UIF: An Objective Quality Assessment for Underwater Image Enhancement

Area of Science:

• Computer Vision
• Image Processing
• Oceanography

Background:

• Underwater imaging suffers from complex lighting, scattering, and noise, degrading visual quality.
• Underwater Image Enhancement (UIE) techniques are crucial for improving visual fidelity.
• Existing objective UIE evaluation methods struggle with deep learning approaches.

Purpose of the Study:

• To propose a novel objective metric, Underwater Image Fidelity (UIF), for evaluating enhanced underwater images.
• To address the limitations of time-consuming subjective evaluations and inadequate objective metrics for modern UIE methods.

Main Methods:

• Developed the UIF metric by analyzing statistical features in CIELab color space.
• Incorporated naturalness, sharpness, and structure indexes, combined via saliency-based spatial pooling.
• Created the Underwater Image Enhancement Database (UIED) with subjective scores for validation.

Main Results:

• The proposed UIF metric demonstrates superior performance compared to existing underwater and general image quality metrics.
• Experimental results validate the effectiveness of UIF in objectively assessing UIE performance.
• The UIED database provides a benchmark for UIE research.

Conclusions:

• The UIF metric offers a reliable and efficient objective evaluation for enhanced underwater images.
• The developed metric and database advance the field of underwater image processing and analysis.
• UIF is particularly valuable for evaluating emerging deep learning-based UIE techniques.