基于AKAZE算法的图像拼接研究  

Research of image stitching based on AKAZE algorithm

作　　者：闫璠 张莹[1] 高赢[1] 涂勇涛[1] 张东波[1] Yan Fan Zhang Ying Gao Ying Tu Yongtao Zhang Dongbo(College of Information Engineering, Xiangtan University, Xiangtan 411105, China)

机构地区：[1]湘潭大学信息工程学院,湘潭411105

出　　处：《电子测量与仪器学报》2017年第1期36-44,共9页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61175075);湖南省教育厅重点项目(14A137)资助

摘　　要：针对基于KAZE特征检测的图像拼接算法实时性问题,提出一种简单有效的AKAZE拼接算法。该算法首先通过AKAZE算法提取图像特征点,接着计算M-LDB描述符从而生成特征向量。随后计算特征向量之间的汉明距离,提取出匹配的特征点对,然后利用RANSC算法估算全局单应性矩阵,根据动态线性变换算法求取重叠区域局部投影关系,结合两者统一投影平面,最后利用加权融合实现两幅图像的拼接。对KAZE、SIFT、SURF、ORB、BRISK进行性能实验比较,所用算法不仅对于高斯模糊、角度旋转、尺度变换和亮度变化等情况下保持良好的性能,而且处理时间大大缩短,实现了有效的图像拼接。To solve the time-consuming problem of image stitching algorithm based on KAZE, a simple and effective image stitching algorithm based on AKAZE is proposed. Firstly, AKAZE feature points are extracted. Secondly, feature vectors are construeted using the M-LDB descriptor and matched by computing the Hamming distance. Thirdly, wrong matches are eliminated by RANSAC and the global homography transform, and then a local projection transform is estimated using moving direct linear transformation in the overlapping regions. The image registration is achieved by combining the two transforms. Finally, the weighted fusion method fuses the images. A performance comparison test can be conducted aiming at KAZE, SIFT, SURF, ORB, BRISK. The experimental results show that the proposed algorithm has better robustness for the various transform, and the processing time is greatly reduced.

关 键 词：AKAZE算法 特征检测 特征匹配 图像拼接 

分 类 号：TN911.7[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]