摘要
边缘检测是图像特征提取的关键技术之一。为强化所提取的边缘特征,设计一种联合Otsu算法和多尺度细节增强的策略对传统Canny算法进行改进。首先针对高斯滤波模糊图像细节的不足,利用多尺度细节增强算法对遥感图像进行处理,增强图像的细节信息,降低高斯算法的影响;然后针对传统Canny算法的人为选择阈值带来的不足,利用Otsu算法进行自适应阈值选择,以提高算法的适应性。实验证明,在主观视觉效果以及客观评价指标下,所设计的改进方法对于不同场景的图像边缘特征增强都具备良好的效果。
Edge detection is one of the pivotal technologies for image feature extraction. To strengthen the extracted edge features, a strategy combining Otsu algorithm and multi-scale detail enhancement is designed to improve the traditional Canny algorithm. Firstly, focused on the deficiency of Gaussian filtering blurring image details, the multi-scale detail enhancement algorithms are utilized to process remote sensing images to enhance image details and reduce the influence of Gaussian algorithm. Secondly, for the deficiencies caused by artificial threshold selection of traditional Canny algorithm, the Otsu algorithm is applied for adaptive threshold selection to improve the adaptability of the algorithm. Under the subjective visual effects and objective evaluation criterions, experiments have proved that the proposed approach has a good impact on the enhancement of edge features for images in different scenes.
作者
李靖
王慧
闫科
阎晓东
杨乐
LI Jing;WANG Hui;YAN Ke;YAN Xiaodong;YANG Le(Information Engineering University,Zhengzhou 450001,China;61206 Troops,Beijing 100042,China;32316 Troops,Urumqi 830000,China)
出处
《测绘科学技术学报》
CSCD
北大核心
2021年第4期398-403,共6页
Journal of Geomatics Science and Technology
基金
中原科技创新领军人才计划资助项目(194200510023)。
作者简介
李靖(1995-),男,河南周口人,硕士生,主要研究方向为数字摄影测量。E-mail:lj_prs@163.com。
