Study On The Technology To Improving Coding Efficiency Of MVC System

With the development of video display technology,3D stereo display technology and multi-view display technology has been gradually applied to the field of daily consumption. Multi-view video sequences are captured at the same time by cameras in different position, including the visual information of the different angles in same scene. Therefore, the amount of data in the multi-view video is far more than the ordinary single-view video. The huge amount of data in multi-view video brings a lot of inconvenience to the storage and transmission. Multi-view video sequence contains a large amount of redundant information, if fully utilized the correlation between the inter-frames and inter view, which will greatly improve the efficiency of multi-view video coding, and the multi-view video can be used in more areas. Therefore, how to improve the coding efficiency becomes a key technology in multi-view video coding.At first, the design was established and realized the multi-view video coding system based on discrete wavelet transform in the inter-frame, intra-frame and inter-view. This system could realize the temporal, spatial, views and quality scalable codec. To improve the speed of motion estimation in multi-view video coding system, a fast motion estimation algorithm based on Markov chain model was proposed. This algorithm established a predicted motion vector state set by making full use of the correlation between the motion vectors of frames and views. According to the transition probabilities of Markov chain model, test of the predicted motion vector are performed in advance. Then fast motion estimation was finished by the early elimination conditions. Experimental results demonstrated that there is a significant improvement in the motion estimation using this algorithm for multi-view video coding.Because of the cameras have different positions and parameters; there are inconsistent illumination and color between different views. Through the analysis of a variety of illumination compensation algorithm, an illumination compensation algorithm was proposed for multi-view video coding system based on discrete wavelet transform. Calculate the different value of illumination change during motion estimation and finished illumination compensation during motion compensated temporal filter. This algorithm can significantly improve the coding quality for the views which have a large difference of illumination. In addition, in order to improve the coding efficiency of multi-view video coding further, a color correction algorithm based on support vector machine was proposed. The matching region between the views was quickly determined by using of the correlation between the views, the vector was global disparity. In order to correct the color of each perspective, the algorithm used the matching region as a training set and established the color correspondence between different views through statistical learning by support vector machine. The experimental results demonstrate that the color correction algorithm can be effectively corrected the color inconformity between different views.