GLF-CR: SAR-Enhanced Cloud Removal with Global-Local Fusion

The challenge of the cloud removal task can be alleviated with the aid ofSynthetic Aperture Radar (SAR) images that can penetrate cloud cover. However,the large domain gap between optical and SAR images as well as the severespeckle noise of SAR images may cause significant interference in SAR-basedcloud removal, resulting in performance degeneration. In this paper, we proposea novel global-local fusion based cloud removal (GLF-CR) algorithm to leveragethe complementary information embedded in SAR images. Exploiting the power ofSAR information to promote cloud removal entails two aspects. The first, globalfusion, guides the relationship among all local optical windows to maintain thestructure of the recovered region consistent with the remaining cloud-freeregions. The second, local fusion, transfers complementary information embeddedin the SAR image that corresponds to cloudy areas to generate reliable texturedetails of the missing regions, and uses dynamic filtering to alleviate theperformance degradation caused by speckle noise. Extensive evaluationdemonstrates that the proposed algorithm can yield high quality cloud-freeimages and outperform state-of-the-art cloud removal algorithms with a gainabout 1.7dB in terms of PSNR on SEN12MS-CR dataset.