一种基于改进SIFT的图像匹配融合算法  

An Image Matching and Fusion Algorithm Based on Improved SIFT

作　　者：肖剑 刘超[1] 于帅 XIAO Jian;LIU Chao;Yu Shuai(School of Spatial Information and Geomatics Engineering,Anhui University of Science and Technology,232001,Huainan,Anhui,PRC;Zhejiang Taihe Land Survey and Planning Co.,Ltd.,324000,Quzhou,Zhejiang,PRC)

机构地区：[1]安徽理工大学空间信息与测绘工程学院,安徽淮南232001 [2]浙江泰和土地勘测规划有限公司,浙江衢州324000

出　　处：《江西科学》2024年第6期1263-1270,共8页Jiangxi Science

基　　金：安徽省教育厅科学研究重点项目(2022AH050849)。

摘　　要：图像匹配与融合是图像拼接的核心环节,分别影响着图像拼接后的精确度和自然度。尺度不变特征转换SIFT(Scale Invariant Feature Transform)因其具有旋转不变性和良好的鲁棒性广泛应用于图像处理领域,但会导致特征提取匹配耗时长、误匹配等问题。鉴于此,提出了一种结合SIFT和AGAST算法的新方法,通过KNN和RANSAC进行特征点提纯,最后通过改进的加权平均法进行融合拼接。结果表明,该方法可以极大地提升特征提取效率和匹配效率,图像融合效果较好。Image matching and fusion are the core processes of image stitching,affecting both the accuracy and naturalness of the stitched images.The Scale Invariant Feature Transform(SIFT)has been widely applied in the field of image processing due to its rotational invariance and robustness.However,it can lead to issues such as time-consuming feature extraction and mismatches.To address these issues,a new method combining SIFT and AGAST algorithms is proposed.Feature points are refined through KNN and RANSAC,and the fusion and stitching are completed using an improved weighted average method.The results demonstrate that this method significantly enhances the efficiency of feature extraction and matching,achieving better image fusion effects.

关 键 词：SIFT 特征提取 图像匹配 图像融合 图像拼接 计算机视觉 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]