A Variational Model with Second-Order Laplacian for Underwater Image Restoration

Images captured underwater are usually degraded by color distortion, blur and low contrast because the light is absorbed and scattered when traveling through water, which limits their applications. To address these problems, based on the underwater image formation model, we establish a Laplacian variation model that includes novel data and regularization terms. Technically, we design a fidelity term to constrain the radiance scene, and a regularization item for divergence to strengthen the structure and texture details. Moreover, a bright-aware blending algorithm and quad-tree subdivision scheme are also integrated into our variational framework to estimate the transmission map and underwater background light, respectively. Accordingly, we provide a fast-iterative algorithm based on alternating direction method of multipliers to solve the optimization problem and accelerate its convergence speed. Experimental results demonstrate that the proposed method achieves outstanding performances on dehazing, detail preserving, and texture enhancement for underwater images. Moreover, extensive qualitative and quantitative comparisons with several state-of-the-art methods also validate the superiority of our proposed method