Related Experiment Video
Updated: May 28, 2026
10:56
Long-term Behavioral Tracking of Freely Swimming Weakly Electric Fish
Published on: March 6, 2014
Physically Guided Attention Mechanism for Underwater Motion Deblurring via Cepstrum-Based Blur Estimation
Ning Hu1, Shuai Li2, Jindong Tan1
1Department of Mechanical, Aerospace and Biomedical Engineering, University of Tennessee, Knoxville, TN 37916, USA.
Journal of Imaging
|May 26, 2026
Summary
This study introduces a new Transformer framework to effectively remove motion blur from underwater images. The method enhances clarity and visual quality, outperforming existing techniques for clearer underwater visuals.
Area of Science:
- Computer Vision
- Image Processing
- Underwater Imaging
Background:
- Underwater images are degraded by motion blur, impacting clarity and vision tasks.
- Existing deblurring methods struggle with the complexities of underwater environments.
Purpose of the Study:
- To develop a physically guided Transformer framework for effective underwater motion deblurring.
- To improve structural clarity, sharpness, contrast, and color consistency in degraded underwater images.
Main Methods:
- A two-stage cepstrum-based blur estimation process.
- Integration of a point spread function (PSF)-guided self-attention mechanism within a Transformer.
- Robust estimation of blur parameters using cepstrum analysis, ellipse fitting, and negative-peak refinement.
Main Results:
- Achieved superior UIQM/UCIQE scores on real underwater datasets (UIEB Challenge-60 and EUVP330).
- Significantly outperformed UFPNet and Phaseformer in perceptual restoration.
- Attained high PSNR (24.23 dB) and SSIM (0.918) on synthetic data, surpassing deep and classical methods.
- Demonstrated consistent top performance in water-tank experiments across various motion speeds.
Conclusions:
- The proposed physically guided Transformer framework offers an effective solution for underwater motion deblurring.
- The method shows strong restoration fidelity, structural consistency, and practical applicability.
- Provides a physically interpretable approach to enhance underwater image quality.
