Abstract: The traditional FCM algorithm has the shortcomings of simple image feature description， and it can be easily disturbed by complex texture in wrong segmenting. An adaptive filtered structure tensor based FCM algorithm for image segmentation was proposed to solve the problems. Firstly， a new anisotropy filtering based structure tensor was proposed to break the constraints of filtering direction and rotation of traditional filtering. Secondly， the image edge density function for adaptively calculating anisotropy filtering proportion was introduced into FCM algorithm to accurately measure image node gliding property. Then， an adaptive filter structure tensor similarity measurement was defined to replace the gray level similarity measurement in the traditional FCM algorithm. Finally， a new distance measure function was adopted to calculate the distance between a node and the clusteringcenter. The simulation results showed that more precise segmentation results could be obtained from complicated texture structure images by the presented algorithm.

Key words: image segmentation, FCM algorithm, structure tensor, Gaussian function, anisotropic filtering