U-shape feature crossover Transformer for underwater image enhancement

Due to the scattering and absorption of light in the underwater environment, underwater images often suffer from color distortion and loss of details. In order to solve these problems, the U-shape feature crossover Transformer for underwater image enhancement is proposed for achieving high-quality underwater image enhancement in this paper. Specifically, a feature crossover strategy is designed to improve the ability of the network to capture global context information. At the same time, the lightweight image reconstruction module (LIRM) is constructed to preserve image detail information to make image clearer. In addition, the dynamic semantic enhancement module (DSEM) is proposed to dynamically learn and extract the deep information of the networks which enhance image textures. Our proposed model solves the problems of low contrast and blurry images in underwater images. Experiments on UIEB and EUVP datasets show that the proposed UFCformer achieves the best results on PSNR and SSIM and obtains more satisfying visual effect compared to other underwater image enhancement methods.