
Frontiers in Marine Science | New and Recent Articles
An attempt at underwater image lightweight super-resolution using transformer and frequency-domain learning
This study introduces the Frequency-domain Learning Transformer (FLT), a novel approach to underwater image super-resolution (SR) that addresses the challenges of low-resolution imaging in complex underwater environments. By leveraging both spatial and frequency domain information, FLT enhances fine-grained detail reconstruction while significantly reducing computational costs. The architecture incorporates Residual Dual-domain Joint Learning Transformer Blocks (RDTBs) and a Multi-scale FeedForward Neural network for improved visual fidelity.































![Help ID please ! [wollongong NSW, aus]](https://preview.redd.it/hhhqvb0zv8zg1.jpeg?width=640&crop=smart&auto=webp&s=796e9f8ac03177900359f4cad3004000d6fd68a7)