Robust underwater image enhancement method based on natural light and reflectivity

The poor visibility of underwater images is caused not only by scattering and absorption effects but is also related to light conditions. To improve robustness, a novel underwater image enhancement method based on natural light and reflectivity is proposed. Aiming at the scattering effects of reflectivity, a dehazing process based on the non-correlation of a foreground scene and background light is first conducted. Then, a more precise reflectivity can be estimated by substituting the captured image with the dehazed image. Moreover, classical methods often regard the dehazed image as the final result, but ignore the fact that attenuated natural light and nonuniform artificial light, which lead to insufficient brightness and halo effects, are included in the dehazed image, and are not robust to all scenes. This phenomenon enables us to remove the artificial light disturbance by introducing the dehazed image in the Lambertian model, and compensate for the loss of natural light energy by exploiting the light attenuation ratio map. Thus, the least-attenuated natural light can be further derived. Experimental results demonstrate that our method is satisfactory in producing more pleasing results under various circumstances.