Research On GPU-based Non-photorealistic Volume Visualization

With the developments of modern medical equipment, large scale volumetricdata can be easily achieved, which presents a greater challenge to data visualization.And how to make a better view, this issue has become a field of scientificvisualization research hotspot. The transfer function design and illuminationrendering method is the core problem of visualization, and this also becomes aresearch focus.This paper studies the three-dimensional medical data visualization methods.In which, we discuss how to use the direct volume rendering method to obtainhigh-quality rendering result by design a transfer function. Besides we use GPU toaccelerate this process. In this paper, we propose a highlight important structuraltransfer function based on LH histogram frequency information. In addition, wealso propose a color transfer function which based on a novel color stretchingmethod. Furthermore, we also propose a multi-channel tone shading model whichcombines various structure enhancement methods in a unified illustration model.Halo and context-preserving are used for depth enhancement. For regionenhancement, we improve the volume rendering model by making use of statisticalinformation in tubular region around projection lines. In addition, we presentGPU-based optimization strategies which significantly improve the performance ofour algorithms. At the same time, we accomplish a visualization platform throughOpenGL for further research. In particular, this platform also supportsfour-dimensional data set (with an additional time dimension).In this paper, we use a variety of data sets for testing, including the VisibleHuman Project (VHP) data set, three-dimensional cardiac CT data set. Bycomparing with the existing methods, we find that the method we proposed hasobvious advantages, which can make a better visualization. Then we build afour-dimensional virtual organ visualization system to render volume data, whichalso has an excellent render result.