Improved Color Balance and Fusion for Enhancement of Underwater Image

Abstract

In underwater photography, the absorption and scattering of light may probably cause low contrast, blurred images, and color cast. We present an effective method for improving the quality of underwater images that have been degraded by medium scattering and absorption. It combines color compensation with multi-scale image fusion. Color compensation consists of independently correcting the value of $\mathbf{R}$ channel and G-B channel of the input image and adjusting the white balance of the corrected image. After color compensation, the color of degraded image is restored effectively, but the blurring of image edges and details receiving scattering effects cannot be remedied. This problem can be solved by multi-scale image fusion. In our method, we use Gaussian pyramid and Laplace pyramid images for fusion, and these two feature pyramids are constructed by using the image after the illumination adjustment. The experimental results illustrate that dark-area sensitivity, global contrast, and edge sharpness of images have been improved by our methods. Moreover, in terms of visual effects and assessment metrics of underwater images, our method excels over other methods.