Improved dark channel prior single image defogging
Tongying Guo, Na Li, Chao Zhang · 2021
The foggy sea image has low contrast, color distortion, blurred edges and other problems. Sea fog brings great difficulties to ocean observation. In response to these problems, this paper proposes an improved single sea fog defogging method based on domain decomposition and dark and bright channel priors. First, it is proposed to segment the sky area through gradient information and edge tracking, and combine the dark channel of the sky area to determine the atmospheric light value. Secondly, it is proposed to optimize the transmission map based on the bright channel of the sky area and the dark channel of the non-sky area, and use the guided filtering to optimize the edge. Finally, the fog-free image is obtained by combining the atmospheric scattering model. Evaluate the performance of our algorithm from a comprehensive test. Experiments on images under different sea fog scenes show that the algorithm can not only effectively overcome the sky color distortion and halo phenomenon, but also restore the details of non-sky areas.