Study On The 3D Reconstruction Of Medical Images

The3D reconstruction of medical images has brought a lot of attention and interests nowadays, which is based on the combination of image segmentation and analysis, computer graphics and virtual reality techniques. Its application to human organs, tissues or infected regime can represent the objects more directly and precisely, helping doctors to analyze them quantitatively, improving the diagnosis. Image segmentation its basis, and affects its speed and results, becoming the bottleneck of its utilization and a worldwide highlight.So far, the standard Marching Cubes (SMC) algorithm is one of the most popular and applicable 3D reconstruction methods. However, it is still necessary to improve its quality, and especially reduce its computation and triangles. This dissertation focuses on the simplification of SMC. By shifting the vertex position to the two endpoints of the interpolation edge respectively, i.e., shrinking or expanding the isosurfaces, two approaches have been proposed. Algorithm analyses and experiments show that both of these methods can simplify the triangulation patterns, reduce the triangles obviously, avoid the interpolation computations for the vertex positions and vectors, and shorten the extraction time remarkably.As to the image segmentation technique, a new algorithm based on the 2D Tsallis entropy has been put forward in the present work. This method avoids the logarithmic calculations of the Shannon entropy, and utilizes the regional gray levels additionally. Segmentation and subsequent 3D reconstruction experiments show that this method can effectively improve the segmentation quality, and reduce the computational complexity.Furthermore, a set of 3D reconstruction software packages have been developed successfully in the present work via Visual C++ OpenGL, which have been employed in the experiments.