基于多尺度各向异性高斯核主方向角度变化的角点检测

A multi-scale corner detection via angle difference of principle directions of anisotropic Gaussian derivatives

角点检测在图像分析和计算机视觉领域有着及其重要的作用。各向异性高斯方向导数(AGDDs)可以很好的提取不同方向的图像局部灰度变化信息,并且具有很强的噪声鲁棒性,在对不同尺度下的轮廓信息和各向异性高斯方向导数进行研究后,将两者相结合提出了一种基于多尺度各向异性高斯核主方向角度变化的角点检测算法。该算法可以有效的降低角点检测的误检率。实验采用了两幅标注了真实角点测试图,在无噪声和加入不同等级噪声的情况下,对该算法和其它三种经典的角点检测算法从角点检测的能力和角点定位的精度进行对比。实验表明,研究所得的算法对于角点的检测具有更好的噪声稳健性和更低的误检率。

Corner detection plays an important role in image analysis and computer vision. Anisotropic Gaussian directional derivatives(AGDDs) can extract the local gray change information of images in different directions, and has strong noise robustness. After studying the contour information and anisotropic Gaussian directional derivatives in different scales, a corner detection algorithm based on multi-scale anisotropic Gaussian kernel principal direction angle change is proposed. The algorithm can effectively reduce the false detection rate of corner detection. Two real corner test images are used in the experiment.In the case of no noise and adding different levels of noise, the corner detection ability and corner positioning accuracy of the algorithm are compared with other three classical corner detection algorithms. Experimental results show that the proposed algorithm has better noise robustness and lower false detection rate for corner detection.