基于自适应几何阈值的SUSAN算法改进

SUSAN Algorithm Based on Adaptive Geometric Threshold

传统的SUSAN算法用于边缘检测时采用固定几何阈值来计算边缘响应,得到的边缘线会较粗,也会出现漏检的情况。为此,在传统SUSAN算法的基础上研究改进,提出结合最大间类方差法来获得自适应几何阈值,从而对原SUSAN边缘检测算法进行改进。每幅图片根据最大间类方差法自适应获得3个几何阈值,多重阈值将图像分割成若干区域,再对各区域点进行分类和合并来获取边缘点集。实验结果表明,该边缘检测算法能够提高原来SUSAN边缘检测的精度,检测到的边缘细窄平滑,并具有更好的抗噪性。

Traditional SUSAN edge detection algorithm uses constant geometry threshold to calculate the edge response.The edge line is thick and missing would happen.The improved SUSAN edge detection algorithm is based on the Otsu＇s method.Otsu＇s method is used to get three thresholds which divide the whole image into several areas.Then the points are classified and merged in the regions to get the edge points set.Experimental results show that the edge detection algorithm can improve the accuracy of original SUSAN edge detection.The detected edge is narrow, smooth and of better resistance to noise.