基于多尺度自适应均衡的遥感图像边缘检测方法

Remote sensing image edge detection method based on multi-scale self-adaptive equalization

针对经典边缘检测算子检测遥感图像边缘时,存在边缘不完整、轮廓不清晰等问题,利用遥感图像地物的多尺度特征,提出了一种基于多尺度自适应均衡的边缘检测方法。采用多尺度和不同方向的结构元素对灰度遥感图像进行形态学重构和梯度运算,获得多幅具有不同尺度和方向边缘特征的图像;求取每幅图像灰度值方差的权重比,利用权重比对多幅图像进行自适应均衡求和,得到边缘轮廓。实验结果表明:与经典边缘检测算子相比,所提方法能够较完整清晰地检测图像边缘,同时能有效抑制遥感图像因采集、传输而引入的椒盐噪声,具有一定的抗噪声性能。

In order to solve the problem of incomplete edge and unclear outline in remote sensing image edge detection using classical edge detection operator,an edge detection method based on multi-scale adaptive equalization is proposed using the multi-scale feature of remote sensing image ground objects. Firstly,morphology reconstruction and gradient operation of gray-scale remote sensing image is carried out by adopting multi-scale and different directional structural elements. Multi-images owning edge features of different scales and directions is obtained. Secondly,compute the weight ratio of grey value variance of each image,adaptive equalization summing is carried out using weight ratio on multi-images,acquire edge profile. The experimental results show that,comparing with classical edge detection operator,the proposed method can detect image edge more clearly and completely,and can effectively suppress the salt and pepper noise caused by the collection and transmission of the remote sensing image,and has certain anti-noise performance.