基于多方向DTCWT和域自适应学习的遥感图像薄云去除

Thin Cloud Removal from Remote Sensing Images Using Multidirectional Dual Tree Complex Wavelet Transform and Transfer Least Square Support Vector Regression

如果遥感图像被薄云污染,将会直接影响图像判读,现给出一种多方向DTCWT分解结合迁移LSSVR低频学习的遥感图像薄云去除算法,对遥感图像进行多方向多尺度分解,再对包含薄云信息的源图像低频系数值进行预测,并增强包含地物信息的高频部分,最终去除含云图像上的薄云。实验结果表明,该方法有助于保持含云图像的地物细节信息,并有效添加多源多时相图像的地物轮廓信息,可以实现较好的薄云去除。

If remote sensing images suffer from thin cloud pollution,image interpretation will be affected.A thin cloud removal algorithm for cloud-contaminated remote sensing images is proposed by combining a multidirectional dual tree complex wavelet transform(M-DTCWT)with domain adaptation transfer least square support vector regression(T-LSSVR).First,M-DTCWT is used to decompose remote sensing images into multiscale and multidirectional sub-bands.Then the low frequency sub-band coefficients of the cloud-free regions on target images and source domain images are used as samples for a T-LSSVR model,which can be used to predict those of the cloud regions on cloud-contaminated images.Finally,by enhancing the high-frequency coefficients and replacing the low-frequency coefficients,the thin clouds on cloudcontaminated images are removed.Experimental results show that our method could keeping the details of the ground objects of cloud-contaminated images and can effectively learn the contour information from multisource and multitemporal images,therefore,the proposed method achieves a good effect of thin cloud removal.