Multi-View Image Coding in 3-D Space Based on 3-D Reconstruction

 In this paper, we propose a multi-view image coding system that operates directly in 3-D space, and it is based on 3-D scene reconstruction. Unlike existing multi-view image coding schemes, in which the 3-D scene information of the images to be encoded is represented by a mesh model as well as the texture data, we use a 3-D voxel model to represent the 3-D scene information of the considered images and then encode the 3-D voxel model. There are several advantages of the 3-D voxel model over the mesh model as well as the texture data. Experimental results show the potential of our proposed multi-view image coding system. Furthermore, we propose important, yet simple, improvements to current 3-D voxel models; these improvements lead to significant coding gain within our 3-D voxel model based compression system.