基于改进SURF算法的图像拼接研究  

Research on image mosaic based on improved SURF algorithm

作　　者：徐启文 唐振民[1] 姚亚洲 Xu Qiwen;Tang Zhenmin;Yao Yazhou(School of Computer Science and Engineering,Nanjing University of Scienceand Technology,Nanjing 210094,China)

机构地区：[1]南京理工大学计算机科学与工程学院,江苏南京210094

出　　处：《南京理工大学学报》2021年第2期171-178,共8页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(61976116);中央高校基本科研业务费专项资金(30920021135)。

摘　　要：快速鲁棒特征(Speeded-up robust feature,SURF)算法在图像匹配、模式识别、图像拼接等众多领域有着广泛的应用。随着摄像机的更新换代,照片分辨率逐渐提升,传统的SURF算法已经无法满足图像拼接的效率要求;针对以上问题,该文提出了一种具有动态阈值的改进SURF算法,该算法依据图像位置的相关性,生成用于规划拼接区域的动态阈值,利用该阈值缩小特征提取和匹配的有效区域,从而提升算法的执行效率。针对传统的渐进渐出图像融合算法失真严重的问题,该文提出了一种新的非线性权重模型,利用该模型,有效降低了拼接图像的重影现象,提升了视觉效果。Speeded-up robust feature(SURF)has a wide range of application in many fields such as image matching,pattern recognition and image stitching.With the updating of cameras,the resolution of photos is gradually improved,and the traditional SURF algorithm has gradually failed to meet the efficiency requirements of image stitching.In view of the problems,this paper proposes an improved SURF algorithm with a dynamic threshold,which generates a dynamic threshold for planning the stitching area based on the correlation of the image position,and uses this threshold to narrow the effective area for feature extraction and matching,thereby improving the execution efficiency of the algorithm.Aiming at the problem of serious distortion of the traditional linear weighted image fusion algorithm,this paper proposes a new nonlinear weight model.Using this model,the ghosting phenomenon of the stitched image is effectively reduced and the visual effect is improved.

关 键 词：快速鲁棒特征算法 图像拼接 动态阈值 图像融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]