Single image dehazing algorithm using complementary saturation prior

In foggy environments, the presence of large amounts of particulate matter in the atmosphere can lead to reduced visibility of the scene. The traditional Dark Channel Prior (DCP) algorithm and its related algorithms have problems such as distortion of high-brightness sky region when removing fog from images. To solve this problem, a fast-dehazing algorithm based on complementary saturation prior is proposed. First, a mathematical model for estimating the transmission based on the image brightness is established, which will simplify the computation. Moreover, this approach also cuts down the running time of the program. Secondly, a method is proposed to estimate the atmospheric light based on the complementary saturation prior, which find the atmospheric light more accurately. The subjective and objective experimental results indicate that the algorithm markedly reduces the computational complexity of fog image recognition and can effectively address the problem of incorrect atmospheric light estimation caused by traditional DCP algorithms in environments with headlights, bright lights, etc. The proposed algorithm has the best Fast Automatic Dehazing Evaluation (FADE) metrics compared to the up-to-date single image dehazing algorithm, which shows that the proposed algorithm removes more fog. It maintains the naturalness of the image while significantly reducing the haze and produces visually pleasing images without halo artifacts.