摘要：

传统聚类算法易陷入局部极值，在数据线性不可分时分类效果较差。为此，提出一种基于最大熵的模糊核聚类图像分割方法。采用最大熵算法对原始图像进行初步分割，求得初始聚类中心；引入Mercer核函数，把输入空间的样本映射到高维特征空间，并在特征空间中进行图像分割。实验结果表明，该方法能减少迭代次数，使分类结果更稳定，从而较好地把目标从背景中分割出来。

关键词: 模糊核聚类, 最大熵, 特征空间, 图像分割

Abstract:

The traditional clustering method is prone to fall into local extremum. It is bad to classify when the data is linear inseparable. This paper proposes a fuzzy kernel clustering image segmentation method based on maximum entropy. It applies maximum entropy algorithm to obtain the initial centers and maps the sample from the input space to the feature space by introducing Mercer kernel function into the method. It completes image segmentation in the feature space. Experimental result shows that the method can reduce the iteration time and steady the class result, and effectively segment the target from its background.

Key words: fuzzy kernel clustering, maximum entropy, feature space, image segmentation

中图分类号: 

• TP391