S Improved Fast Two Cycle by using KFCM Clustering for Image Segmentation

Among available level set based methods in image segmentation, Fast Two Cycle (FTC) model is efficient and also the fastest one. But its efficiency is highly dependent to contour initialization. This paper tries to improve this method by using a kernel-based fuzzy c-means (KFCM) clustering algorithm. The proposed approach consists of two successive stages for image segmentation. Firstly, the KFCM is used to cluster the input image. Then ROI’s fuzzy membership matrix is used for next stage as an initial contour. Eventually, FTC model is utilized to segment the image by curve evolution based on level set. As a result, two benefits are provided for image segmentation in addition of advantages of FTC model. They are independency of curve initialization and reduction of user intervention. Experimental results show promising outputs in segmentation of different kinds of image including medical data imagery, natural scene and synthetic data.