基于快速区域划分的多焦点图像融合方法  

Multi-focus image fusion based on fast region segmentation

作　　者：陈旸[1] 覃征[1] 胡浩[2] 宋铭越[2] 

机构地区：[1]清华大学计算机科学与技术系,北京100084 [2]清华大学软件学院,北京100084

出　　处：《清华大学学报（自然科学版）》2010年第10期1733-1737,共5页Journal of Tsinghua University(Science and Technology)

基　　金：国防"十一五"预研基金(102060206;402040202);国家自然科学基金资助项目(60673024)

摘　　要：针对传统的基于像素点和基于窗口的多焦点图像融合方法中,处于像平面的清晰物体呈现效果不佳的问题,提出了基于快速区域划分的图像融合方法。首先使用边缘提取算子对图像进行初始的轮廓提取,然后对初始轮廓中存在的非连续、非闭合的缺陷进行消除。再综合各自图像的轮廓得到联合区域划分,最后基于联合区域划分进行图像融合。实验结果表明:该方法能有效的发现处于像平面中的清晰物体,使得物体在融合结果中的呈现更清晰。与传统的基于像素点和基于窗口的多焦点图像融合方法相比,该方法融合效果更好。Deficiency of image exhibition in the image plane exists in the traditional approaches of pixel-based and window-based image fusion.A region-based image fusion method was developed using fast region segmentation.The edge detection operator was used to extract the initial contour,in which the defects of the discontinuous and non-enclosed phenomenon were later eliminated.Joint segmentation was generated through synthesizing contours of multi-images with the region-based image fusion being guided by means of the joint segmentation.Experimental results demonstrate that the proposed method effectively discovers objects lying in the image plane,making objects clearly exhibited in the fusion result and that the method also promotes the performance compared with the approaches using the pixel-based and window-based image fusion.

关 键 词：多焦点 图像融合 区域划分 轮廓提取 

分 类 号：TP311.11[自动化与计算机技术—计算机软件与理论]