Illuminant Intensity Compensation with Depth Estimation for Underwater Image Restoration

Visual quality restoration in underwater images has been paid more attention in recent years as it is a fundamental task in many relevant image processing applications. This paper investigates a new underwater image restoration algorithm based on a simplified image formation model by the integration of the Jaffe-McGlamery and Lambertian systems. The Retinex theory is introduced into the prototype to explicitly disclose the illuminant intensity, which is computed using an efficient gray index scheme for light source attenuation compensation. Subsequently, an improved scene depth estimation method is exploited to extract the background scene, on which an efficient compensation method is proposed for better background light estimation. Finally, an ensemble color gain is appraised to correct color deviation. A wide variety of underwater images with various scenarios in three different datasets were employed to evaluate the proposed image quality restoration system. Experimental results demonstrated the advantages of our underwater image restoration algorithm over many state-of-the-art methods both qualitatively and quantitatively. It is believed that the developed image restoration framework is of potential in many underwater image processing applications.