基于边缘概率的层次交互式图像分割算法  

A novel hierarchical interactive image segmentation based on edge probability

作　　者：薛丽霞[1] 孙伟[1] 汪荣贵[1] 杨娟[1] 胡敏[1] XUE Lixia;SUN Wei;WANG Ronggui;YANG Juan;HU Min(School of Computer and Information, Hefei University of Technology, Hefei 230601, China)

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230601

出　　处：《合肥工业大学学报（自然科学版）》2020年第3期342-348,共7页Journal of Hefei University of Technology：Natural Science

基　　金：国家自然科学基金资助项目(61672202)。

摘　　要：在基于区域合并的分割算法中,超像素分割算法过分割现象严重,不利于后续区域合并操作。文章提出了一种新颖的基于边缘概率的层次交互式图像分割算法。首先,利用一种性能优异的边缘检测算子(structured edge,SE)作为输入,得到边缘概率图,并作用于超像素分割算法进行超像素的提纯,相比于原始的超像素分割算法,该方法产生的超像素数量较少;其次,选用归一化的颜色直方图对超像素进行特征描述;最后,采用区域邻接图(region adjacency graph,RAG)和最近邻图(nearest neighbor graph,NNG)实现快速、高效的层次区域合并。仿真实验结果表明,该文方法的鲁棒性、分割质量更具有优势。In the interactive image segmentation algorithm based on region merging,the over-segmentation of superpixel segmentation algorithm is serious,which is not conducive to following merging process.In this paper,a novel hierarchical interactive image segmentation method based on edge probability is proposed.Firstly,a structured edge(SE)detector is used to obtain the edge probability map,which is applied to superpixel segmentation algorithm.Compared with the original superpixel segmentation algorithm,the proposed method refines superpixels and produces less number of superpixels.Secondly,the normalized color histogram is used to describe superpixels.Finally,region adjacency graph(RAG)and nearest neighbor graph(NNG)are used to realize hierarchical region merging in a fast and efficient way.The results of simulation experiments demonstrate the effectiveness and robustness of the proposed method.

关 键 词：交互式图像分割 超像素 边缘概率 区域合并 区域邻接图(RAG) 最近邻图(NNG) 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]