基于改进Canny算子高低阈值选取的遥感图像边缘检测

Remote Sensing Image Edge Detection Based on Improved Canny Operator High and Low Threshold Selection

针对传统Canny算子存在梯度算子简单、阈值选取缺乏自适应性等问题,提出一种改进的高低阈值选取算法。首先将卷积模版扩大到3×3,并对45°和135°方向上的权值进行优化;然后利用改进的Otsu法自适应地生成高低阈值;最后对局部阈值进行优化,从而完成对Canny算子的改进。实验结果表明,针对不同的影像,改进算法能自适应地生成符合局部特征的高低阈值,在纹理细节丰富和图像对比度较小的区域,能够在避免过多冗余信息的情况下最大限度地保留局部细节,改进后的方法能够为边缘提取提供参考。

Due to the traditional Canny algorithm has the problems of simple gradient operator and the threshold selection is not adaptive,an improved high and low threshold selection algorithm is proposed.First,the convolution template is expanded to 3×3,and the weight in 45°and 135°direction is optimized.Then,the improved Otsu method is used to adaptively generate high and low thresholds.Finally,the local threshold is optimized to improve the Canny operator.The experimental results show that,for different images,the improved algorithm can adaptively generate high and low thresholds consistent with local features.In areas with rich texture details and low image contrast,the local details can be retained as much as possible without too much redundant information.The improved method can provide a reference for edge extraction.