DBSD-Net boosts sea surface temperature resolution with dual-branch AI

Sea surface temperature (SST) is a critical climate indicator, but satellite imagery often lacks the resolution needed to capture fine thermal features like ocean fronts. A new paper on arXiv introduces DBSD-Net, a Dual-Branch State-Displacement Network designed specifically for SST super-resolution. The architecture uses two parallel branches: one applies discrete wavelet transform to explicitly separate low- and high-frequency components, while the other—dubbed VGGUNet—extracts multi-scale semantic features from a frozen pre-trained VGG backbone. This combination lets the model preserve global context while recovering sharp, localized thermal gradients.

Inside the wavelet branch, the authors introduce a Structural State Space Module (SSSM) with a Gated Structure Refinement (GSR) unit to efficiently model long-range dependencies without the computational overhead of attention. A Displacement Gate Module (DGM) learns a displacement field for geometry-aware modulation of high-frequency details, which helps handle spatially varying degradation common in satellite data. Experiments across multiple public SST datasets show DBSD-Net consistently outperforms current state-of-the-art methods, making it a promising tool for climate research, oceanography, and environmental monitoring. The paper is accepted for publication in IEEE JSTARS.