An Improved Dark Channel-Based Algorithm for Underwater Image Restoration

Abstract

Underwater imaging is crucial to a wide variety of research and realistic applications in marine biology, water fauna identification and assessment, archaeology, mine detection, oceanic mapping, and autonomous underwater robotics. However, due to specific propagation properties of light in water such as absorption and scattering as well as unstable environment such as light changing and water turbidness, the images captured are highly disturbed with low contrast, blurring, darkness, and color diminishing. This paper proposes a new underwater image restoration algorithm that consists of three major phases: haze removal, color correction, and contrast enhancement. To estimate the transmission coefficient function, we first compute the dark channel map, which is the set of “dark” pixels having very low intensity values in at least one RGB color channel. To accommodate the blue color distortion phenomenon, the red channel value is excluded from the calculation if the peak histogram intensity is smaller than a specified threshold. To optimize the medium transmission map, we adopt the matting Laplacian matrix associated with the sparse linear system to generate a cost function, followed by the guided filtering method to accelerate the computation. Finally, the contrast limited adaptive histogram equalization method is used to enhance the contrast while maintaining the color fidelity. We have applied this new approach to a wide variety of underwater images. Experimental results indicated that this new method is of potential in facilitating the interpretation and perception of underwater images in the fields of ocean engineering, ocean biology, and ocean science.