DehazeMist: Research on Image Dehazing System Based on Improved Dark Channel Prior

Image dehazing is a critical technology for enhancing the quality of images captured under low-visibility conditions. To address limitations of the classical single-image dehazing algorithm based on Dark Channel Prior (DCP), such as edge artifacts and color distortion in sky regions, this paper proposes an improved algorithm and develops an image dehazing system named DehazeMist. Specifically, the proposed method employs adaptive window sizing for transmission map calculation to balance haze removal intensity and detail preservation. It introduces guided filtering to optimize the rough transmission map for edge protection, effectively preventing halo artifacts in the restored images. On the system side, a graphical user interface (GUI) is designed with integrated functions for dynamic parameter tuning and real-time preview of results. Experiments conducted on images from various scenarios compare the proposed method with existing algorithms. The results indicate that the improved algorithm effectively enhances the clarity and naturalness of the dehazed images while maintaining computational efficiency.