利用小波和分层聚类进行彩色图像分割

Image Segmentation Based on Wavelet and Clustering

提出了一种有效的基于颜色和位置相关信息的图像分割方法。图像以块为单位进行划分,在YUV空间,提取块的颜色特征,并对每小块进行小波变换,获得其小波能量特征。将颜色特征和能量特征作为每小块的特征向量,采用分层聚类的方法进行区域的合并。该方法在聚类过程中考虑了小块的位置关系,并能最终自动确定聚类的类别数。

This text proposes an effective method for image segmentation based on the color and position relevant information. An image is divided into blocks with pixels, and In YUN color space, a feature vector which includes color and wavelet energy by wavelet transformation is extracted for each block, then the method of layered clustering incorporates those blocks. This method considers the position relevant information of those blocks in the course of clustering and, finally, determines the number of sort automatically.