Underwater Image Enhancement by Gaussian Curvature Filter
Jiaying Xiong, Yuxiang Dai, Peixian Zhuang · 2019
We develop a new Retinex-based variational model for enhancing single underwater image by imposing Gaussian curvature priors on the illumination and the reflection. Gaussian curvature filters are employed to estimate the illumination and the reflection efficiently for solving the proposed variational model. And Gaussian curvature filter can capture better underwater image details and prevent image over-enhancement. In addition, Gaussian curvature filter can reduce the runtime of underwater image enhancement without calculating both partial derivative operations and the gradient of overall energy functional. Ultimately, we provide numerous experiments to demonstrate the effectiveness of the proposed method, which outperforms several underwater enhancement algorithms in terms of enhancement performance, visual improvements and runtime.