Two-stage framework for correcting underwater distorted image sequences

Area of Science:

• Optics and Photonics
• Image Processing
• Fluid Dynamics

Background:

• Underwater imaging is challenged by geometric distortion and blur from water surface fluctuations.
• Existing correction methods suffer from residual distortion due to poor reference templates and outlier screening.

Purpose of the Study:

• To develop a robust two-stage framework for high-quality underwater imaging despite water surface disturbances.
• To improve geometric accuracy and detail clarity in underwater target images.

Main Methods:

• A two-stage framework separating preliminary correction and refined reconstruction.
• Stage 1: Rényi entropy for reference frame selection and Cox-Munk theory for centroid displacement compensation.
• Stage 2: Median selection-based inversely proportional weighting reconstruction (MSIPWR) for outlier rejection and detail enhancement.

Main Results:

• The proposed method significantly reduces residual distortion compared to existing techniques.
• Demonstrated superior performance in edge sharpness, structural integrity, and specular highlight suppression.
• Achieved precise noise suppression and enhanced fine texture details.

Conclusions:

• The novel two-stage framework offers a stable and effective solution for underwater imaging in dynamic water conditions.
• The method overcomes limitations of traditional approaches by using high-quality references and robust outlier rejection.
• Provides a significant advancement for clear and accurate underwater target visualization.