一种基于属性均值聚类的图像分割方法

Image segmentation method based on attribute mean clustering

为了降低图像噪声以及边界模糊对图像分割的影响,把属性均值聚类的思想应用于图像分割。首先,采用自己设计的非线性自适应滤波器对图像进行降噪预处理,接着采用模糊熵进行粗分割。最后,针对由于人工设置初始隶属度矩阵不当而导致属性均值聚类收剑速度慢的问题,把模糊熵最终得到的隶属度矩阵作为属性均值聚类的初始参数进行二级分割。通过实验表明该方法最终分割出了自己感兴趣的区域,效果较好。

To solve the problem about noise sensitive and fuzzy edge, Attribute Mean Clustering （AMC） method was applied in the image segmentation. Nonlinear adaptable filter was done firstly to pretreat image to lighten the influence of noise. Then fuzzy entropy was used to segment the image. At the end, due to the speed of convergence in the fuzzy Cauchy AMC, the fuzzy subjection got from the last result of fuzzy entropy was used to segment the image. Experiments show that the method precisely gets the interesting region and segments the image better.