摘要
针对遥感影像薄云去除过程中丢失影像细节及薄云去除不彻底等问题,本文提出了一种基于小波变分法去除遥感影像薄云的方法。首先,对含薄云的遥感影像进行小波分解,可以使低层低频分量中的局部瞬变信号(不具备低秩性的有用信息)被分解到高频分量中;其次,考虑第五层低频分量中有用信息具有低秩性及薄云具有稀疏性特点,建立一种能够从第五层低频分量中分离出薄云的影像恢复正则化模型,并通过交替方向乘子算法解求出去除薄云后的第五层低频分量及薄云影像,从而达到去除薄云的目的;最后,通过小波重构,获得去除薄云后的影像。通过与其他算法比较,本文算法不仅可以有效地去除薄云,而且能够很好地保留影像的有用信息,去除薄云后的影像质量得到明显提高。
There are some problems in the process of removing thin cloud in remote sensing images,such as the large loss of image details,in complete removal of thin cloud and so on.According to these problems,we propose an effective method to remove the thin cloud without destroying the image details.First,we conduct five-level wavelet decomposition on the GF-2 image which makes the local transient signal in the low frequency component(useful signal without low rank)decompose into the high frequency component;Second,we construct an image recovery regularization model,which contains the regularized item of the low frequency component and the regularization term of the thin cloud from the angle of low rank resistance of the low frequency component and the sparse characteristics of thin cloud,we solve the low frequency component after removing thin cloud by the alternating direction method of multipliers(ADMM)algorithm;Finally,we obtain the de-cloud image by wavelet reconstructing.Experimental results show that the proposed method can not only effectively remove the thin cloud,but also preserve the image details very well.The quality of image is improved obviously.
作者
郭东升
GUO Dongsheng(Tieling Natural Resources Affairs Service Center,Tieling 112000,China)
出处
《测绘与空间地理信息》
2022年第3期120-125,共6页
Geomatics & Spatial Information Technology
关键词
正则化项
GF-2影像
薄云
低频分量
小波分解
regularization model
GF-2
thin cloud
low frequency component
wavelet decomposition
