一种基于多分辨分析的简化的分裂-合并图像分割算法  

A simplified split-merge image segmentation method based on multi-resolution analysis

作　　者：刘国英[1] 傅明[1] 唐贤瑛[1] 孟爱国[1] 

机构地区：[1]长沙理工大学计算机与通信工程学院,湖南长沙410076

出　　处：《长沙理工大学学报（自然科学版）》2006年第4期77-80,共4页Journal of Changsha University of Science and Technology:Natural Science

基　　金：湖南省教育厅科研资助项目(03C083)

摘　　要：为了减少分裂-合并算法的计算复杂性,提出了一种基于多分辨率分析的分裂-合并简化算法.首先,将原始图像用多分辨的形式分层表示,对最低层分辨率的图像用简化的分裂-合并算法进行图像分割,同时用该层的边缘信息对分割结果进行优化,得到该层的分割图像;然后,用直接影射的方法将低分辨率的分割图像映射到高分辨率空间中,并用相应的边缘信息进行优化,直到在原始分辨率空间完成上述工作,得到最终的分割结果.试验证明,所提算法简单有效,较好地解决了分裂-合并算法计算复杂的问题.In order to induce the computation of split-merge algorithm, a new method was presented for image segmentation of simplified split-merge algorithm based on multi-resolution. First, the initial image to be segmented is described in the form of multi-resolution, then the lowest resolution level image is segmented by simplified split-merge method, and at the same time the segment result is refined by the edge information of the image on the same resolution level, after these the segmented image on this level is obtained. Then, the segmented image on the lower level is projected into a higher level by directly project method, and the project result is also refined by the edge information of the image on the corresponding level. Once the segmentation and refining finish on the initial resolution level, the segment work will be completed. Experiments have shown that our proposed method is simple and effective, it can solve the computation problems of split-merge method.

关 键 词：多分辨分析 分裂-合并算法 图像分割 小波变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]