Comparison of Single Image Enhancement Techniques for Underwater Color Imaging

Abstract

Underwater color images suffered from low visibility and color cast effect because of light attenuation by water and the floating particles. In order to enhance the quality of underwater images, we applied single image enhancement techniques and compared their performances with real underwater images taken in Korean sea. In this study, we considered dark channel prior-based, gradient-based, image fusion-based, and underwater generative adversarial network (UGAN) methods for single image enhancement. And the performances were evaluated in terms of underwater image quality measure (UIQM) and gray-world assumption. Image fusion and UGAN showed reasonable qualitative and quantitative performances enhancing the visibility and colorfulness.