Volume Rendering And 3D Geometry Filtering Based On Point Model And The Application In Medical Image Processing

Triangle is regarded as queen of graphics primitives.since a great success has been made in rendering and modeling by using of triangular meshes. With the development of three dimension (3D) scanning technology, such as Computed Tomography, Magnetic Resonance, Ultrasound, Laser Scanning and Structure Light Scanning, volume data and point sets become more and more large and triangular meshes constructed from volume data or point sets will become more and more large. Large meshes propose new difficulty for mesh processing because of connection information between meshes. It is more important that many slim triangles will be produced in large deformation (such as soft tissue simulation) as may decrease accuracy of simulation. Topology structure must be changed continuously in hydro-simulation. To simulate these physical phenomenons, meshes must be reconstructed real time. These difficulties raise enormous challenge for graphics hard. With the great success being made in visualization filed, many researchers realize that point model can be used to represent surface and can be used as a uniform model for volume and surface. In this background, point model is researched in the paper.The purpose of the paper is to do some research based on point model about Craniofacial surgery aided design system. It focuses on two kinds of data. One is volume data, another is point sampled surface. For volume data, isosurface extraction and hard acceleration for EWA splatting have been done. Iso-differential points instead of triangular meshes are used to represent iso-surface, where iso-differential points are the points have attritures of position, normal, principle curvature and principle direction. So the speed of iso-surface is increased. The obtained results show that the extraction time of isosurfaces can be reduced by a factor of four. An accelerating algorithm for EWA Volume Splatting is developed based on modern GPU. It mainly made use of point sprite, frame buffer object and programmable function of GPU. Experiments results demonstrate that the algorithm can be used in high quality and interactive visualization.For point sampled surface, some research about filtering has been researched deeply. First, a features-preserving filtering algorithm for point set surface and surface attributes is developed,which is based on robust statistics. Second, a nonuniform bilateral filtering algorithm for point set surface and surface attributes is developed, in which the tangent shift of points is avoided. The algorithm can be applied when points are sampled nonuniformly. Finally, a novel meshless method for point set processing is developed via partial differential equations (PDEs), which introduces well-known meshless local Petrol-Galerkin (MLPG) method to the field of graphics. The proposed approach does not need to construct the local or global triangular meshes, nor does need the global parameterization. It is based on the local tangent space and the local interpolated surface. By constructing the local symmetric weak form (LSWF) for every point, we can easily assemble PDE-specific mass and stiffness matrices. Corresponding sparse linear system can be solved by iterative solver. This algorithm can be directly applied to deformation for volume and shell. So it provides the base for surgery simulation.At the end, Craniofacial surgery aided design system has been introduced.