Fattal's Single Image Dehazing Algorithm Implementation

This MATLAB code implements Fattal's single image dehazing algorithm. The implementation employs Independent Component Analysis (ICA) methodology based on the fundamental assumption that surface shading and atmospheric transmission functions are statistically uncorrelated within local image patches. The core algorithm operates by first estimating the atmospheric light and transmission map through statistical analysis of color channels and local image statistics. Key functions include patch-based correlation analysis, ICA decomposition for separating illumination components, and transmission refinement using edge-aware filtering techniques. While demonstrating excellent performance on certain image types with moderate haze distribution, the method shows limitations when handling images with dense fog, strong atmospheric light interference, or scenes violating the underlying statistical assumptions. Fattal's research provides a mathematically rigorous approach to image dehazing, though further improvements in robustness and generalization remain valuable research directions.