基于RVCF的大视角差异图像匹配方法

文章针对 SIFT算法对大视角变化图像匹配效果差的问题,提出一种改进的 SIFT算法——抗视角变化特征 提摘取算法(Resistance to Viewpoint Change Feature,RVCF)。首先,利用最大稳定极值区域(MSER)算法提取图像中 的仿射不变特征区域；然后,利用协方差矩阵将 MSER检测出的椭圆区域转化成符合尺度空间条件的圆形区域；最 后,利用 SIFT算法对获得的圆形区域进行仿射不变特征的提取与匹配。实验结果表明:RVCF算法能够在大的视 角和尺度变化下成功实现图像间的匹配。

A Modified RVCF Algorithm for Matching Images with Different Viewing Angle

In order to solve the problem of invalidation of SIFT algorithm when viewpoints had been changed with large af.fine angle, in this paper a M-SIFT algorithm that the RVCF was proposed. Firstly, the algorithm detected affine invariantfeature regions by MSER algorithm. Then covariance matrix was used to transform anisotropic patches into isotropic patch.es by rotating and squeezing to meet the condition of scale-space theory. Finally, the affine invariant key points on isotro.pic patches were detected and matching using SIFT algorithm. Experiments showed that RVCF worked well with large af. fine angle changes and scale changes.