Research And Optimizing On Some Key Technologies Concerning Multiview Video Based On H.264

Multiview video coding (MVC), which comprises rich3D information of a scene, can provide splendid visual perception compared with the traditional media. MVC is based on the video coding standard H.264and it introduces advanced technologies such as disparity estimation, hierarchical B picture and so on. All of these methods have obtained the good performance on encoding. However, the computational complexity of MVC encoders is tremendously increased. Fast algorithms are very desirable for the practical applications of MVC.To reduce the complexity, the structure of the encoding is analysed and experiments are have done in JMVC(Joint Multiview Video Coding), which is the platform of the multiview video coding. Some optimization algorithms,which include early termination, the selection of inter-view prediction and the optimizing of disparity/motion estimation,are implemented on JMVC. The fast algorithms are proposed for MVC, as follow.1. Each macroblock is checked whether the rate-distortion(RD) cost of the Skip mode is below a predefined ’low’ threshold or above the ’high’ threshold, to make an early termination or narrow the candidate modes. The mode of Inter8*8is also optimized.2. By utilizing Inter16*16mode prediction results, the disparity estimation of other Inter modes is guided to bypass the unnecessary inter-view prediction. Meanwhile, joint the proposed method in part1, a fast mode selection algorithm of MVC is proposed. As compared to the full mode decision in MVC, the experiment results show that the proposed algorithm can reduce up to78.13%on average encoding time.3. The disparity/motion vectors correlation of Inter modes are analyzed and the prediction disparity/motion vectors are compared in the collection. Optimizing has been done on the selection of the search center among the candidates. The proposed method is implemented on TZsearch of JMVC. The experiment results show the efficiency of the proposed algorithm.