Super-Resolution Sea Surface Temperature From Multiple Thermal Infrared Bands Using Residual Channel Attention Network

Super-resolution (SR) techniques are commonly used to obtain high-resolution (HR) sea surface temperature (SST) data. However, existing SST SR methods require collocated optical and thermal infrared images as inputs, making them completely unusable at night. This letter presented a novel SST SR method that uses two thermal infrared sensors (TIRSs) by the residual channel attention network (RCAN). The process is tested by 1 km-resolution Sentiel-3/SLSTR and Terra/MODIS data to obtain 250 m SST data. Evaluation results indicate that this method outperforms the scheme that uses visible light data as input in terms of better metrics in root mean square error (RMSE) of $0.12~^{\circ }$ C, structural similarity index measure (SSIM) of 0.784, and peak signal-to-noise ratio (PSNR) of 22.35 dB. The case studies also demonstrate its capability for night-time applications.