CAJ | 학술논문

提出一种基于各向异性高斯方向导数滤波器提取图像粗轮廓检测角点的方法。首先利用多方向的各向异性高斯方向导数滤波器提取图像多方向的灰度变化信息，然后利用像素间梯度相关性提取图像粗轮廓。最后在提取的图像粗轮廓的基础上构造自相关矩阵并求解其特征值，利用特征值归一化的乘积做为角点的测度。实验证明本算法具有噪声鲁棒性及角点定位准确性。
제출일충기우각향이성고사방향도수려파기제취도상조륜곽검측각점적방법。수선이용다방향적각향이성고사방향도수려파기제취도상다방향적회도변화신식，연후이용상소간제도상관성제취도상조륜곽。최후재제취적도상조륜곽적기출상구조자상관구진병구해기특정치，이용특정치귀일화적승적주위각점적측도。실험증명본산법구유조성로봉성급각점정위준학성。
This paper presents proposes a corner detector method based on anisotropic Gaussian directional derivative filters to extract the crude outline of image.Firstly,multi-direction anisotropic Gaussian directional derivative filters are used to extract the variation of gray change information,and then the crude outline is extracted by using the gradient correlation of pixels.Finally,we construct the auto-correlation matrix on the basis of the extracted crude outline and normalize eigenvalues of the auto-correlation matrix,then the product of normalized eigenvalues is served as the measure of corner extracting.Experiments show that the algorithm is noise-robust and location-accurate.