Dynamic Mesh Coding Using Edge Length-Based Adaptive Subdivision

Video-based Dynamic Mesh Coding (V-DMC) compresses dynamic mesh data by using a low-resolution mesh and displacement data that is added to the subdivided low-resolution mesh at reconstruction. By increasing the number of subdivisions, more displacement data is applied, and higher quality is obtained. However, the number of reconstructed faces becomes larger than the input face count. In an environment with limited computing resources, it is necessary to keep the number of triangles of the reconstructed mesh the same as the one of the original mesh. It is a challenge to achieve high quality while keeping the number of reconstructed faces. This paper describes the edge length-based adaptive subdivision method adopted in V-DMC. This method restricts refinement of small faces by checking edge length of each face. A target number of decoded mesh faces can be reached by setting a threshold. Experimental results show the proposed method provides finer control over the bitrate and quality of the V-DMC stream.