Research On Image Segmentation Algorithms Based On FCM

The essence of image segmentation is to decompose the image into several different regions according to certain similarity criteria,and then extract the interested objects from them.Image segmentation technology has been attached great importance to by people.In the fields of medical image analysis,traffic control,meteorological prediction,geological exploration and detection,face and fingerprint recognition,image segmentation has achieved significant pioneering results.Many researchers have studied image segmentation deeply and put forward many effective image segmentation algorithms.Among them,the segmentation algorithm based on fuzzy clustering is a widely used algorithm,and the most representative one is FCM(fuzzy c-means,fuzzy C-means)algorithm.This paper mainly studies and improves the image segmentation algorithm based on FCM.Firstly,based on the analysis of image segmentation and existing FCM algorithms,an improved FCM algorithm is proposed,which combines genetic algorithm with FCM algorithm which fuses spatial neighborhood information,aiming at the problem that the initial clustering center of FCM algorithm is sensitive to local convergence and is sensitive to noise.The genetic algorithm has a strong ability of global optimization.By using it,the initial clustering center ofthe optimized FCM algorithm can be obtained,which can avoid the algorithm falling into local extremum.The FCM algorithm which integrates spatial neighborhood information can improve the anti-noise ability of the FCM algorithm to a certain extent.The combination of the two algorithms can achieve high quality segmentation effect and stronger robustness in image segmentation.Then,this paper applies the proposed improved segmentation algorithm to color image,combines the three-channel information of color image to improve the shortcomings of traditional FCM algorithm in color image segmentation.The color image is transformed from RGB color space to Lab color space,and the image is segmented on L,a and b channels respectively.Three initial segmentation results are obtained,and the three segmentation results are fused to get the final segmentation result.Finally,the algorithm is implemented and compared with the existing FCM algorithm through experiments.The experimental results show that the proposed algorithm achieves better segmentation effect.