基于区域检测的多聚焦彩色图像融合  

Multi-focus Color Image Fusion Based on Region Detection

作　　者：向昌成[1] 

机构地区：[1]阿坝师范高等专科学校计算机科学系,四川郫县611741

出　　处：《重庆师范大学学报（自然科学版）》2010年第4期80-83,共4页Journal of Chongqing Normal University:Natural Science

基　　金：四川省教育厅自然科研基金(No.07ZC002)

摘　　要：针对在空域中基于分块的区域检测的多聚焦图像融合算法存在着对清晰部分的提取不完全的问题,对图像边缘部分无法准确提取的缺陷,本文提出一种基于非规则区域检测的彩色图像融合方法。首先将待融合的彩色图像由RGB模式转换到HSV模式,并且比较其对应的V分量图像像素点的邻域内方差的大小,将较大者标识为清晰像素,较小者标识为模糊像素;然后在此基础上根据像素邻域内的统计特征得到具有连续性和不规则性的图像的清晰区域;最后,根据多幅待融合图像的的清晰区域重构得到融合结果图像。实验结果表明,与基于分块的区域检测相比,本文所提出的方法能更为准确的提取图像的清晰部分。对比于传统的小波方法,从主观的视觉效果和客观数据分析来看,本文算法是有效的。For most multi-focus image fusion algorithms which based on block region detection identify the clear region inaccurately, and also distinguish the image edge from source image less effectively. This paper puts forward a kind of color image fusion method based on irregular region detection. First, the RGB images are transformed into HSV images. Comparing the variance of the pixel neighborhood in the V component images, the bigger is regarded as clear pixel. On the contrary, the smaller is regarded as ambigulous pixel, then clear region is obtained based on the statistical characteristics according to the image, finally, the result image is reconstructed with the clear region of the source images. Experimental results show that the proposed method, comparing to the block region detection, can effectively distinguish the clear region from source image, and solve the problem about multi-focus color image fusion more effectively.

关 键 词：区域检测 彩色图像 清晰像素 方差 多聚焦 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]