测绘学报

• 学术论文 •    

一种基于频谱段能量的高分辨率遥感图像边缘特征检测方法

吴桂平1,肖鹏峰1,冯学智2,王珂3,黄秋燕1   

  1. 1. 南京大学
    2. 南京大学城市与资源学系
    3. 南京大学地理与海洋科学学院地理信息科学系
  • 收稿日期:2010-09-06 修回日期:2010-11-16 出版日期:2011-10-25 发布日期:2011-10-25
  • 通讯作者: 吴桂平

A Method of Edge Feature Detection from High-resolution Remote Sensing Images Based on Frequency Spectrum Zone Energy

  • Received:2010-09-06 Revised:2010-11-16 Online:2011-10-25 Published:2011-10-25

摘要: 从图像信号的能量角度出发,探讨了一种基于频谱段能量的高分辨率遥感图像边缘检测的方法。该方法采用滑动窗口傅立叶变换和频谱能量分段叠置的手段,将含有噪声的图像信号分解到不同的频谱段中,然后根据直流中心频谱段图像良好的噪声抑制特性,在此基础上进一步分析其高阶频谱段能量对图像边缘特征的贡献作用,将3×3的滑动子窗口中频谱半径为1-1.5范围内的频谱能量系数进行求和叠置,并量化至0–255的灰度范围,得到最终的边缘检测图像。实验结果表明,基于频谱能量所检测出的图像边缘特征响应较为显著,视觉效果较好,并且具有特定方向上的边缘检测能力。研究结果对于后期遥感图像的分割以及面向对象的遥感图像识别具有重要的理论意义和实际应用价值。

Abstract: In view of the characteristics of high-resolution remotely sensed imagery, a novel edge feature detection algorithm based on frequency spectrum zone energy was introduced in this paper. From the perspective of energy of image signal, the methods of sliding-window fourier transformation and frequency energy sectional superposition were used to decompose signal into different frequency spectrum zones. Then according to good noise suppression capacity of DC (Direct Current) central frequency spectrum zone image, contribution degree of higher-order frequency spectrum zone energy to the image edge feature was further analysed. Finally, we obtained the edge detection result through aggregating spectrum energy coefficient within each window radius of frequency spectrum range from 1 to 1.5 and quantizing image gray between 0-255. The experimental results shows that this algorithm remains good edge of image and has good capability of edge detection in a specific direction as well. The proposed method and thought has certain theoretical significance and practical reference value for image segmentation and object-oriented image recognition.