基于CenSurE-star改进BRISK图像匹配算法

An improved BRISK image matching algorithm based on CenSurE-star

针对BRISK(binary robust invariant scalable keypoints)算法匹配准确性较低,提出了一种基于CenSurE-star改进BRISK的图像匹配方法。首先检测图像CenSurE-star特征点,并构建BRISK二进制描述符,基于汉明距离相似度量方法采用KNN算法进行粗匹配筛选,并结合RANSAC(random sample consensus)算法对剩余不匹配点对进一步剔除。经实验研究对比,在关于光照、模糊、旋转、压缩比多种变化下该方法相较于BRISK算法准确性更高,保持了BRISK算法的实时性,与SIFT、SURF常见算法相比,该方法具有较高的准确性,且明显提高了匹配速度。

For the low matching accuracy of BRISK(binary robust invariant scalable keypoints)algorithm,an improved BRISK image matching method based on CenSurE-star was proposed.First detect image CenSurE-star feature points and build BRISK binary descriptor.KNN algorithm was used for rough matching screening based on the hamming distance similarity measurement method,and the random sample consensus(RANSAC)algorithm was combined to further eliminate the remaining mismatched point pairs.Compared with the BRISK algorithm,the accuracy of this method is higher than that of the BRISK algorithm under various changes of illumination,blur,rotation and compression ratio,and the real-time performance of the BRISK algorithm is maintained.Compared with the common algorithms of SIFT and SURF,this method has higher accuracy and significantly improves the matching speed.