| The development of scientific visualization makes virtual endoscopy(VE) of medical images become a hot research topic which relates to medical imaging, digital image processing, computer graphics, computer visualization and virtual reality. It has a wide application in medical diagnosis because it is a method of noninvasive, non-touched non-hurt examination compared to traditional endoscopy.Some key technologies and implementations of VE are systematically investigated in this thesis. Firstly, several segmentation algorithms of medical image are presented, by analyzing their advantages and disadvantages, we use the seed-filling methods based on scanning line to realize a three-dimension(3D) segmentation method for medical volume dataset. Secondly, a introduction of Marching Cubes(MC), volume rendering and GPU-based MC is given, and virtual endoscopies based on these three kinds of rendering approaches are also realized. Then, aiming at the path planning of VE, distance transformation is introduced to compute an accurate distance from boundary (DFB) field for the 3D tubular object, construct the maximum spanning tree (MST)and extract the centerlines of organs. Finally, based on all the key technologies above, a VE system with effective interaction is developed which provides the automatic roam, manual roam and inductive roam. The experimental results with the real clinic image data illustrate the effectiveness and practicability of our VE system. |
PDF Full Text Request