Current Issue Cover
一种快速鲁棒核空间模糊聚类分割算法

吴其平,吴成茂(西安邮电大学)

摘 要
目的 针对现有鲁棒核空间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模糊聚类分割算法。方法 利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,为进一步提高算法时效性,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,设计一种基于二维直方图的鲁棒核空间模糊聚类快速分割算法。结果 对大幅面图像在一定强度的高斯、椒盐以及混合噪声进行对比实验,本文算法比基于邻域空间约束的核模糊C-均值聚类等算法峰值信噪比值至少提高1.5dB,误分率约降低5%,聚类性能评价的划分系数提高约10%,且划分熵降低约9%,运行速度比核模糊C-均值聚类和基于邻域空间约束的核模糊C-均值聚类算法至少提高30%,且与一维直方图核空间模糊C-均值聚类算法具有相当的时间开销。结论 本文算法相比现有空间邻域信息约束的鲁棒核空间模糊聚类等算法具有更强的抗噪鲁棒性,更优的分割性能和实时性,对大幅面遥感、医学等影像快速解译具有积极的促进作用,能更好地满足实时性要求较高场合图像分割的需要。
关键词
Fast robust kernel-based fuzzy C-means clustering segmentation algorithm

WU Qiping,WU Chengmao(Xi’an University of Posts and Telecommunications)

Abstract
Objective Considering that the existing robust kernel-based fuzzy C-means clustering segmentation algorithm is too difficult to meet the needs of fast segmenting image that has large size and been interfered by strong noise , a fast robust kernel-based fuzzy C-means clustering segmentation algorithm is proposed. Method The linear weighted filtering image which is firstly obtained by combining the neighborhood pixels gray information with the space information. Secondly, To improve currency of the algorithm, Two-dimension histogram between the clustered pixels and the mean value of its neighborhood pixels is embedded into the robust kernel-based fuzzy C-means clustering. In the end, a fast robust kernel-based fuzzy C-means clustering segmentation algorithm based on the two-dimension histogram is presented. Result Segmentation results of large scales images interrupted by Gaussian, salt-and-pepper and mixed noise show that compared with the kernel-based fuzzy c-means clustering with spatial constraints of the neighborhood information,and other kernel-based fuzzy clustering segmentation algorithms. the proposed clustering segmentation algorithm has raised about 1.5dB of the peak signal to noise ratio, reduced at least 5% of the misclassification rate, increased about 10% of the partition coefficient, and decreased about 9% of the partition entropy. In the meantime, the running speed of it is enhanced 30% compared with the kernel-based fuzzy c-means clustering method and kernel-based fuzzy c-means clustering with spatial constraints of the neighborhood information, and the running speed of the proposed clustering segmentation algorithm is similar to that of kernel-based fuzzy c-means clustering method based on one-dimension histogram. Conclusion Compared with the existing robust kernel-based fuzzy clustering segmentation algorithm based on spatial constraints of the neighborhood information, the proposed clustering segmentation algorithm has stronger ability against noise and advantages of better segmentation performance and real time. Meanwhile it has a positive effect on rapid interpretation of images such as large scales remote sensed images, medical images and so on, and that the proposed clustering segmentation algorithm meets the needs of the effectiveness of image segmentation.
Keywords
QQ在线


订阅号|日报