Atmospheric Scattering Model Combined With Image Segmentation in Foggy Image Enhancement

This project presents a unified MATLAB framework for image enhancement and quality evaluation using the Otsu-ASM (Adaptive Sky Modeling) algorithm. The proposed method integrates dark channel prior, Otsu-based sky segmentation, and adaptive white region refinement to recover high-quality images from degraded or hazy inputs. The framework automatically estimates atmospheric light and transmission maps, enabling natural color restoration and improved visibility. Quantitative metrics such as MSE, PSNR, SSIM, entropy, and average gradient are computed to assess image quality. Additionally, a dataset evaluation module generates PSNR fitness curves, precision–recall plots, and average runtime bar charts to analyze model consistency and computational efficiency. The results demonstrate that Otsu-ASM achieves competitive PSNR and SSIM values while maintaining a balanced runtime performance. The framework’s modular structure supports single-image testing and batch dataset processing, offering a robust tool for image dehazing and enhancement analysis.

Keywords: Image Enhancement; Otsu-ASM; Dark Channel Prior; Atmospheric Light Estimation; Sky Segmentation; Transmission Refinement; PSNR Fitness Curve; Structural Similarity (SSIM); Image Quality Metrics; MATLAB Framework