Wavelet Integrated Convolutional Neural Network for Thin Cloud Removal in Remote Sensing Images

Cloud occlusion phenomena are widespread in optical remote sensing (RS) images, leading to information loss and image degradation causing difficulties subsequent applications such as land surface classification, object detection, change monitoring. Therefore, thin cloud removal is a key preprocessing procedure for RS has great practical value. Recent deep learning-based methods have achieved excellent results. However, these common problem that they cannot obtain large receptive fields while preserving detail. In this paper, we propose novel wavelet-integrated convolutional neural network (WaveCNN-CR) images can larger without any loss. WaveCNN-CR generates cloud-free an end-to-end manner based on encoder–decoder-like architecture. the encoding stage, first extracts multi-scale multi-frequency components via wavelet transform, then further performs feature extraction each high-frequency component at different scales by multiple enhanced modules (EFEM) separately. decoding recursively concatenates processed low-frequency scale, feeds them into EFEMs extraction, reconstructs high-resolution inverse transform. addition, designed EFEM consisting of attentive residual block (ARB) gated (GRB) used emphasize more informative features. ARB GRB enhance features from perspective global local context, respectively. Extensive experiments T-CLOUD, RICE1, WHUS2-CR datasets demonstrate our significantly outperforms existing state-of-the-art methods.