Medical image segmentation is critical in understanding pathological changes and computer-aided diagnosis. Most of the existing medical segmentation models focus on the overall segmentation effect of the model and lack of thinking about the problems of boundary blurring and model generalization ability. Given this, a bipartite segmentation model YNet based on boundary enhancement is proposed, which is based on the encoder-decoder architecture and consists of two core components: boundary enhancement module (BEM) and feature fusion module (FFM). The BEM utilizes the wavelet transform to separate the frequency domain information from the original image, allowing the model to dynamically adjust the boundary details in the original image, thus enhancing the model’s ability to perceive the boundary information and mitigating the effect of noise on the model. An attention mechanism is introduced in the FFM to enhance the model’s generalization ability by dynamically adjusting the channel and spatial information weights to emphasize critical features and suppress redundant information. Experimental results comparing other methods on CVC-ClinicDB, Kvasir-SEG, DSB2018, and ISIC2018 datasets show that the model has more explicit boundaries and better segmentation generalization. The source code of our YNet will be made available at https://github.com/DeadlyCodeGod/YNet.