Can the Structure Similarity of Training Patches Affect the Sea Surface Temperature Deep Learning Super-Resolution?

Meso- and fine-scale sea surface temperature (SST) is an essential parameter in oceanographic research. Remote sensing efficient way to acquire global SST. However, single infrared-based microwave-based satellite-derived SST cannot obtain complete coverage high-resolution simultaneously. Deep learning super-resolution (SR) techniques have exhibited the ability enhance spatial resolution, offering potential reconstruct details of fields. Current SR research focuses mainly on improving structure model instead training dataset selection. Different from generating low-resolution images by downscaling corresponding images, high- are derived different sensors. Hence, similarity patches may affect and, consequently, reconstruction. In this study, we first discuss influence selection performance, showing that determined index (SSIM) 0.6 can result higher reconstruction accuracy better image quality. addition, practical stage, between input objective output a key factor for SR. Moreover, obtained actual AMSR2 MODIS more suitable because skin sub-skin difference. Finally, accuracies models relatively consistent, yet differences reconstructed quality rather significant.