.Firstly, the application background and the research situation of distributed video coding is introduced. The main structure and the basic principle of distributed video coding are introduced. Secondly, the methods of generating the side information are summarized. The key steps in the side information generation, such as block matching and motion vector estimation and so on, are analyzed. Finally, based on the classic side information generation algorithm MCTI, an improved scheme is proposed. In this method, the frames before and behind the input key frame are segmented in the frequency domain. The low-frequency part of the image is maintained according to the human visual characteristics(HVS). Then, the matching block is classified and the larger movement range ones are subdivided into smaller search block. For each search block, the three step search algorithm is used to block matching. Then the quadratic function is used to estimate the movement trajectory. At last, the objective quality assessment of SSIM and VIF is adopted to evaluate the generated side information.The simulation results show that, compared with the traditional edge information of simulation results, the proposed algorithm improves the interpolation frame overlap block effect and side information generation quality. So the result of the image quality evaluation is more consistent with human subjective feelings.