基于多尺度Gabor滤波器的角点检测

Corner detection using multi-scale Gabor filters

为了克服在不同图像上的尺度选择问题,提出了一种基于边缘轮廓线的多尺度Gabor滤波器的角点检测算法。该算法首先利用Canny边缘检测算子提取图像的边缘轮廓;进而用一组构建好的4个尺度8个方向Gabor滤波器的虚部对图像进行平滑,并计算每个像素在其相同尺度下各个方向上Gabor滤波器虚部响应的归一化的和;最后将每个边缘像素点在所有尺度下的乘积作为新的角点测度,当角点测度大于预设阈值时,则认定该点为角点。将实验结果与经典的Harris、CPDA和He&Yung角点检测算法进行比较,提出的算法在检测准确率、定位误差、噪声稳健性性能指标上,都取得了更好的结果。

A contour-based corner detector using multi-scale Gabor filters is proposed,which improves the selection of scale in different images.Firstly,the edge contours are extracted by Canny detector.Secondly,a set of Gabor filters including four scales and eight directions are constructed to smooth the input image,and then compute the sum of the normalized magnitude responses at each direction.Thirdly,at each edge pixel,the product of the normalized sum at all scale is defined as corner measure.When the corner measure is above a previously specified threshold,this contour pixel will be labeled as a corner.Finally,the proposed corner detector is compared with three classic corner detectors including Harris detector,CPDA detector and He Yung detector.The results show that the proposed detector has the better consequence in detection accuracy,localization accuracy and noise-robustness.