基于模糊聚类的小波变换图像去噪算法改进  

IMPROVING DENOISING ALGORITHM FOR FUZZY CLUSTERING-BASED WAVELET TRANSFORM IMAGE

作　　者：张焰林[1] 

机构地区：[1]温州职业技术学院计算机科学系,浙江温州325035

出　　处：《计算机应用与软件》2010年第8期133-135,171,共4页Computer Applications and Software

基　　金：浙江省教育厅科研项目(Y200805669)

摘　　要：介绍一种改进的较优的基于模糊聚类的小波变换图像去噪算法。首先分析了模糊C均值聚类算法中加权指数m的重要性,采用基于模糊决策的方法,分别构造模糊目标和模糊约束,由模糊目标和模糊约束的交集来共同确定最优的加权指数m以获取较为理想的聚类分类结果。再利用该种加权模糊聚类算法把小波系数划分成包含信号与只包含噪声的小波系数两类,将只包含噪声的小波系数置为零,将包含信号的小波系数利用软阈值法进行收缩,最后对处理后的系数根据M带小波变换的局部时频分析能力及其良好的信噪分离能力进行M带小波变换,得到去噪效果较好的图像。This article introduces an improved acceptably good denoising algorithm for fuzzy clustering-based wavelet transform image. First of all, we analysed the importance of weighted index m in fuzzy c-means clustering algorithm, in which fuzzy decision-making method is applied to constructing fuzzy goals and fuzzy constraints. Then, the intersection of fuzzy goals and fuzzy constraints determines jointly the opti- mal weighted index m in order to obtain better classification results of clustering. After that, the weighted fuzzy clustering algorithm is used again to divide wavelet coefficients into two types -- the wavelet coefficients with signal and the wavelet coefficients with noise only. The latter will be assigned zero, and the former will be shrunk using soft threshold method. At last, the processed coefficients are executed M-band wavelet transform according to the abilities of local time-frequency analysis and good noise separation of the M-band wavelet transform to get the images with fairly good denoising effect.

关 键 词：模糊聚类 M带小波 图像去噪 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置]