基于多尺度二阶广义高斯方向导数滤波器角点检测算法

Corner detection algorithm based on multi-scale second-order generalized Gaussian directional derivative filter

检测性能良好的角点检测算法能够应对多幅图像中的不同特征及变化,且能够很好的检测图像中的不同或相似特征。二阶各向异性高斯方向导数(SOGGDD)可以很好的提取不同方向的图像信息,并且可以很好的应对噪声干扰,研究了,将不同尺度的图像信息和二阶各向异性高斯方向导数结合提出了一种基于多尺度二阶广义高斯方向导数滤波器的角点检测算法,该算法可以有效提高检测的平均重复性,有效应对不同条件的变化。实验采用了标准数据库和角点检测文献中的28张场景图进行比较。通过对该算法和多尺度Gabor角点检测算法进行平均重复性的对比,实验表明提出的算法具有更好的角点检测性能。

Corner detection algorithm with good detection performance can deal with different features and changes in multiple images, and can well detect different or similar features in images. The second-order anisotropic Gaussian directional derivative(SOGGDD) can extract image information in different directions and deal with noise interference. The image information of different scales and the second-order anisotropic Gaussian directional derivative are studied. Through the combination of the two,a corner detection algorithm based on multi-scale second-order generalized Gaussian directional derivative filter is proposed,The algorithm can effectively improve the average repeatability of detection and effectively deal with the changes of different conditions. The experiment uses 28 scene maps in the standard database and corner detection literature for comparison. By comparing the average repeatability between the algorithm and multi-scale Gabor corner detection algorithm, the experimental results show that the proposed algorithm has better corner detection performance.