Three-Dimensional Reconstruction Based On Sectional Medical Image Of Human Body

With the research development of medical image application in clinic, three-dimensional reconstruction based on sectional medical image of human body has become more and more significant. There are two methods in 3D reconstruction of medical images: volume rendering and surface rendering. Surface rendering can get acceleration with current hardware which is suitable in run-time interaction, so it is widely used.Marching Cubes algorithm is a typical method of the surface rendering using triangular faces as the basic elements to construct isosurface. The algorithm itself is relatively simple and effective in support of graphics hardware acceleration, which has a profound impact on the surface rendering. Based on the Marching Cubes algorithm, an approach is proposed for medical three-dimensional reconstruction in view of the characteristics of modern medical image in this paper.The traditional Marching Cubes algorithm traverses the entire image space to get the isosurface, whereas in medical image the isosurface passes through only a small percentage of the total cubes. The proposed approach in this paper combines with seed-occupying algorithm to expand isosurface, which can avoid detection of the empty cube and improve the algorithm speed greatly. The existing medical image data sets have higher density and good drawing effects, but tend to generate too many triangular faces. These faces are increasingly small and the individual differences are more subtle, so calculation of the interpolation position is of little significant. The proposed approach uses intersected edge midpoints instead of intersection points in processing the single cube, which can be exploited to reduce the computational cost and produces little effect on the isosurface. When generating the isosurface, because the approach uses polygons which are formed of isopoints directly instead of triangles without resorting to the pre-defined cases, it reduces the number of faces greatly and gives a good solution to the ambiguity problem inherent in MC algorithm. On the basis of detailed description of the design and implementation of 3D reconstruction of medical images, the thesis focuses on the depiction of improved method. Experimental results illustrate that our method gives a great improvement on topological structure, storage space and processing speed. The generated model is made up of fewer points and faces, so it allows real-time operations and interactive performance with computer by less hardware cost.