A sample and strong method for image cloud removal

Optical remote sensing imagery is at the core of many Earth observation activities. The regular, consistent and global-scale nature of the satellite data is exploited in many applications, such as cropland monitoring, climate change assessment, land-cover and land-use classification, and disaster assessment. However, optical remote sensing system are often affected by clouds during the acquisition process, resulting in low clarity of the acquired images, which affects the judgment of feature information and subsequent use, and thick cloud occlusion causes the surface information below to be completely invalid. We propose a spatial attention recurrent neural module based on an encoder-decoder network, which can solve the cloud occlusion problem. Furthermore, contexture transformer network is proposed by composed contexture extractor and correlation embedding to consider the relationship of remote dependencies in the network, and our network achieves good results on the RICE1 and RICE2 datasets.