Atmospheric Scattering And Segmentation Based Foggy Image Enhancement

Abstract

We Propose A Novel Foggy Image Enhancement Pipeline That Integrates An Improved Atmospheric Scattering Model (ASM) With Otsu-based Segmentation. The System First Converts The RGB Input To Grayscale, Then Applies Otsu’s Method To Segment Fog-dense Regions. This Segmentation Guides A Region-specific Inverse ASM Dehazing: We Estimate Atmospheric Light And Transmission Differently For Fog And Non-fog Areas To Avoid Global Artifacts. Each Color Channel Is Then Enhanced According To The Refined ASM And Recombined To Preserve Color Fidelity. The Method Is Evaluated On Synthetic And Real Foggy Images Using Standard Metrics: Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), And Mean Squared Error (MSE). Results Show That Segmenting Out Heavy-fog Regions Before Applying ASM Yields Clearer, More Natural Images Compared To Baseline Defogging. For Example, Our Approach Achieves Higher PSNR And SSIM (closer To 1) And Lower MSE Than Conventional Methods, Confirming Its Effectiveness. We Include Example MATLAB Code Illustrating Grayscale Conversion And RGB Reconstruction.

Keywords