基于尺度限制SURF的图像配准  

Image registration based on scale restricted SURF

作　　者：詹曙[1] 王少武[1] 蒋建国[1] 

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009

出　　处：《电子测量与仪器学报》2013年第10期963-967,共5页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61174170);安徽省高校自然科学研究重点项目(KJ2010A193)

摘　　要：图像配准是数字图像处理中一个重要研究领域,是图像拼接、图像超分辨率重建的基础,基于尺度不变特征的图像配准已成为最近几年的热门研究领域,SURF算法是其中之一。对于具有尺度差异(SD)同时满足仿射变换的图像,传统SURF算法存在较大误匹配,从而影响最后的配准精度。针对误匹配问题,采用基于尺度限制(SR)的SURF算法并利用样本统计的方法来剔除误匹配,实验结果表明基于尺度限制的SURF算法能够有效的剔除误匹配,从而提高图像配准的精度。Image registration is an important research field in digital image processing, which is the basis of image stitching and image reconstruction. Recently, image registration based on scale invariant feature becomes a hot study topic, and SURF algorithm is one of hot topic. For an affine transformation and scale difference between images, the traditional SURF algorithm has a large mismatch,which affects the final registration accuracy. Aiming at the mis- match problem, in this paper, scale restrict and sample statistics are used to reject mismatch. The experimental re- suits demonstrate that this method can effectively reject mismatch and improve precision of the image registration, therefore the accuracy of image registration could be improved.

关 键 词：尺度不变 图像配准 仿射变换 SURF算法 尺度差异 

分 类 号：TN929.52[电子电信—通信与信息系统]