A Study on View-Dependent Representation of 3-D Mesh Models Using Hierarchical Partitioning

As the demands of high quality visual service have increased recently from consumers, many researchers and companies have focused on the development of multimedia applications based on 3-D models, avatars and animations. However, it will be a difficult work to develop some applications related to 3-D mesh model transmission since they should accept the tremendous amount of transmission data for 3-D mesh models basically. Therefore, the study of efficient 3-D mesh model transmission scheme will be needed to manipulate those models in limited bandwidth channels. In this thesis, we propose a new scheme for the transmission of 3-D mesh model information according to human viewing. The proposed scheme is able to say a version combining sequential and progressive mesh transmissions with view-dependent transmission. A view-dependent nonuniform mesh information can be transmitted through a network by setting higher priority to visible parts than invisible parts. The proposed system can be divided into three parts: hierarchical mesh partitioning to represent a mesh model, resolution decision of partitioned meshes and view-dependent 3-D mesh model transmission. The first part is to represent a 3-D mesh model with a hierarchical partitioning structure. Hierarchical partitioning structure is obtained by doing mesh partitioning with previous partitioned meshes iteratively according to the specified number of initial partitioned mesh and level. In mesh partitioning, we use K-means algorithm for selecting start vertices. We also constitute multi-layers for progressive transmission based on the hierarchical mesh partitioning structure. Base layer and enhancement