【摘要】 在图像分割领域,模糊C-均值聚类算法得到了广泛的应用,但存在计算量大、易受噪声影响、目标与背景对比较弱时对边界处的像素分辨能力低等问题。针对以上问题对标准模糊C-均值聚类算法进行了改进:利用一维灰度直方图来降低计算量;并在此基础上,考虑每一层灰度级的邻域像素之间的空间一致性;然后,构造特征散度来重构聚类算法的目标函数。最后用一幅测试图像和两幅航天高分辨率图像对改进的方法进行试验,结果表明,对于削弱上述问题的影响,算法较标准模糊C-均值聚类算法有较大提高。更多还原

【Abstract】 Fuzzy c-means(FCM) clustering algorithm has been widely used in the field of image segmentation.It is remarkable that there are such problems as large computation cost,sensitivity to noise and low resolution of pixels in boundaries because of poor contrast between objects and background.The traditional FCM was modified in this article according to these difficulties.The one-dimension gray-level histogram was utilized to decrease the computation complexity.Moreover,the spatial consistency of the neighborhood pixels in each gray level was taken into account.Furthermore,the objective function of the proposed clustering algorithm was remodeled through constructing feature divergence.Finally,the modified algorithm was tested with one test image and two high resolution spaceflight images.The segmentation results illustrate that the proposed algorithm could improve the capability in dealing with these problems prominently.更多还原