A Segmentation-based Triangular Mesh Generation Scheme for Mesh-based Video Coding

A new mesh generation algorithm is proposed for mesh-based video coding. The still segmentation result of watershed segmentation is projected onto motion segmentation result of change region detection in proposed algorithm. Therefore, mesh presentation generated by our new method is able to reflect both spatial and temporal features of input video sequence. This experimental results show that meshes generated by proposed algorithm perfectly match still image contours and motion object boundary, and that it significantly reduce nodes and patches with similar PSNR performances of conventional methods. Thus accurate motion compensation, low computation complexity and high compression ratio can be achieved in mesh-based coding using proposed algorithm.