<p>Due to the complexity of underwater environments and the diversity of image degradation, underwater image enhancement (UIE) is an extremely challenging task. Existing single-stage networks are difficult to simultaneously solve multiple degradation problems. Therefore, this article proposes a two-stage deep learning framework called multi-information excitation network (MIE-Net), which uses cascaded contrastive learning to guide the training of each stage. In the first stage, raw images and reference images are used as negative samples and positive samples, respectively, to establish a contrastive loss for constraining the network training, ensuring that the color correction result is better than the input. In the second stage, the output of the first stage is used as negative samples to ensure that the final enhanced results of the second stage are better than the output of the first stage. Experiments demonstrate that our MIE-Net can achieve superior performance compared with many state-of-the-art methods. The ablation study validated the effectiveness of each key component.</p>