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

TL;DR

This paper introduces UVEB, a large-scale underwater video enhancement benchmark, and proposes UVE-Net, a novel supervised method leveraging inter-frame information for improved underwater video quality.

Contribution

The paper provides the first large-scale high-resolution underwater video benchmark and a new supervised enhancement method utilizing inter-frame information exchange.

Findings

UVEB contains 1,308 paired video sequences with over 453,000 high-resolution frames.

UVE-Net effectively utilizes inter-frame information for superior enhancement performance.

Experiments demonstrate the effectiveness of UVE-Net and the utility of the UVEB benchmark.

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.