Image Dehazing Using Curvelet Transform and Gaussian Inverse Filtering

This project presents a MATLAB implementation for image dehazing using Curvelet transform and Gaussian inverse filtering. The algorithm consists of two main computational phases: first, the Curvelet transform decomposes the hazy image into multiscale directional components using the Frequency Wrapping method, which effectively captures curved singularities and edge information. Second, Gaussian inverse filtering operates in the frequency domain with carefully tuned sigma parameters to mitigate atmospheric scattering effects while preserving image details. Key implementation aspects include: - Utilizing the CurveLab toolbox for efficient Curvelet coefficient computation - Implementing Wiener filter-based regularization to handle noise amplification during inverse filtering - Applying multiscale fusion techniques to combine enhanced components from different Curvelet scales - Adaptive parameter optimization based on atmospheric light estimation using quad-tree decomposition The algorithm generates comparative results demonstrating significant improvement in image clarity and detail recovery. Performance metrics including PSNR and SSIM measurements validate the effectiveness of our approach. Through this project, we explore advanced image processing techniques that substantially enhance visual quality and provide superior user experience for fog-affected images.