Single Image Defogging Based on Illumination Decomposition for Visual Maritime Surveillance

Area of Science:

• Computer Vision
• Image Processing
• Remote Sensing

Background:

• Single image defogging is crucial for surveillance, but existing methods often confuse fog and haze, failing to account for distinct scattering properties.
• Maritime surveillance is particularly challenged by fog's significant influence on illumination and scattering, necessitating specialized defogging techniques.

Purpose of the Study:

• To propose a novel single image defogging method tailored for visual maritime surveillance applications.
• To address the limitations of current defogging techniques that inadequately model fog's unique scattering properties and illumination effects.

Main Methods:

• A comprehensive scattering model is developed to represent fog images under glow-shaped environmental illumination.
• An illumination decomposition algorithm is employed to remove glow effects and recover a fog layer with uniform luminance for distant objects.
• Transmission-map estimation utilizes a non-local haze-lines prior to constrain the transmission map within a reasonable range.
• An illumination compensation algorithm is applied to preserve the natural illumination of the defogged image.

Main Results:

• The proposed method effectively removes fog from images captured in maritime surveillance scenarios.
• Experimental results demonstrate superior performance compared to state-of-the-art methods based on both subjective and objective evaluations.
• The defogging process successfully maintains the natural illumination characteristics of the original images.

Conclusions:

• The developed method offers a significant advancement in single image defogging for maritime visual surveillance.
• It provides a robust solution for enhancing image quality in foggy conditions, outperforming existing approaches.
• The method's ability to preserve naturalness makes it highly suitable for real-world surveillance applications.