A Simple Method for Backscattered Light Estimation and Image Restoration in Turbid Water
Frans Shafuda, Hayato Kondo · OCEANS 2021: San Diego – Porto · 2021
In a turbid water environment, scattering and absorption phenomena due to suspended particles in the water affect the propagation of light. This results in the degradation of captured images due to poor visibility and low object contrast. The image restoration process in turbid water involves estimation of the amount of backscattering, once estimated correctly it can be removed from the total radiance of an image to recover the scene radiance. Our study proposes a simple method for backscattered light estimation and image restoration based on the dark channel prior (DCP) [1] and the light channel prior [2]. We show that backscattered light in a degraded underwater image can be estimated by simply computing its light channel image and directly use this as the backscattered light map in the image restoration process. The proposed approach is suitable for uniformly illuminated underwater turbid scenes. We have tested our method using underwater images from the TURBID dataset in [3] and compared it to standard underwater image restoration algorithms such as the underwater dark channel prior (UDCP) method. The experimental results show that the proposed method yields better image restoration results in comparison to state-of-the-art methods at very low computational load. The proposed method has low computational load due to its simplicity in estimating the backscattered light.