基于边缘流与区域归并的彩色图像分割方法  

Color image segmentation algorithm based on edgeflow and region merging

作　　者：徐丽燕[1] 刘复昌[1] 曹国[1] 孙权森[1] 夏德深[1] 

机构地区：[1]南京理工大学计算机科学与技术学院,江苏南京210094

出　　处：《光电子．激光》2011年第10期1582-1587,共6页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金资助项目(60773172);江苏省自然科学基金资助项目(BK2008411);中国博士后科学基金资助项目(20070411055);江苏省博士后科学基金资助项目(0701037B)

摘　　要：为了克服边缘流各向异性扩散(EAD,edgeflow-driven anistropic diffusion)过分割和最小生成树(MST,min-imum spanning tree)方法计算复杂度高的缺点,提出了结合边缘流与区域归并的彩色图像分割方法。首先利用EAD方法对图像进行预分割,然后利用MST方法依据全局最优化准则对EAD的过分割区域进行归并,最后进行相应的后处理,得到最终的分割结果。这里,由于MST方法是基于EAD的过分割区域而非像素点,因此算法效率得到了很大的提高。另外,EAD方法可以有效利用图像的局部信息,而MST方法则考虑到了图像的全局特征,因此本文方法综合了两者的优点。实验结果表明,本文方法不但能够取得很好的分割效果,而且运行时间较短。In order to overcome the over segmentation phenomenon of edgeflow-driven anistropic diffusion （EAD） and the high computational complexity of （MST） minimum spanning tree, a color image segmentation algorithm based on edgeflow and region merging is presented. First of all, the EAD is applied to the image to get a preliminary result. Then the MST is used to perform the globally optimized region merging. Finally,segmentation results are achieved after proper post-process. Because the MST is based on segmented regions instead of image pixels, this algorithm requires much lower computational complexity. In addition,the GAD focuses on local details while the MST captures global property, so the proposed algorithm combines both advantages. Experimental results clearly indicate that the approach we propose can not only get satisfactory segmentation results, but also decrease the runtime of the process.

关 键 词：彩色图像 分割 边缘流 边缘流各向异性扩散(EAD) 最小生成树(MST) 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]