Cloud Removal with SAR-Optical Data Fusion and Graph-Based Feature Aggregation Network

In observations of Earth, the existence clouds affects quality and usability optical remote sensing images in practical applications. Many cloud removal methods have been proposed to solve this issue. Among these methods, synthetic aperture radar (SAR)-based more potential than others because SAR imaging is hardly affected by clouds, can reflect ground information differences changes. While used as auxiliary for may be blurred noisy, similar non-local spectral electromagnetic features cannot effectively utilized traditional methods. To overcome weaknesses, we propose a novel method using SAR-optical data fusion graph-based feature aggregation network (G-FAN). First, cloudy contemporary are concatenated transformed into hyper-feature maps pre-convolution. Second, inputted G-FAN reconstruct missing cloud-covered area aggregating backscattering image, neighborhood non-neighborhood pixels image. Finally, post-convolution long skip connection adopted final predicted cloud-free images. Both qualitative quantitative experimental results from simulated real experiments show that our outperforms deep learning removal.