Improved dark channel priori single image defogging technique using image segmentation and joint filtering.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Foggy images present challenges in analysis due to low definition and blurred details.
• Existing haze removal techniques struggle with sky regions, leading to misestimated atmospheric light and halo artifacts.

Purpose of the Study:

• To propose an improved dark channel prior single-image defogging technique.
• To effectively segment sky areas and mitigate halo artifacts in hazy images.
• To enhance the detail recovery and visual quality of defogged images.

Main Methods:

• Image segmentation using K-means clustering and probability density distribution functions to estimate atmospheric illumination values.
• Optimization of the image transmittance map via joint filtering (fast-guided filtering and weighted least-squares filtering).
• Post-processing with gamma correction and automatic level optimization for brightness and contrast enhancement.

Main Results:

• Effective segmentation of sky and non-sky regions in hazy images.
• Significant suppression of halo artifacts compared to traditional dark channel prior methods.
• Improved image detail recovery and overall visual quality in defogged images.

Conclusions:

• The proposed technique offers superior performance in defogging images with sky areas.
• It effectively addresses limitations of existing methods, providing better subjective and objective evaluations.
• This approach enhances the practical applicability of image defogging in challenging environments.