Research On The Methods Of Image Segmentation Based On Fuzzy Clustering

Image segmentation is a technology and process in which the image to be divided into characteristic regions,and extract the target of interesting.It is the basis of image analysis and understanding,and playing an important role in the image project.Image segmentation has been attention for many years as one of the basic computer vision technology.Due to there is some uncertainty in images,and fuzzy theory precisely has a better ability to describe it,thus fuzzy clustering has been applied into image segmentation problems and achieved better effect.However,fuzzy clustering algorithm also has some problems,such as not good at anti-noise performance and very sensitive to initialization.In this paper,based on the theory of fuzzy clustering,through the division of data sets to lead fuzzy C-means algorithm,and the defects of fuzzy C-means clustering algorithm for image segmentation has been improved as following:(1)Based on the one-dimensional histogram of FCM clustering is only use some image information for image segmentation,so it may has the problem of incomplete image segmentation,through the use of spatial information,using the average gray value of neighborhood pixel to construct the two-dimensional space,instead of gray value as the sample sets,more completely utilizing the image imformation.(2)The traditional FCM algorithm is essentially a local search algorithm,very sensitive to the initialization,it is difficult to obtain the global optimal solution,and easy to fall into local optimal solution.Through genetic algorithm,the use of the powerful general of genetic algorithms to solve the problem,to solve the initialization sensitivity and falling into the local optimum.Genetic algorithm is suited to optimization in the large-scale space.Experimentations show that the fuzzy clustering based on genetic algorithm can achieve a good result.