Dark Channel Prior-Based Image Dehazing Algorithm

Implementation of the CVPR best paper award-winning dark channel prior image dehazing algorithm with soft matting, featuring algorithm improvements including color distortion correction and a fast algorithm achieving soft matting effects

Research Papers on Dark Channel Prior-Based Image Dehazing

Best Paper award at CVPR 2009 - He Kaiming's seminal work on dark channel prior-based image dehazing, including associated papers, presentation materials, and MATLAB implementation. The dark channel prior represents a statistical observation of haze-free images: in most local image patches, at least one color channel contains pixels with very low intensity values. This principle enables direct estimation of haze concentration from a single input image and recovery of high-quality dehazed results through optimized atmospheric scattering modeling and transmission map calculations.