Underwater Image Enhancement Using a Diffusion Model with Adversarial Learning.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Underwater images suffer from low contrast, color distortion, and noise due to environmental factors.
• Existing underwater image enhancement methods lack generalization across diverse underwater conditions.

Purpose of the Study:

• To develop a robust and generalizable underwater image enhancement method.
• To improve the visual quality and quantitative metrics of degraded underwater images.

Main Methods:

• Introduced adversarial learning diffusion underwater image enhancement (ALDiff-UIE), a diffusion model-based approach.
• Employed an adversarial learning strategy with a generator and discriminator for noise elimination and quality enhancement.
• Incorporated a multi-scale dynamic-windowed attention mechanism for effective feature fusion.

Main Results:

• ALDiff-UIE demonstrated significant improvements in underwater image enhancement.
• Achieved an average increase of 12.8% in PCQI and 15.6% in UIQM across benchmark datasets.
• Outperformed several mainstream underwater image enhancement approaches in qualitative and quantitative evaluations.

Conclusions:

• ALDiff-UIE offers a superior solution for enhancing underwater images compared to existing methods.
• The proposed method shows strong generalization capabilities across various underwater environments.
• ALDiff-UIE effectively enhances visual quality and quantitative metrics for underwater imagery.