Underwater Image Clarification Based on Double-Opponency Light Estimation and Red Channel Prior

Abstract

Underwater images typically exhibit color bias and low contrast resulting from the absorption and scattering effects of light propagating in water. Such degraded images often have difficulty in satisfying the requirements of underwater operations. Traditional underwater image restoration and enhancement methods generally require long computation times with unsatisfactory results. In this study, we propose a new underwater image restoration method involving background light estimation based on double-opponency and transmission estimation based on the red channel prior. For background light estimation, double-opponency is adopted to imitate human visual color constancy, whereby background light information consistent with human perception is obtained. Transmission estimation uses the prior knowledge that red wavelength has the fastest attenuation in water. To enhance the contrast effect of the resulting image, the restored image is fused with that of contrast stretch processing in HSV space to obtain the final output image. Experiments indicate that the proposed method can reduce computation time by 40% or more relative to comparison methods, thereby producing a clearer normal color image according to both objective and subjective evaluations