Scalable multiview video coding using wavelet

Most existing multiview video coding (MVC) techniques are based on the traditional hybrid DCT-based video coding schemes, e.g., MPEG-2, MPEG-4 and H.264. They neither fully exploit the redundancy among different views nor provide an easy way of implementation for scalabilities. In this paper, we propose an MVC scheme based on wavelet, which can provide temporal, spatial, SNR as well as view scalabilities. To the best of our knowledge, wavelets have not been used for MVC in the literature before. In particular, we consider a multiview video as a 2D matrix with 1D along the temporal axis, 1D along the view axis, and each element in the 2D matrix represents a video frame. We first apply 1D wavelet decomposition to the 2D matrix along the temporal axis with motion compensation, and then along the view axis with disparity compensation. After that, 2D spatial wavelet decomposition is applied to each element in the matrix. Finally, all the subbands are encoded by 3D-ESCOT with rate-distortion optimization. Compared with traditional 3D wavelet coding, our proposed wavelet-based MVC scheme applies one more dimensional wavelet transform along the view direction to exploit the redundancy between adjacent views. Some preliminary experiments are carried out using standard multiview video sequences and the efficiency of the proposed system is confirmed.