A polarization-based image restoration method for both haze and underwater scattering environment

Area of Science:

• Computer Vision
• Image Processing
• Optical Engineering

Background:

• Existing polarization-based defogging algorithms are limited in scenes with low polarized light.
• These methods often rely on polarization degree or angle, reducing their effectiveness.
• Scattering media like haze and water degrade image quality.

Purpose of the Study:

• To propose a universal image restoration method for both haze and underwater scattering environments.
• To develop a technique that is effective even in scenes with minimal polarized light.
• To improve upon existing polarization-based defogging algorithms.

Main Methods:

• Utilizes the assumption that clear images have higher gray variance and average gradient than scattered images.
• Calculates maximum (I_best) and minimum (I_worst) variance polarimetric images from four captured polarization images.
• Estimates transmittance to remove scattering light and fuses I_best and I_worst for color and clarity restoration.

Main Results:

• Successfully restores clear images in both haze and underwater scattering conditions.
• Demonstrates effectiveness in scenes with low polarized light.
• Achieves superior performance compared to algorithms dependent on polarization degree or angle.

Conclusions:

• The proposed method offers a more universal approach to image restoration in scattering media.
• It effectively addresses the limitations of existing polarization-based techniques.
• The technique is suitable for a wider range of challenging visual environments.