基于模糊熵和RPCL的彩色图像聚类分割

提出了一种基于模糊熵和RPCL(rival penalized competitive learn ing)的彩色图像聚类分割算法。该算法可以自动确定图像的颜色类数目和初始类中心,从而提高了聚类的收敛速度,并且能够解决模糊熵阈值化分割算法所造成的过度分割问题。首先,计算彩色图像各颜色分量的模糊熵,获得分量模糊熵曲线,并根据模糊熵原理确定各分量的分割区域及聚类中心;然后,对各分量的聚类中心进行组合,形成彩色图像可能的聚类中心。但是,组合的聚类中心数目会多于实际的聚类数目,造成过度分割。因此,本文采用RPCL算法,对这些组合的聚类中心颜色进行学习来确定实际的颜色类数目以及聚类中心,并用学习后的聚类中心对原图像进行聚类分割。实验结果表明,该算法能有效地分割彩色图像,无需事先给定聚类数目和初始类中心。

Color Image Clustering Segmentation Based on Fuzzy Entropy and RPCL

This paper presents a clustering segmentation approach for color image based on fuzzy entropy and RPCL.It not only can adaptively detect the appropriate number and centers of the initial clusters of color image for RPCL and improve the learning rate,but also avoid over-segmentation caused by fuzzy entropy thresholding approach.Firstly fuzzy entropy of each color component is computed and initial clusters' centers of each color component are determined according to the fuzzy entropy curve.Then,these centers of different color components are combined to form the initial clusters' centers of color image.But the number of these combined clusters may be larger than that of the actual clusters,which may result in the over-segmentation.Therefore,RPCL is utilized to converge some of initial centers to actual centers of original color image and image is segmented by these learned cluster centers.The experiment shows that the method can effectively and adaptively segment color images without specifying the number and centers of initial clusters in advance.