Chinese Journal of Computers   Full Text
Title	A High-Accuracy Contour Extraction Algorithm for MR Images Based on Mumford-Shah Model
Authors	YUAN Da1),2) ZHANG Cai-Ming1) LI Jin-Jiang1) LIU Xiao-Hua2)
Address	1)(School of Computer Science and Technology, Shandong University, Jinan 250100) 2)(School of Computer Science and Technology, Shandong Institute of Business and Technology, Yantai, Shandong 264005)
Year	2009
Issue	No.2(268—274)
Abstract &   Background	Abstract This paper proposes a high-accuracy contour extraction algorithm based on curve evolution model. Face to the blurred edges and high noise in MR image, Mumford-shah model is improved here and fuzzy clustering is used to establish active contour model to lower sensitivity to irregular details and noises. The derivation of the level set and the semi-implicit implementation based on the additive-multiplicative operator splitting is performed in order to improve the computing efficiency and accuracy. Experimental results are given to demonstrate the feasibility of the proposed method in extracting contour from the blurred edge and high-noise images. Keywords curve evolution; contour extraction; level set；operator splitting Background The research of this paper is supported by the National Natural Science Foundation of China under grant Nos.60773053, 60673153，60673003. There are some essential tasks of medical image processing，such as image segmentation, image smoothing and shape recovery. For a wide range of computer vision, a unified solution has been given by curve evolution，which provides an effective way for contour extraction. Furthermore, considerable results have been achieved on the research of curve evolution (e.g., classical snakes, GAC, Mumford-Shah and Chan-Vese method). However, there still exist some problems, such as original curves’ great impacts on analysis result，difficulty to obtain accurate segmentation of objects，etc. Based on curve evolution，we devote to the research of some critical problems of contour extraction. Our research aims to improve the extraction’s accuracy, reality and stability. In the modeling process of curve evolution and the process of numerical approximation, fuzzy clustering and the curve evolution framework are comprehensively used. Our main result is the development of an automatic extraction algorithm with four characters, i.e.: Application adaptive evaluation，independent of original curve, fast convergence and accurate location capability. The evaluation and comparison analysis of our algorithm are performed by theoretical and experimental analysis, and the results are given finally.