Haze Removal Using Radial Basis Function Networks for Visibility Restoration Applications.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Hazy image restoration is crucial for outdoor computer vision applications.
• Existing methods often struggle to balance visibility, brightness, and artifact reduction.
• Need for advanced techniques to handle complex atmospheric conditions.

Purpose of the Study:

• To propose a novel haze removal method using radial basis function (RBF) artificial neural networks.
• To effectively remove haze while preserving image brightness and visible edges.
• To address limitations of traditional single-atmospheric-veil models.

Main Methods:

• Development of RBF artificial neural networks for dynamic multi-atmospheric veil learning.
• Utilizing scene complexity to adapt the neural network's learning process.
• Employing a specific activation function during testing to enhance image brightness.

Main Results:

• The proposed RBF network method successfully removes haze from images.
• Restored images exhibit improved brightness and more visible edges compared to traditional methods.
• Qualitative and quantitative evaluations demonstrate superior performance on benchmark hazy images.

Conclusions:

• The RBF-based artificial neural network offers a robust solution for haze removal.
• The method effectively enhances image quality by increasing brightness and edge visibility.
• This approach represents a significant advancement over current state-of-the-art haze removal techniques.