[1]郑倩,刘珊,邓璐娟,等.基于平行四边形对角线理论的角点检测算法[J].郑州大学学报(工学版),2021,42(04):19-25.[doi:10.13705/j.issn.1671-6833.2021.02.017]
　Zheng Qian,Liu Shan,Deng Lujuan,et al.Corner detection algorithm ba＜x＞sed on parallelogram diagonal theory[J].Journal of Zhengzhou University (Engineering Science),2021,42(04):19-25.[doi:10.13705/j.issn.1671-6833.2021.02.017]
点击复制

基于平行四边形对角线理论的角点检测算法()
分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
42
期数:
2021年04期
页码:
19-25
栏目:
出版日期:
2021-07-30

文章信息/Info

Title:
Corner detection algorithm ba＜x＞sed on parallelogram diagonal theory
作者:
郑倩; 刘珊; 邓璐娟; 王强; 张世征
郑州轻工业大学软件学院;

Author(s):
Zheng Qian; Liu Shan; Deng Lujuan; Wang Qiang; Zhang Shizheng;
Zhengzhou Light Industry University School of Software;

关键词:
Keywords:
DOI:
10.13705/j.issn.1671-6833.2021.02.017
文献标志码:
A
摘要:
目的 角点检测是图像处理和计算机视觉领域的基本任务，角点响应函数构造的复杂性或对曲线进行多次平滑的操作会制约角点检测方案的检测效率。针对这一问题，本文提出了一种利用平行四边形对角线之比快速估计曲率的角点检测算法。方法 首先，利用Canny边缘检测器提取边缘轮廓线，并通过各向异性高斯方向导数滤波器对边缘线进行光滑，其次，利用提出的角点响应函数估计曲线上每个像素点的“离散”曲率，将曲率值大于设定阈值的像素作为候选角；最后，对候选角进行非极大抑制，保留精确的角点集，删除弱角点和伪角点。结果 与现有五种基于轮廓的角点检测算法相比，本文算法不需要平方根运算，在相同的图像测试集下，重复率较高，位置更加准确，而且角点检测速度约是CTAR的3倍。结论 提出了一种新的基于轮廓的角点检测算法，利用平行四边形对角线之比计算角点响应函数，避免了平方根运算，降低计算复杂度。而且，该方法不仅具有优异的角点检测性能，还对噪声具有良好的鲁棒性，并且还可用于从复杂的医学图像提取角点。
Abstract:
ob＜x＞jective Corner detection is one of the fundamental topics in image processing and computer vision. The complexity of the construction of the corner response function or the multiple smoothing of the curve often restricts detection efficiency of the corner detection scheme. Thus, in this paper we propose a novel method for image corner detection ba＜x＞sed on the diagonal of a parallelogram to estimate the curvature value. Method First, the Canny edge detector is used to extract each edge contour from the input image. Second, we smooth curves using anisotropic Gaussian directional derivative filter, and estimate the discrete” curvature of each pixel on the curve according to the corner response function proposed in this paper. And then, non-maximum suppression is applied to the candidate corner sets. Finally, the refined corner sets are retained with unstable and false corners removed. Result Compared with the existing five contour-ba＜x＞sed corner detection algorithms, the proposed algorithm does not require square root operation. The extensive experiments showed that the developed method gives the highest average repeatability and low localization error than the five other detectors, while the corner detection speed is about 3 times that of CTAR. Conclusion A new and fast contour-ba＜x＞sed corner detection algorithm is proposed, which uses ratio of the diagonals of the parallelogram (FRPD) to calculate the corner response function reducing the computational complexity. The proposed method shows excellent performance in terms of detection accuracy, noise robustness and can be applied to corner extraction from medical images
更新日期/Last Update: 2021-08-26