Two-Stage Image Dehazing Algorithm Based on Atmospheric Model and Fusion Strategy

Traditional dehazing methods primarily enhance global contrast but often lead to the loss of fine details and distortions in sky regions. To overcome these limitations, a twostage dehazing algorithm is designed based on an atmospheric model and a fusion strategy. The atmospheric light and transmission maps are estimated using a quadtree algorithm and a linear model, followed by the inversion of the atmospheric scattering model to obtain an initial haze-free image. To generate images with varying exposure levels, gamma correction with different values is applied. A multi-scale fusion process is then employed using a non-subsampled difference-of-Gaussian pyramid, where weight maps are constructed to effectively integrate multi-scale components and restore a clear, haze-free image. Experimental results demonstrate that this algorithm reduces image distortion while preserving fine details, outperforming existing dehazing methods in both subjective and objective evaluations.