The Visualization Of Large-scale Hierarchical Atlas

Hierarchical atlas is popular in our life. But when atlas contains millions of images, traditional browsing and moving display becomes inefficient and unbearable. Visualization can offer effective ways for large-scale atlas combined with data classification, data variation and multidimensional tags. This thesis focuses on the visualization for large-scale hierarchical atlas; it can be organized as following:(1) Design visualization scheme for hierarchical atlas. Voronoi Treemap is used to present hierarchical information since it has advantage in dynamic and stable layout. As Photomosaic show images in an aesthetic way, so it is used to express images in atlas and their information.(2) Optimize Voronoi Treemap computation and propose a stable layout algorithm. The optimization includes a mean-value method to reduce the error of CVT computation, and an iteration ending condition based on the error variant rate and node’s weight error. Thus the time consumption of CVT computation decreases nearly 1/3. To fit for the data variability, we designs stable layout by focusing on topological structure and weight. And a set of contour schemes such as template based, image contour extraction and user-defined are designed to support our Voronoi Treemap layout.(3) Extend and optimize Photomosaic to visualize atlas. We extend Photomosaic method to interaction and multi-layered atlas. FFT is used to speed up tile matching process to fit for large-scale images filtering. A target image selecting strategy is proposed by comparing the output of mosaic images.(4) Build hierarchical atlas visualization prototype system. The system contains data processing, Voronoi Treemap and Photomosaic module. At last, instants of "China Plants Pictures Library" are introduced into the prototype system as case study. The system runs well for thousands of layered images by Voronoi Treemap for hierarchical representation and Photomosaic for atlas representation.Through the study of the large-scale hierarchical atlas, we combine Voronoi Treemap and Photomosaic to show data’s information. We improve the computational accuracy of Voronoi Treemap and the efficiency and dynamic interaction of Photomosaic.