UVEB: A Large-scale Benchmark and Baseline Towards Real-World Underwater Video Enhancement

TL;DR

The paper tackles the lack of large-scale real-world paired data for underwater image/video enhancement (UIE) and the underutilization of inter-frame information in underwater videos. It introduces UVEB, a large-scale benchmark with 1,308 video pairs and 453,874 high-resolution frames (including 173,797 UHD 4K frames), and a GT generation protocol based on evaluating 20 UIE baselines, yielding reliable ground-truth quality scores. It then presents UVE-Net, a supervised underwater video enhancer that exchanges inter-frame information via convolution-kernel guidance generated from a downsampled middle frame, using FEGM and FRGM modules to produce $f_i^e$ and $f_i^r$ without frame alignment. Experiments show significant PSNR gains over 20 baselines and real-time capability with a lighter variant, underscoring UVEB’s diversity and UVE-Net’s efficiency for practical underwater vision applications.

Abstract

Learning-based underwater image enhancement (UIE) methods have made great progress. However, the lack of large-scale and high-quality paired training samples has become the main bottleneck hindering the development of UIE. The inter-frame information in underwater videos can accelerate or optimize the UIE process. Thus, we constructed the first large-scale high-resolution underwater video enhancement benchmark (UVEB) to promote the development of underwater vision.It contains 1,308 pairs of video sequences and more than 453,000 high-resolution with 38\% Ultra-High-Definition (UHD) 4K frame pairs. UVEB comes from multiple countries, containing various scenes and video degradation types to adapt to diverse and complex underwater environments. We also propose the first supervised underwater video enhancement method, UVE-Net. UVE-Net converts the current frame information into convolutional kernels and passes them to adjacent frames for efficient inter-frame information exchange. By fully utilizing the redundant degraded information of underwater videos, UVE-Net completes video enhancement better. Experiments show the effective network design and good performance of UVE-Net.

Figures (29)

• Figure 2: Distribution
• Figure 3: Degradation types
• Figure 4: Resolution
• Figure 6: MOS of the samples before and after enhancement.
• Figure 8: The rating increases as the video quality improves.