A Study For The Color Image Segmentation Based On Fuzzy Clustering

Color image segmentation is one of the critical techniques for the intelligent transportation system (ITS). This paper proposes a method feasible to the color image segmentation. Algorithms for color image segmentation have been developed in three steps. 1) The selection of color space model. The suitable color space is selected by calculating the evaluation value of object area and background area in each image according to the equation given in the paper. 2) The establishment of eigen-image through K-L transform. Eigenvectors picked-up by K-L transform set out eigen-image subspace which constructs the center of clustering; 3) Image segmentation by application of the fuzzy C-means algorithm. Fuzzy C-means algorithm is applied in supervised method and non-supervised method respectively. In this research, FCM based on eigen disseminated-degree in supervised method has the better segmentation effect. And at the same time Dynamic FCM based on comparability between sample and kernel in non-supervised is testified having ideal segmentation effect. Results by testing through a series of color vehicle images taken from various backgrounds, show that the algorithms developed in this paper are relatively effective and robust. Further research suggestions are given at the end of this paper.