基于区域合并和图割的非监督彩色图像分割算法  

Unsupervised color image segmentation using region merging and graph cuts

作　　者：李磊[1] 金良海[1] 宋恩民[1] 

机构地区：[1]华中科技大学计算机科学与技术学院,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2013年第12期44-47,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60972098);河南省教育厅自然科学研究资助项目(13A520177)

摘　　要：提出一种区域合并和图割技术相结合的非监督彩色图像分割算法,以解决单独使用图割和基于超像素合并分割图像所产生的问题.首先,使用多尺度四元数Gabor滤波提取图像的纹理特征;然后以能量最小化为准则,对粗尺度下图像所生成的超像素进行合并,并使用α-expansion对合并结果进行优化;而后,把结果映射到原始尺度,再次使用α-expansion对其进行优化;最后,去除结果中的弱边界.实验结果表明该算法有效提高了分割质量.The purpose o{ this study is to improve graph cuts and superpixel-based region merging. A new unsupervised color segmentation method based on region merging and graph cuts was proposed. First, the multi-scale quaternion Gabor filter with more discriminating power was employed to extract texture feature of a given image. Then the superpixels obtained at the coarse scale of two-level Gaussi- an pyramid of the image were merged based on the criterion of energy minimization, and the merged result was refined by a-expansion. Then the result was mapped to the original scale and refined by a- expansion again, and the final segmentation result can be obtained automatically after removing weak edges. Experimental results demonstrate the good performance of the proposed method.

关 键 词：彩色图像分割 区域合并 图割 多尺度四元数Gabor滤波 能量最小化 高斯金字塔 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]