A Joint De-Rain and De-Mist Network Based on the Atmospheric Scattering Model

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Rain and mist cause visual distortions like streaks and halos, degrading image quality.
• Existing methods struggle to effectively remove both rain and mist simultaneously.

Purpose of the Study:

• To develop a novel approach for simultaneous de-raining and de-misting of images.
• To enhance the clarity and quality of images captured in rainy and misty conditions.

Main Methods:

• A rain and mist model based on atmospheric scattering for initial image reconstruction.
• Deep Memory Block (DMB) for selective extraction of rain and mist layer transfer spectra.
• Multi-scale Convolution Block (MCB) for feature extraction and model enhancement.

Main Results:

• The proposed Joint De-rain and De-mist Network (JDDN) effectively removes both streaks and halos.
• JDDN outperforms state-of-the-art deep learning methods on synthetic and real-world datasets.
• An average improvement of 0.29 dB was achieved on heavy-rainy-image datasets.

Conclusions:

• The JDDN model offers a robust and accurate solution for joint de-raining and de-misting.
• This approach significantly improves image quality in adverse weather.
• The method demonstrates superior performance compared to existing techniques.