Current challenges in image dehazing are mostly related to scene variability, including darkened images and distortions caused by non-uniform fog density, atmospheric variability, and depth differences. This paper proposes a novel approach to address these issues by integrating Extreme Channels Prior (ECP) into restoration models. Specifically, an \(\ell ^0\) norm constraint is applied to the extreme channel values, which effectively guides the iterative restoration process towards a haze-free condition. Building on our ECP-based optimizer for the proposed dehazing model, we integrate the algorithm into established transmission and atmospheric-light estimation pipelines, including the dark channel prior, combined radiance–reflectance, and saturation-line frameworks, and we further align it with haze-lines and light absorption–enhanced strategies. The proposed ECP fundamentally diverges from existing dehazing methods in two key aspects: (1) It effectively reduces color distortions caused by errors in the estimation of transmission rates; (2) It significantly enhances the quality of the restored images, achieving more accurate and clearer restoration of the real scenes. Experiments on the SOTS, I-HAZE and RESIDE- \(\beta \) datasets are conducted to evaluate the proposed method, and test results verify that the integration of ECP consistently enhances the performance of the dehazing algorithm across various models, and that our approach surpasses several state-of-the-art dehazing methods with a computationally acceptable overhead.
