Color restoration in turbid medium

Abstract

Light scattering and color distortions are two major issues with underwater imaging. Scattering occurs due to turbidity of the medium and color distortions are caused by differential attenuation of wavelengths as a function of depth. As a result, underwater images taken in a turbid medium have low contrast, color cast, and color loss. The main objective of this work is color restoration of underwater images i.e, produce its equivalent image as seen outside of the water surface. As a first step, we account for low contrast by employing dark channel prior based dehazing. These images are then color corrected by learning a mapping function between a pair of color chart images, one taken inside water and another taken outside. The mapping thus learned is with respect to a reference distance from the water surface. We also propose a color modulation scheme that is applied prior to color mapping to accommodate the same mapping function for different depths as well. Color restoration results are given on several images to validate the efficacy of the proposed methodology.