This paper addresses the lack of a comprehensive and real-world underwater image dataset for evaluating underwater image enhancement algorithms. To bridge this gap, the authors construct the first large-scale real-world underwater image benchmark dataset, named Underwater Image Enhancement Benchmark (UIEB), containing 950 real underwater images, 890 of which have corresponding reference images. The remaining 60 images are challenging data due to the difficulty in obtaining satisfactory reference images. Using this dataset, the authors conduct a comprehensive study of state-of-the-art underwater image enhancement algorithms, both qualitatively and quantitatively. Additionally, they propose Water-Net, a deep learning-based underwater image enhancement model trained on the UIEB, to demonstrate the generalization of the dataset for training Convolutional Neural Networks (CNNs). The evaluations and the proposed model highlight the performance and limitations of current algorithms, providing insights for future research in underwater image enhancement. The dataset and code are available online.This paper addresses the lack of a comprehensive and real-world underwater image dataset for evaluating underwater image enhancement algorithms. To bridge this gap, the authors construct the first large-scale real-world underwater image benchmark dataset, named Underwater Image Enhancement Benchmark (UIEB), containing 950 real underwater images, 890 of which have corresponding reference images. The remaining 60 images are challenging data due to the difficulty in obtaining satisfactory reference images. Using this dataset, the authors conduct a comprehensive study of state-of-the-art underwater image enhancement algorithms, both qualitatively and quantitatively. Additionally, they propose Water-Net, a deep learning-based underwater image enhancement model trained on the UIEB, to demonstrate the generalization of the dataset for training Convolutional Neural Networks (CNNs). The evaluations and the proposed model highlight the performance and limitations of current algorithms, providing insights for future research in underwater image enhancement. The dataset and code are available online.