Underwater Image Restoration via Maximum Attenuation Identification

Underwater images are difficult to process because of low contrast and color distortion. The in-water light propagation model was proposed several years ago but is relatively complicated to be used in reality. In this paper, the full underwater light propagation model is simplified to be used as the transmission model. On the basis of this model, we propose a new method, called maximum attenuation identification, to derive the depth map from degraded underwater images. At the same time, regional background estimation is implemented to ensure optimal performance. Experiments on three groups of images, namely, natural underwater scene, calibration board, and colormap board, are conducted. We report the quantitative and qualitative comparisons of our approach with existing state-of-the-art approaches. The performance evaluation on contrast enhancement and color restoration validates that our approach outperforms existing state-of-the-art approaches.