2M-dimensional Vector Orthogonal Transform Nuclear Mtrix And Application In Multi-view Video Coding

Matrix and Application in Multi-view Video Coding With the growing needs of human towards material and cultural, science and technology develop rapidly nowadays. Telecommunication networks, multimedia and other technologies promote the development undoubtedly. Such as, 3DTV, HDTV and FTV, these technologies change the original 2-dimensional (2D) video can not give people the enjoyment of the real situation of three-dimensional. Multi-view video (MVV) is multiple video sequences that integrated different viewpoints of the same three-dimensional (3D) scene. The slightly different between angles give people the true immersive experience. Meanwhile, another problem must be considered, that is, the data of the MVV is several times more than ordinary video, in order to ensure the continuous when the viewer moves to a certain angle to watch the image, people need more than 100 video points. Such huge amount of data, ordinary video coding technology could not do efficiently. Thus we need a new mathematical model to handle large amounts urgently.So our laboratory proposed multi-dimensional vector matrix theory (MDVM), extend the matrix tool of mathematical from 2D to MD. Based on the theory, this paper proposed the 2M-dimensional (2MD) vector orthogonal transform nuclear matrix that based on DCT and basic algorithm, and verify the orthogonality and concentration of energy. We introduce the concept of 2MD vector orthogonal transform, apply the theory to multi-view video coding (MVC), express video data in MD mathematical model, and transform and inverse transform. The MVV that we choose is 2- viewpoint, yuv format, earlier 8 frames. We divide the Y, U, V component into cubes separately, combine the two views` data, then transform. Because ordinary video data is 2D, this paper introduce the view dimensional, the MVV data is 4-dimensional (4D), 4D model: f (?). Then devote the dimensional into two groups, denote as f uNuu 1u×uu Nuu u2u u×u uNuu 3v×uNuu4v , Then transform using the 6-dimensional vector orthogonal transform nuclear matrix (6DVM) proposed in this paper and 2D DCT operator. In the experiment, M is 3. The analysis of experimental results we verify the method in our paper have good energy concentration properties. With the evaluation criteria mentioned in section 4.1, satisfactory results are obtained. We realize the algorithm in this paper by use of the Windows system and the Visual C++6.0 software. The even-dimensional generalized, that is 2M-dimensional vector orthogonal transform nuclear matrix (2MDMV), is a effective tool to process multi-dimensional data, can be applied in many ways. We apply the theory to MVC innovatively, unlike the previous method of dealing with MVV, and there are many additional area need to improve. For example, the quantization towards to the coefficients that transformed, scanning methods, effective methods of entropy coding, and the evaluation standard for MVV, all are we need to study further. This work is supported by the National Science Foundation of China under Grant 60702036 and 60911130128.