LALG: Underwater image enhancement via luminance-aware variational color correction and joint local-global contrast restoration

Abstract

Underwater image enhancement collected by optical electronic devices often suffers from low contrast, blurry details and color casts due to the interference of scattering and absorption effects, most of current methods enhance the underwater image by designing specialized priors on the illumination and reflectance. Nevertheless, these methods may still result in incomplete color correction and low illumination. To deal with these poor issues, we propose an underwater image method based on luminance-aware variational color correction and joint local-global contrast restoration. Concretely, we propose a luminance-aware variational framework to correct the color casts, which designs an adaptive luminance-aware factor based on the illuminance of underwater scenes, and provides an augmented Lagrange multiplier based alternating direction minimization to solve the variational optimization problem, guiding the correction and illumination improvement simultaneously. In addition, we propose a local contrast enhancement based on the local characteristic of the mean value and the variance value, which can better adapt to the local differences of contrast loss of underwater image. Meanwhile, a simple global contrast is proposed to adaptively compensate the contrast intensity. Comprehensive comparative results on three underwater image enhancement benchmarks demonstrate that our method has a superior performance in revising the color, enhancing the details and improving the contrast of underwater images. It also can produce more excellent quantitative values than state-of-the-art methods, i.e., compared with the second-best method, the average PCQI and $\overline{r}$ values of our method at least improve by 1.70% and 28.14%, respectively, while the average Blur value of our method at least decline by 11.33%, respectively.