Search papers, labs, and topics across Lattice.
This paper introduces Frequency Reconstruction via Spectral Harmonization (FReSH-IR), a novel lightweight method for restoring images affected by adverse weather conditions. By employing a hierarchical encoder-decoder architecture that decomposes features into high- and low-frequency components, FReSH-IR effectively captures essential frequency information while maintaining spatial detail. The method achieves comparable restoration quality to transformer-based models while utilizing 80% fewer parameters and operations, making it particularly suitable for resource-constrained environments.
Achieving high-quality image restoration with 80% fewer parameters, FReSH-IR redefines efficiency in weather-affected image processing.
Adverse weather conditions such as rain, haze, and snow significantly degrade image quality, posing challenges for both human perception and physical AI. Existing restoration methods require large computational budgets, struggling to process high-resolution images and handle different degradations. In this paper, we present Frequency Reconstruction via Spectral Harmonization, a novel lightweight all-in-one restoration method that explicitly decomposes feature representations into high- and low-frequency components at each scale of a hierarchical encoder-decoder architecture. By combining spectral decomposition with spatial processing through Fourier-based skip connections, FReSH-IR captures complementary frequency information without sacrificing spatial detail. Our approach achieves similar restoration quality with 80% fewer parameters and operations than transformer-based models. Extensive experiments demonstrate that our method offers a great efficiency-performance trade-off, highlighting its practical applications in constrained-resource systems.
