摘要： 角点是图像轮廓线最重要的特征，为了准确而快速地检测角点，提出基于轮廓尖锐度的计算方法。该算法在多尺度空间中对轮廓曲线进行高斯平滑，在局部支撑区域内计算轮廓每一点的尖锐度，结合角点筛选规则确定角点。数学推导表明该算法具备合理性和可行性。实例分析表明其抗干扰性好、运算量小、定位准确。

关键词: 图像轮廓, 多尺度, 高斯平滑, 角点检测

Abstract: Corner is one of the important features of image contour. A technique for detecting corners precisely and rapidly is proposed. The technique is based on the calculation of contour sharp degree. The procedure processes the contour through Gaussian multi-scale smoothing, calculates the shape degree of each contour point within the region of local support, and determines corners by a process of corner filtering at some criteria. Rationality and feasibility of the algorithm are verified mathematically. Contrast experiments show that the new approach is robust and accurate.

Key words: image contour, multi-scale, Gaussian smoothing, corner detection

中图分类号: 

• TP311