Dichromatic Scattering Model for Image Dehazing Algorithm

The Dichromatic Scattering Model for image dehazing is a physics-based approach that enhances the clarity of foggy images by modeling light scattering effects. This algorithm reconstructs original scene illumination information through mathematical modeling of haze's light scattering properties, effectively removing atmospheric interference from image quality.

Core Principle The dichromatic scattering model assumes haze scattering consists of two primary components: atmospheric light (global illumination) and object-reflected light. The algorithm estimates these components combined with depth information (distance between scene points and camera) to compute the dehazed image. Depth map acquisition is critical and can be achieved through stereo vision, deep learning methods, or sensors like LiDAR.

Algorithm Pipeline Atmospheric Light Estimation: Typically selects the brightest image regions as approximate atmospheric light values using intensity-based filtering. Transmission Map Calculation: Computes light attenuation through haze using depth information, often implemented with exponential decay functions. Image Restoration: Applies the dichromatic model to correct illumination distribution, utilizing matrix operations for efficient pixel-wise processing.

Performance and Optimization Compared to traditional methods like Dark Channel Prior, the dichromatic scattering model demonstrates superior detail preservation and color fidelity, particularly in non-uniform haze distribution scenarios. Optimization focuses on improving depth map accuracy through sensor fusion and enhancing computational efficiency via GPU parallelization.

Application Scenarios This algorithm is widely deployed in autonomous driving systems, aerial photography enhancement, and surveillance systems, significantly improving image usability in low-visibility environments through real-time processing implementations.