Font Size: a A A

Study On Theory And Techniques Of Geometry Reconstruction From Slice-Images

Posted on:2003-11-08Degree:DoctorType:Dissertation
Country:ChinaCandidate:F X JiFull Text:PDF
GTID:1118360065456269Subject:Mechanical Manufacturing and Automation
Abstract/Summary:PDF Full Text Request
Geometry reconstruction from slice-images is a main topic of visualization of 3D volume data. Reconstruction of model is the premise of analysis, emulation and decision-making. This dissertation mainly studies and deals with the technique of geometry reconstruction from cross-sectional contours and some related techniques, such as gray-level interpolation or shape-based interpolation of slice-images, contours extraction, surface reconstruction from planar contours and 3D solid model transformation.The high similarity in shapes and large overlapping areas of corresponded contours on adjacent cross-sectional are main conditions in reconstructing surface correctly. In accordance with the disadvantage of blurring caused by traditional gray-level interpolation methods, this dissertation introduces a method to interpolate 3D images based on point matching approach to get 3D isotropy volume data. With respect to non-similar shapes or no superposition of corresponded contours on adjacent slices and concave contours, this dissertation presents a method for shape-based interpolation based on nonuniform scale bounding box.Extracting contours is a main pre-processing of 3D reconstruction. Wavelets transform has good local character and multi-scale character. This dissertation proposes an algorithm to detect edges and extract contours from medical images based on quadric spline wavelets transform. The algorithm can generate single, continuous and close contours.There are two main methods of surface reconstruction ?explicit connecting contours method and implicit function method. In order to reconstruct object with holes, a method is presented to detect contour inclusion relationships based on seed-filling approach. Cylinder growing method is used to determine the linking relationships of contours between slice-images. Branching problem is a difficulty in surface reconstruction from contours. This dissertation proposes an algorithm to construct transition contours using mathematical morphology operation. This algorithm can ensure transition contours are similar to trunk contours and branch contours and can ensure that reconstructed transition surfaces at branch points are smooth. A method is also presented to solve tilling problem based on Delaunay triangulation approach. The method considers the aspect character of triangles and uses the angle-optimal of triangulation as optimization criterion in reconstructing surface. By the method of local adjustment, surface is optimally triangulated based on the principle of Delaunay triangulation.This dissertation also studies surface reconstruction using implicit function methods. This method can avoid losing surface details, but its calculation is computationally expensive. The projection characters of adjacent contours and their connected surfaces are analyzed. A new algorithm for reconstruction is proposed based on projection. A new concept of "Connected Surface Projection Region " is introduced. Using distance function, only the points whose projections are inside this region are calculated to generate transition volume and reconstructed surface. Much time of computation is saved. Marching Cubes (MC) algorithm and Marching Tetrahedra (MT) algorithm are introduced. This dissertation chooses MT algorithm to extract zero isosurface in order to avoid ambiguity of MC algorithm. In accordance with the intersection feature between isosurface and voxel, this dissertation improves MT algorithm and proposes nine configurations of distance function values of tetrahedron vertices. The definition of distance function is modified and a method is presented to reconstruct connective branch objects.Finally, this dissertation studies the transform process from reconstructed surface model to solid model. The key techniques for reverse engineering based on CT image are introduced. 3D reconstructed surface model doesn't possess full geometry and topology information as solid geometry model. This dissertation presents an algorithm for rebuilding solid model from triangulated patch su...
Keywords/Search Tags:Slice-Images, Cross-Sectional Contours, Surface Reconstruction, Image Interpolation, Edge Detection, 3D Surface Model, Solid Geometry Modeling
PDF Full Text Request
Related items