摘要
针对尺度不变的二值化角点(BRISK)算法抗噪性能较低,未充分利用图像的边缘的问题,提出了一种基于非线性尺度空间的图像配准技术。该算法在构造尺度空间的时候采用非线性滤波器构造图像非线性尺度空间,同时利用AGAST算法在构建的非线性空间里提取具有显著特征的角点,结合旋转不变性的BRIEF算法构造128位的二值化描述子,采用汉明距离匹配描述子。实验结果表明,该算法能大幅度提高关键点的提取质量,获得了较好特征点重复检测率,增强了特征点鲁棒性以及提高了描述子的匹配率。
In allusion to the problem of that the lower anti-noise performance and the indistinct boundaries and smoothed details of an image have existed with the BRISK algorithm, a new image registration technology based on nonlinear scale space is put forward in this paper. While the scale space is structured by applying the algorithm, the nonlinear filter is used for constructing the nonlinear scale space of the image; meanwhile the comers of having the obvious features are extracted from the constructed nonlinear space by utilizing the AGAST comer detection operator, that combines with the rotation invariant BRIEF algorithm to construct the 128-bit binaryzation descriptor and Hamming distance matching descriptor is adopted. The experimental result shows that the extracting quality of the key points can be improved substantially, the higher duplicate detection rate of the feature points is obtained, the robustness of the feature points is enhanced and the matching ratio of the descriptor is raised by using the algorithm.
出处
《计算机光盘软件与应用》
2014年第8期72-74,共3页
Computer CD Software and Application
基金
国家自然科学基金资助项目-高效稳健的自适应绝对偏度滤波算法研究(项目编号:F010305)
