Skip to content
Open access

WaveUIE: Efficient Wavelet-Domain Underwater Image Enhancement

2026 · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · Vol 19, pp. 35082-35098 · 0 citations · 70 references

Abstract

Underwater image enhancement (UIE) remains challenging because underwater degradations are not uniform in the image space: color distortion and illumination attenuation mainly affect global appearance, while scattering and blurring largely degrade local textures and structural details. However, most existing deep UIE methods formulate enhancement as a generic spatial-domain restoration problem, leaving the frequency-dependent nature of underwater degradations insufficiently explored. In this article, we revisit UIE from a wavelet-domain perspective and identify a degradation-frequency correspondence: low-frequency components are closely related to global color and illumination correction, whereas high-frequency components are more responsible for texture and detail restoration. Based on this observation, we propose WaveUIE, a degradation-aware wavelet-domain enhancement framework that explicitly decomposes underwater images into low- and high-frequency components and restores them with targeted modeling. Rather than treating different frequency components independently, WaveUIE further introduces low-frequency-guided high-frequency refinement to achieve coherent restoration between global appearance and local details. Moreover, we exploit degraded underwater images not merely as inputs but also as informative negative samples, and propose a wavelet fine-grained contrastive policy to learn discriminative degradation-aware representations across multiple wavelet sub-bands. Extensive experiments on UIEB, large-scale underwater image, and enhancement of underwater visual perception demonstrate that WaveUIE achieves state-of-the-art performance in both full-reference and no-reference metrics, while maintaining a favorable accuracy-efficiency tradeoff.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.