<p>Due to the complexity of the underwater imaging environment, underwater images experience severe visual degradation, such as color bias, low contrast, and blurred details. Recently, underwater image enhancement algorithms have been proposed to solve these problems. However, these methods have limitations in recovering high-frequency details and suppressing noise. In this paper, we propose a novel underwater image enhancement network with embedding frequency and Haar wavelet, termed FHW-Net. Firstly, we embed the Fourier transform into the encoder and propose a frequency domain and spatial domain feature extraction module that extracts features from both domains and separately processes the amplitude and phase of underwater images to avoid introducing noise during enhancement. Meanwhile, a Haar wavelet downsampling is utilized to replace the traditional pooling operation to reduce information loss. Then, a multi-scale cross-axis attention module is deployed in the bottleneck layer to learn multi-scale contexts and emphasize critical features. Finally, in the decoder, we propose a cross-spatial learning feature enhancement module to fuse the features of different levels and further recover image color and detail. The proposed FHW-Net is extensively evaluated on popular datasets, and the experimental results show that our method achieves competitive performance in terms of color restoration and detail preservation. The source code is available at <a href="https://github.com/jin123f/FHWNet">https://github.com/jin123f/FHWNet</a>.</p>