GLF-CR: SAR-enhanced cloud removal with global–local fusion

The challenge of the cloud removal task can be alleviated with aid Synthetic Aperture Radar (SAR) images that penetrate cover. However, large domain gap between optical and SAR as well severe speckle noise may cause significant interference in SAR-based removal, resulting performance degeneration. In this paper, we propose a novel global–local fusion based (GLF-CR) algorithm to leverage complementary information embedded images. Exploiting power promote entails two aspects. first, global fusion, guides relationship among all local windows maintain structure recovered region consistent remaining cloud-free regions. second, transfers image corresponds cloudy areas generate reliable texture details missing regions, uses dynamic filtering alleviate degradation caused by noise. Extensive evaluation demonstrates proposed yield high quality outperform state-of-the-art algorithms gain about 1.7 dB terms PSNR on SEN12MS-CR dataset.