一种基于Mean-Shift聚类算法的图像区域分割方法

A METHOD OF REGION-BASED IMAGE SEGMENTATION WITH MEAN-SHIFT CLUSTERING ALGORITHM

提出一种基于均值平移(Mean-Shift)聚类算法的图像区域分割方法。该方法首先选用适当的彩色空间对图像中的每个像素点抽取颜色、纹理及空间位置等特征,形成特征空间;然后,利用Mean-Shift聚类算法,在像素点特征空间中进行聚类,利用提出的方法,确定最佳窗口半径参数,进而确定聚类簇数、聚类中心等参数,将像素初步划归不同的组,并利用相邻像素之间的连接原理对图像区域进一步分割。分割方法提供了丰富的区域描述特征。实验结果表明这种方法具有图像分割速度快,分割效果好等特点。

A method of region-based image segmentation with Mean-shift clustering algorithm is introduced. This method first extracts with selected appropriate colour space the colour, texture, and location features from each pixel to form feature space. Then,the feature space of pixels are clustered with Mean-shift Clustering algorithm, and the optimal parameter of window radius is decided by the proposed method followed by selecting other parameters of clustering numbers and clustering centres, etc. Every pixel is primarily grouped and labelled, the regions with the same label are segmented again according to the neighbour connection theory for pixels and a lot of the features which describe the regions are provided in the segmentation. Experiment results show that this method can segment images quickly and has good segmentation results.