基于空间位置约束的K均值图像分割

K均值聚类分割是一种有效的基于聚类的图像分割算法.传统的K均值聚类分割算法采用特征空间中的相似性测度来度量像素的归属类别.由于自然景物图像的复杂性,位置邻近且本应属于同一分割区域的像素点,由于它们视觉特征的差异性,导致其在特征空间中相距甚远而被分割为不同的区域.以投票的方法将像素的局部空间位置信息引入到K均值聚类分割算法中,达到了改善分割效果的目的.实验结果证实了该方法的有效性.

An improved method of K-means image segmentation based on spatial position information

K-means clustering is an effective algorithm for image segmentation,which attempts to separate objects of interest from their background.Traditional K-means clustering algorithms use the visual similarity measures of pixels in the feature space to determine which segmentation region the pixels belong to.Because of the complexity of natural images,neighboring pixels with different visual features,which should be treated as part of the same object,may end up in separate regions.As a result,it is hard to get s...