基于图像灰度变化与像素强度比较的角点检测算法  

Corner Detection Algorithm Based on Image GrayLevel Change and Pixel Intensity Comparison

作　　者：苏玟萱 任劼[1] 章为川 SU Wenxuan;REN Jie;ZHANG Weichuan(School of Electronics and Information,Xi’an Polytechnic University,Xi’an 710048,China;School of Electronic Information and Artificial Intelligence,Shaanxi University of Science and Technology,710016,China)

机构地区：[1]西安工程大学电子信息学院,陕西西安710048 [2]陕西科技大学电子信息与人工智能学院,陕西西安710016

出　　处：《长江信息通信》2025年第2期77-82,共6页Changjiang Information & Communications

基　　金：陕西省教育厅重点项目(23JY029)资助;陕西高校青年创新团队支持;中国博士后科学基金(2022M712555)资助。

摘　　要：文章提出了一种基于图像灰度变化与像素强度比较的创新性角点检测算法,旨在克服单一利用这两种方法的局限性。该算法结合了高斯方向导数滤波器与像素比较策略,前者在光照和视角变化下提供良好的检测精度,但对图像旋转较为敏感,而后者在尺度变化上表现出色但对噪声、光照和视角变化的鲁棒性不足。为了解决这些问题,本文设计了一种优化的组合策略,有效融合了两种方法的优点,从而在复杂场景中实现更高的鲁棒性和检测精度。大量实验结果表明,所提算法在仿射变换、JPEG压缩及高斯噪声条件下具有优越的重复性和定位精度。在HPatches数据集的图像匹配实验中,算法在视角和光照变化下展现出显著优势,证明了其在实际应用中的广泛适用性。In this paper,we propose an innovative corner detection algorithm based on image gray level variation and pixel intensity comparison,aiming to overcome the limitations of using either method alone.The algorithm combines a Gaussian directional derivative filter with a pixel comparison strategy.The former provides high detection accuracy under illumination and viewpoint variations but is more sensitive to image rotation,while the latter performs well on scale variations but is not robust enough to noise,illumination and viewpoint variations.In order to solve these problems,this paper designs an optimized combination strategy that effectively combines the advantages of the two methods to achieve higher robustness and detection accuracy in complex scenes.Numerous experimental results show that the proposed algorithm has superior repeatability and localization accuracy under affine transform,JPEG compression and Gaussian noise conditions.In the image matching experiments on the HPatches dataset,the algorithm demonstrates significant advantages under changes in viewing angle and illumination,proving its wide applicability in practical applications.

关 键 词：高斯方向导数滤波器 像素比较 角点检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]