一种改进的模糊核聚类红外图像分割算法

An Improved Infrared Image Segmentation Algorithm Using Fuzzy Kernel Clustering

针对模糊核聚类对红外图像分割存在的不足,提出了一种改进的模糊核聚类红外图像分割算法。首先在模糊核聚类的基础上引入了隶属度和空间约束关系,有效抑制了野点;然后定义了像素对类别的认同度指数和类别对像素的排斥性度量,并将之引入到隶属度函数中,判断像素的分类合理性,提高聚类的精度,更好地分割目标和背景区域,保护目标的完整性和精确性。实验结果表明,与传统的模糊聚类分割结果相比,该算法能准确完整地分割出目标,防止背景像素和野值点对目标区域的干扰,获得良好的分割效果。

Due to the problems of infrared image segmentation using fuzzy kernel clustering, an improved method for infrared image segmentation was proposed. Firstly, the membership and spatial constrained were introduced. The outliers were suppressed. Then the degree of self- recognition for pixel to categories and repulsive metrics for categories to pixel were defined. And the membership functions were revised using them. The rationality of classification was judged. The precision of classification was improved. The target area and background region could be segmented better. The integrity and precision of target were preserve. The experimental results show the infrared image can be segment well by the proposed method using compared with the conventional method. The obstruction to targets recognition by background pixel and outliers was prevented.