The Research And Implementation Of 3D Reconstruction Based On Improved MC And MT Method

Presently, the 3D reconstruction of medical images has gained internal and overseas comprehensive attention in the fields of medical treatment and image processing. Its application in human organs, tissues or focus can represent the objects more directly and precisely, helping doctors analyze them quantitatively, improving the veracity of diagnosis. Image segmentation is basis of 3D reconstruction and can directly affect the speed and results of 3D reconstruction, so the images are segmented before reconstructed in this thesis. Image segmentation is a difficulty and bottleneck of image processing, so it is one of hard points in this thesis.The thesis researches the traditional 3D reconstruction methods. The methods of 3D reconstruction are mainly divided into two groups:Surface Rendering and Direct Volume Rendering. The Surface Rendering method is applied widely in correlative fields because of its celerity and better interactive capability. Marching Cubes algorithm (MC) is one of classical Surface Rendering methods, but has the problem of topology ambiguity. Marching Tetrahedral (MT) method was developed based on MC method and it can avoid the topology ambiguity in a certain extent, but it has decomposing ambiguity and produces overabundance triangles. In terms of the problems above, this thesis proposes the conception of pyramid decomposing, combines MC method and MT method and puts forward a self adaptive mixed-decomposing method. The method can not only resolve the problem of too many triangles, but also avoid topology ambiguity and decomposing ambiguity. Finally, a rendering accelerating algorithm based on data buffer storage technology is suggested to improve the efficiency of mixed-decomposing method. In order to represent the organs syllabify, this thesis proposes an improved region growing method that can grow in adjacent images. Then, based on the above mixed-decomposing method, the segmented images are reconstructed. At last, the methods in this thesis are validated through experiments.