采用聚类分割和直方图均衡的图像增强算法  

Image enhancement using clustering and histogram equalization

作　　者：曹军峰[1,2,3] 史加成[4] 罗海波[1,2] 常铮[1,2] 惠斌[1,2] 

机构地区：[1]中国科学院沈阳自动化研究所,辽宁沈阳110016 [2]中国科学院光电信息处理重点实验室,辽宁沈阳110016 [3]中国科学院大学,北京100049 [4]清华大学电子工程系,北京100084

出　　处：《红外与激光工程》2012年第12期3436-3441,共6页Infrared and Laser Engineering

基　　金：国防973项目

摘　　要：针对现有直方图均衡化算法难以兼顾图像细节增强和噪声抑制问题,提出了一种基于聚类分割和直方图均衡的图像增强算法。引入模糊C均值聚类(FCM)算法将低通滤波后的图像分成若干类,并利用原图的边缘信息和形态学操作修正聚类结果;根据各类的均值和标准差确定映射后的动态范围,对其进行直方图均衡化处理;采用平滑滤波方法消除类间的边界效应。实验结果表明:经过该算法处理后的图像,对比度得到明显增强,噪声得到有效抑制。该方法不但可用于图像增强,还可用于图像动态范围压缩,具有广泛的应用前景。It was difficult for the conventional HE methods to give consideration to restraining noise while emphasizing features. A novel image enhancement technique based on image segment by clustering and histogram equalization was proposed to overcome the drawback. The input image was smoothed by a Gaussian filter, and the fuzzy C-means clustering algorithm was used to segment the low-pass image into several partitions, and then the edge information and morphological operations were used to fix the segmentation result. Each partition was assigned to a new dynamic range according to its mean and standard deviation. Based on the dynamic range, the histogram equalization process was applied independently to these partitions. Smooth filtering was applied to eliminate the blocking-effect. The results of experiments showed that been processed by the proposed algorithm, the contrast was significantly enhanced and noise was effectively suppressed. The proposed algorithm not only can be used for image enhancement, but also can be used for dynamic range compression, so it has broad application prospects.

关 键 词：图像增强 图像分割 直方图均衡 模糊C均值聚类 

分 类 号：TN911.73[电子电信—通信与信息系统]