An Improved Dehazing Algorithm for Remote Sensing Images Using Multi-Scale Atmospheric Light Estimation and Adaptive Side Window Filtering

Image dehazing is one of the key preprocessing steps necessary for enhancing the visibility and quality of images taken in hazy conditions. This paper proposes an advanced dehazing algorithm that integrates multi-scale atmospheric light estimation and adaptive side window filtering (ASWF) to address the limitations of existing methods. The proposed system begins with white balance correction to remove color biases, followed by Contrast Limited Adaptive Histogram Equalization (CLAHE) for local contrast enhancement. Atmospheric light is estimated adaptively using dynamic patch sizes, ensuring accurate estimation in heterogeneous haze conditions. The transmission map is refined using ASWF, which preserves edges and reduces artifacts. A fusion step combines the dehazed image with the CLAHE-enhanced version to achieve optimal results. Evaluated on the Haze1k dataset [9], the proposed method demonstrates superior performance in terms of PSNR, SSIM, NIQE, and BRISQUE metrics, making it suitable for remote sensing and photography applications. The proposed method outperforms existing techniques by effectively handling non-uniform haze and preserving fine details, as demonstrated through extensive experiments and comparisons.