A Computer-Aided Design Method For Segmentation Of Volumetric

Image segmentation is the process of partitioning an image into meaningful regions. Those regions represent objects or their parts. The current image segmentation methods are unable to accurate extract the various objects of medical image data. In addition, even image segmentation can convey accurate image information, difficulties arise when properties within objects vary or boundaries of objects become blurred, and these variations are often unpredictable. These reasons make it impossible to develop an automatic method that can segment all images correctly. Because a high accuracy is demanded in the segmentation of medical images, the user has a critical role in examining the results and correcting the possible errors. The form of changes could be 2-D(2-Dimensional) or 3-D(3-Dimensional), and the characteristics of this paper is that we can edit the 2-D data by changing the 3-D model.In this paper, a computer-aided design method is introduced that enables the user to effectively correct such errors. The proposed method approximates a 3-D region by a parametric surface which is rational Gaussian surfaces. The user can edit the surface shape by changed the parameters of the position of control points. First, we use the coarse-to-fine method to parametrize a digital shape. This algorithm major includes: subdivision of a single triangle algorithm, approximation of a digital shape by a triangular mesh and parametrizing the shape points algorithm. Then, the paper studied the principle and generation methods of rational Gaussian surfaces, and completed the algorithm of approximating digital shape. At last, the method implements the 3-D visualization in Visual C++.NET platforms with OpenGL, and gives the interface of edited image segmentation results, and completes editing the segmentation results functions.This method provides tools for the user to effectively revise the surface by viewing both the surface and the volumetric image and pulling or pushing the surface locally until the desired region is formed. The only limitation of the method is that it works only on spherical topology regions. With this method, the user typically needs only a fraction of the time to correct local errors obtained by an automatic segmentation method.