RGB Attention-Aware Enhancement for Underwater Image Enhancement

Underwater image enhancement is an important tool for acquiring deep-sea resources, marine environment monitoring, submarine pipeline maintenance and other fields. Due to the special characteristics of underwater environment, underwater images often suffer from incomplete image information such as blurring, noise, and uneven illumination, which seriously limits the value of underwater images in scientific research and engineering applications. To address these issues, this paper proposes the RGB attention-aware enhancement conditional generative adversarial network for underwater image enhancement. The network learns the end-to-end mapping from input images to enhanced outputs. A key innovation in this paper is the RGB attention-aware enhancement module incorporated into the generator. This module effectively extracts features and adjusts the contrast and saturation of underwater image colors. Meanwhile, L1 loss function and adversarial loss function are proposed in this paper, which can focus on ground truth image features in a targeted way. In this study, we use numerous comparative tests to confirm the superiority of the suggested strategy in real underwater image enhancement. The qualitative and quantitative evaluation results demonstrate that the proposed method outperforms other advanced techniques in enhancing underwater images. Moreover, it exhibits superior generalization performance. It can also adapt to the needs of underwater image enhancement in a variety of scenarios and conditions.