Underwater image enhancement via adaptive white-balancing and multi-restoration image fusion

Underwater images often suffer from color shift and visual blur due to light absorption and scattering in the water medium. An adaptive image enhancement method is proposed for underwater image quality improvement. First, adaptive color compensation is proposed based on the color features of the image, and selectively apply the gray world assumption according to the degree of color unevenness. Second, we use the color correction result combined with a rough estimation of ambient light and medium transmission map to restore the brightness and contrast of the underwater image. Next, based on different ambient light estimation values, we obtain multiple restoration results with different local features. Finally, we introduce the well-known multi-scale fusion strategy, using gradient matrix as weight, to fuse multiple restoration results. Extensive experiments on natural and synthetic underwater image datasets show that our method is robust in underwater image enhancement to achieve color, high contrast, moderate brightness in various complex underwater environments. Moreover, its effectiveness is also verified in the applications of underwater image saliency detection and natural image enhancement.