Underwater image enhancement combining low-dimensional and global features

The physical transformation of light will cause the quality of underwater images to decrease, which impacts the precision of object detection, recognition, and segmentation in underwater circumstances. In this study, we suggest an underwater image enhancement combining low-dimensional and global features (UIELG). This model can availably heighten the texture minutiae and global key features of subaquatic images. Moreover, we advise a new loss function, which can effectively boost the structure and texture similarity of underwater images. Finally, we train and test the model on the synthetic subaquatic images. The experimental outcomes declare that this model is preferable to the existing models in both SSIM and PSNR scores. And the experimental outcomes on the real-world subaquatic image dataset present the generalization and robustness of the suggested model.