Research on Haze Image Enhancement based on Dark Channel Prior Algorithm in Machine Vision.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Foggy images suffer from low visibility due to noise, low resolution, and uneven illumination.
• Traditional dark channel prior (DCP) methods can lead to uneven colors and dark overall image appearance.

Purpose of the Study:

• To propose an improved foggy image enhancement method based on dark channel prior.
• To address limitations of traditional methods, such as color unevenness and image darkening.

Main Methods:

• Refining transmittance and optimizing atmospheric light value.
• Enhancing brightness (V component) using MSRCR with bilateral filtering and saturation (S) with adaptive stretching.
• Converting images between HSV and RGB color spaces for processing.

Main Results:

• The proposed method effectively overcomes noise amplification and edge blur common in conventional enhancement algorithms.
• It significantly improves image darkening and prevents distortion across various image formats (JPEG, BMP, GIF, PNG, PSD, TIFF).
• Quantitative analysis shows superior performance in PSNR, SSIM, and IE compared to DCP, SSR, MSR, MSRCR, and CLAHE.

Conclusions:

• The improved algorithm demonstrates superior performance in enhancing heavily polluted haze images.
• It effectively preserves edge information, showing good adaptability and stability.
• The method provides a robust solution for improving the visual quality of foggy images.