JDPNet: A Network Based on Joint Degradation Processing for Underwater Image Enhancement

Area of Science:

• Computer Vision
• Image Processing
• Underwater Imaging

Background:

• Underwater images suffer from complex, non-linearly coupled degradations due to water's properties.
• Existing methods often address degradations individually, neglecting their coupled nature.
• Processing these coupled degradations effectively is a significant challenge in computer vision.

Purpose of the Study:

• To develop a unified framework for processing multiple, coupled underwater image degradations.
• To effectively mine and unify information from coupled degradation features.
• To improve the quality of underwater images by addressing nonlinear interactions.

Main Methods:

• Proposed JDPNet (Joint Degradation Processing Network) with a joint feature-mining module.
• Introduced a probabilistic bootstrap distribution strategy for feature adjustment.
• Developed AquaBalanceLoss to balance color, clarity, and contrast during training.

Main Results:

• JDPNet demonstrated state-of-the-art performance on multiple underwater image datasets.
• The network effectively captured and processed nonlinear interactions of coupled degradations.
• Achieved a favorable tradeoff between performance, parameter size, and computational cost.

Conclusions:

• JDPNet offers a robust solution for joint underwater image degradation processing.
• The proposed methods successfully address the limitations of existing approaches.
• This work advances the field of underwater image restoration with an efficient and effective network.