一种快速鲁棒核空间图形模糊聚类分割算法

针对现有鲁棒图形模糊聚类算法难以满足强噪声干扰下大幅面图像快速分割的需要,提出一种快速鲁棒核空间图形模糊聚类分割算法。该算法将欧氏空间样本通过核函数映射至高维空间;采用待分割图像中像素邻域的灰度和空间等信息构建线性加权滤波图像,对其进行鲁棒核空间图形模糊聚类;并引入当前聚类像素与其邻域像素均值所对应的二维直方图信息,获得鲁棒核空间图形模糊聚类快速迭代表达式。对大幅面图像添加高斯和椒盐噪声进行分割测试,实验结果表明:本文算法相比基于图形模糊聚类等分割算法的分割性能、抗噪鲁棒性和实时性有了显著提高。

A fast and robust clustering segmentation algorithm for kernel space graphics

Addressing the existing problem of difficulty in realizing fast segmentation of large-scale images under strong noise interference, a fast robust kernel space graph fuzzy clustering segmentation algorithm is proposed. This algorithm first mapped the samples in European Space to the high dimensional feature space through the kernel function; subsequently, it constructed the linear weighted filtering image using the gray scale and spatial information of the pixel neighborhood in the image to be segmented and carried out the robust kernel space pattern fuzzy clustering on the image. The fast iterative expression of robust kernel space graph fuzzy clustering was obtained by introducing the two-dimensional histogram information corresponding to the mean value of the current clustering pixel and its neighboring pixels. Experimental test results of large size images interrupted by Gaussian and salt-and-pepper noise show that the segmentation, robustness, and real-time performance of the proposed segmentation algorithm have improved more significantly than those of the picture-based fuzzy clustering, and other fuzzy clustering segmentation algorithms.