基于顾及像素空间信息的加权FCM聚类的图像分割

Image segmentation based on weighted fuzzy C-means clustering accounting for pixel spatial information

针对标准的FCM算法没有考虑像素的空间信息而对噪声比较敏感和没有考虑不同样本数据对聚类效果的不同影响的不足,提出了一种顾及像素空间信息的基于图像的灰度直方图加权的FCM聚类算法,它在Szilagyi等提出的算法基础上通过引入图像的灰度直方图加权对算法中的目标函数进行修改.对人工合成图像和真实图像的数值模拟结果均显示出该算法的优良性能.

The standard FCM algorithm is noise sensitive because of not taking spatial information into account, and it considers that each feature datum has the same contribution to classifying results. To overcome the above problems, this paper presented a modified FCM algorithm accounting for pixel spatial information based on gray histogram weight. The proposed algorithm was realized via introducing a gray histogram weight in the objective function given in Szilagyi＇ s algorithm. Experimental results on both artificial synthesized images and realistic images demostrated the sound performances of the proposed algorithm.