Mathematical Modelling and Computational Approach for Underwater Image Enhancement Using Dark Channel Prior and Filtering Techniques

Abstract

Underwater imagery is a critical tool for marine exploration, infrastructure inspection, and biological research.  However, the aquatic medium severely degrades image quality through wavelength-dependent light absorption  and scattering, leading to pervasive color casts, diminished contrast, and loss of detail. This paper presents a  robust computational framework for underwater image enhancement by integrating the Dark Channel Prior  (DCP), originally developed for haze removal in atmospheric environments, with advanced filtering techniques.  The proposed model first employs the DCP to estimate the depth map and the global background light,  effectively characterizing the veiling effect. A transmission map is subsequently refined using a guided filter to  suppress noise and block artifacts while preserving edge information. The restored image is then reconstructed  by inverting the underwater optical model. Furthermore, a post-enhancement white balancing and contrast  stretching module is incorporated to rectify residual color inaccuracies and improve dynamic range.  Experimental results on a diverse dataset of underwater scenes demonstrate that the proposed method achieves  superior performance in color fidelity, contrast enhancement, and edge clarity compared to several state-of-the art techniques, providing a more reliable computational solution for autonomous underwater vehicles and  marine scientific analysis.