摘要： 提出了一种利用各向异性扩散和平均值位移聚类的彩色图像分割的算法。它主要的思想是把一幅彩色图像分成彩色通道和亮度通道,再分别对两种通道进行各向异性扩散,然后再利用平均值位移算法进行聚类分割,最后把两者的结果合并起来就可得到最后的分割结果。在这种算法中,各向异性扩散起着去噪、增强和粗分割的作用,它让一幅图中彩色信息更均匀。平均值位移在扩散的基础上,作最后的分割处理。使用CIE L*u*v*色彩空间作为图像的表示进行分割,能很有效地在彩色分量上分割出彩色区域,而彩色信息区别不明显的区域则可以通过对亮度信息分割得到。实验数据表明,它是一种有效、稳定、健壮的算法。

关键词: 图像分割, 各向异性扩散, 平均值位移

Abstract: This paper presents a new color image segmentation algorithm using anisotropic diffusion and mean shift clustering. The color image data is separated into chromaticity and brightness, which are separately diffused. The results of the two diffusions are separately clustered and segmented by means of mean shift algorithm, and their combination results in final segmented image. In this paper, anisotropic diffusion serves for denoising, enhancement, coarse segmentation, and mean shift finally segments the two channels. We apply the algorithm to the filtering of color images represented in CIE L*u*v* space. Experiences show that it can segment the color image effectively and robustly. ;;

中图分类号: 

• TP391.4