基于扩散距离的SURF特征图像匹配算法

Images matching algorithm based on diffusion distance of SURF

针对高维特征向量存在的使用传统欧氏距离计算最近邻匹配正确率低的问题,文章提出了一种基于SURF和扩散距离的图像匹配算法。首先用Fast Hessian检测子进行特征点检测,生成SURF特征描述向量,然后利用扩散距离代替欧氏距离进行匹配,使用随机抽样一致从候选匹配中排除错误的匹配。实验证明该算法提高了SURF算法匹配的正确率,并在图像形变、光照变化方面具有较高的鲁棒性。

In this paper,an image matching algorithm based on speeded-up robust features（SURF）and diffusion distance is proposed in view of the low calculation accuracy of nearest neighbor matching of high-dimensional feature vector using the Euclidean distance.In this algorithm,Fast Hessian detection is used to find features,and the feature vector of SURF descriptors is generated.Then a SURF matching algorithm based on diffusion distance is proposed which replaces the Eculidean distance with the diffusion one.And the random sample consensus（RANSAC）is presented to exclude the mismatching points.The experimental results show that the algorithm can improve the matching accuracy of SURF algorithm,and has higher robustness in image deformation and illumination change than the traditional one.