(A) 3-D rendering system using wavelet transform-based data compression and mesh approximation for very large terrain data

Abstract

In this dissertation, we present a new mesh approximation method for the regular gird 2-D data in order to implement a fly-through simulator. The proposed mesh approximation method uses wavelet coefficients as the local area complexity measures. To utilize the characteristics of wavelet transform (WT), we examine the characteristics of WT and devise a new mesh approximation scheme. Since the wavelet transform has the spatio-frequency localization characteristic, this method only selects the wavelet coefficients larger than a given threshold and constructs quadtree based triangular meshes using them. Computer simulation results show the advantages of the proposed method in PSNR and visual quality. As the size of terrain data becomes larger, the need for the effective terrain data management scheme is increased. Hence, in this dissertation, we also propose and implement a new 3-D rendering system to visualize the very large terrain data. It combines an image compression method and the visualization scheme mentioned above. For the efficient data compression, the set partitioning in hierarchical tree (SPIHT) codec, which is a kind of progressive codec based on wavelet transform and the block-based data compression method are adopted. The rendering system decodes the compressed terrain data block by block depending on the visual resolution based on the local level of detail method and makes a triangular mesh based on the proposed mesh approximation method. Since the compression of terrain data and level of detail in mesh approximation are combined algorithm, the rendering system can manage the large terrain data effectively. Computer simulation shows that the proposed rendering system can provide good visual quality as well as high performance. The proposed system can be applicable to various fields, such as, flight simulator and computer games