Adapting a Dehazing System to Haze Conditions by Piece-Wisely Linearizing a Depth Estimator

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Road safety is significantly impacted by haze, a common weather condition leading to numerous accidents.
• Existing image dehazing algorithms lack robustness across varied haze densities, necessitating advanced solutions.

Purpose of the Study:

• To develop an adaptive image dehazing method capable of reliably handling diverse haze conditions.
• To enhance road safety by improving the performance of dehazing systems in real-world scenarios.

Main Methods:

• A novel method estimates haze density to discriminate various haze conditions.
• A piece-wise linear weight is generated based on haze discrimination to modify the depth estimator.
• A real-time hardware implementation is developed for seamless integration into embedded systems.

Main Results:

• The proposed method effectively handles arbitrary input images across different haze levels.
• Comparative assessments confirm the efficacy of the adaptive dehazing method against benchmark designs.
• The hardware implementation demonstrates real-time performance for practical applications.

Conclusions:

• The adaptive image dehazing approach offers a robust solution for varying environmental conditions.
• The developed system significantly contributes to improving visibility and safety in hazy driving environments.
• The real-time hardware implementation facilitates widespread adoption in automotive and embedded systems.