Image Dehazing with Improved Dark Channel Prior Algorithm

This code implements a simulation based on He Kaiming's proposed dark channel prior dehazing algorithm. The dark channel prior algorithm is a widely-used image dehazing method that leverages the observation that in images with sky backgrounds, most regions contain pixels with relatively low intensity values, while sky regions exhibit even lower intensities due to minimal fog interference. The algorithm utilizes this characteristic by calculating global minimum values to estimate transmission rates in sky regions, then performs dehazing based on the derived transmission map. The implementation involves key computational steps including: dark channel calculation using minimum filters across color channels, atmospheric light estimation through brightest pixel selection, and transmission refinement using soft matting techniques. The primary improvement in this version addresses sky region distortion issues by incorporating color adjustment mechanisms that dynamically correct atmospheric light estimates in sky areas. This optimization employs sky region detection algorithms and adaptive color correction functions to eliminate distortion artifacts, significantly enhancing the visual quality of processed images while maintaining natural color balance across different image regions.