<p>As an essential technology in the field of ocean engineering and research, underwater image enhancement plays a vital role in underwater vision tasks. However, the acquired underwater images inevitably suffer from degradation, including color casts and blurred edge details, due to absorption and scattering during light propagation in water. These issues seriously interfere with their application in high-level underwater vision tasks. While many models have been developed recently to address these degradation problems, challenges like color casts and blurred details remain. Furthermore, existing methods are limited in multi-path cooperation and edge preservation, both of which are essential for enhancing complex underwater scenes. In this paper, we present a dual-path cross-collaboration and edge information-aware network (DC-EINet), which aims to exploit the degraded image features to mine richer supervised information and enhance the network’s focus on color and edge detail. DC-EINet consists of a dual-path cross-collaboration branch and an edge information-aware branch. Specifically, we first design a dual-path cross-collaboration branch based on the encoder-decoder structure, which takes degraded images and their corresponding bilateral blurred images as inputs to learn diversity features. A wavelet-based multi-attention aggregated residual block is designed to improve the network’s capability to model both global background information and local texture detail information. To better capture detailed edge information, an edge information-aware branch is designed and integrated with the dual-path cross-collaboration branch. Extensive qualitative and quantitative experiments on both synthetic and real-world datasets demonstrate the effectiveness of the proposed method. Additionally, we conducted ablation studies to verify the contribution of the proposed components.</p>