改进的基于模糊C-均值聚类的图像分割算法

Study on improved image segmentation algorithm based on fuzzy C-means clustering

目前的FCM类型的算法聚类数目的确定需要聚类原形参数的先验知识,否则算法就会产生误导.为了提高图像分割算法的抗噪性能,用K均值聚类算法简单、快速的优点对模糊C均值聚类算法进行改进.结合图像的邻域信息,对图像的直方图作均衡化处理,改善图像质量,通过自适应滤波,降低噪声对分割效果的影响.先用K均值聚类算法对图像进行分割,快速的获得较为准确的聚类中心和初次分割图像,避免了FCM算法中初始聚类中心选择不当造成的死点问题.用邻域灰度均值信息代替传统模糊C均值聚类算法中的灰度信息,对K均值聚类得到的图像作二次分割.该方法能更好的抑制噪声的干扰,提高了聚类算法的分割精确度.

The number of clustering in the FCM algorithm need prior knowledge of the cluste- ring prototype parameters, or the algorithm will produce misleading. This paper used simple and fast advantages of K- means clustering algorithm to improve fuzzy C -means clustering algorithm in order to improve the performance of the noise in the algorithm, and combined with the neighborhood information of the image and make equalization processing on the his- togram of the image. The quality of the image was improved. This paper reduced the noise effect to the segmentation through the adaptive filter. Used K- means clustering algorithm in the image segmentation to obtain accurate clustering center and the initial segmentation image in order to avoid the dead point problem caused by the inappropriate initial clustering center, obtained secondary division image with the neighborhood grayscale average information in- stead of the gray information of traditional fuzzy C - means clustering algorithm. This method could reduce the interference of noise better and improve the accuracy of segmentation of the clustering algorithm.