Underwater Image Restoration Based on A New Underwater Image Formation Model

Abstract

Underwater image restoration is crucial for compute applications and consumer electronics. However, restoring underwater image from a single image is an odd-ill problem due to the complicated underwater environment. To improve the visual quality of underwater image, we propose an underwater image restoration method. First, we present a new underwater image formation model, which takes the properties of underwater imaging and light into account. Then, a medium transmission estimation method for underwater image based on joint prior is proposed, which, respectively, predicts the medium transmissions of three channels of an underwater image. Moreover, we replace the global background light, which is always used in previous underwater image restoration method, with the colors of light source to correct the color casts appeared on the degraded underwater image. The performance of the proposed method is evaluated on the degraded underwater images taken from different scenes by qualitative and quantitative comparisons. Experimental results demonstrate that our results look more visually pleasing and outperforms the results of several existing methods, especial for the colors and contrast.