Study On Image Segmentation Algorithm And Application In Industrial CT

Image segmentation is the technology and procedure which divide image into different parts and find interested object while according to some standard. It is the precondition of further work which is to analyze, explain, and recognize image. It is also the important base work of a lot of advanced image processing technique as image video, image measurement and image compression. Image Segmentation is one of the key techniques in digital image processing domain, this significance make it become the hot issue which is researched by many scholars.The image segmentation method have been developed decades of years as yet, scholars have proposed and applied lots of algorithms from different theory point of view. In this thesis, several familiar theories and algorithm characters of image segmentation method which based on edge detection were studied first; also the corresponding experiment and analysis were given. Then, aiming at active contour algorithm which is based on partial differential equation theory and facet model which is based on edge detection, special research, improvement and actual application were done.Active contour model is one of the image segmentation method based on partial differential equation, the main idea is the definition of energy function. The local minimum of the energy function is searched when the contour approaches to the real contour of object from initial contour gradually, namely, it approaches to the real contour of object by dynamic optimization of energy function. This method could conquer the affection of noise effectively, so that it is adapted to the processing of image with big noise. First, the fundamental of active contour model was expounded, the algorithm of traditional parametric active contour model and improved approaches were researched, in allusion to the problem that the model is sensitive to the initial contour, the means of getting initial contour was improved, then the improved algorithm was applied in practical problem of industrial CT, also the defect segmentation and parameter measurement of practical CT image were done, the satisfied results were obtained.Facet model is an image segmentation method based on image edge. Using facet model, we can obtain the precise image edge to meet the needs of high accuracy. But it will consume too much time when be used to detect 3D image edge, so it is difficult to be applied in 3D edge detection of actual industrial CT image. Against this problem, a method of 3D edge detection based on wavelet transform and facet model was presented. Firstly, 3D wavelet transform was used to 3D data of Industrial CT; the initial candidates of edge could be obtained by setting smaller threshold. Secondly, 3D facet model was used to decide if the candidate is the true edge. Then the 3D image edge of high accuracy was located. The improved method not only has the equivalent accuracy with direct facet model, but also increases the speed greatly. The satisfied results were obtained when the improved method was used to actual industrial CT image.