基于空间纹理相似性的图像角点特征匹配算法

本文针对传统图像角点特征匹配算法的匹配速度慢且准确率低等问题，提出一种基于空间纹理相似性的图像角点特征匹配算法。首先，计算图像目标上角点对应的空间距离矩阵；然后，通过计算图像角点的空间距离矩阵在对应角点邻域LBP特征向量上的瑞利商，将角点在图像灰度特征空间内的度量问题转换为纹理特征空间内幅值的度量问题；最后，根据角点对应的瑞利商的大小，实现不同图像间的角点特征匹配。对不同条件下采集的图像进行角点特征匹配，得到的匹配结果表明本文算法不仅能够很好的适应图像光照、几何变化，得到的匹配正确率较高，同时与传统算法相比本文算法在运行时间上也有大幅度的降低，当处理特征数量较小时平均降低48ms，而匹配特征数量较多时能够降低2408ms。

Image corner matching algorithm using the similarity of spatial texture

This paper proposes a novel image corner matching algorithm using the similarity of spatial texture aiming at addressing the low matching rate and long computational time of traditional image corner matching algorithms. First, the paper calculated the spatial distance matrix of the corners in the image objects. Second, it transformed the measure of the image corners into the measure of spatial texture amplitudes by calculating the rayleigh quotient of the spatial distance matrix in the LBP feature space. Finally, it matched the image corners between different images by comparing their corresponding rayleigh quotients. This paper carried out the corner matching between different images captured under different circumstances. The experimental results demonstrate that our proposed image feature matching algorithm is robust on the image transformation and produces higher matching rate with less computational time. Compared with the-stated-of-art corner matching algorithm, the computational time is decreased by 48ms and 2408ms when the calculated features numbers are low and high respectively.