Abstract: Chronic wounds not only seriously affect the quality of life of patients,but also prone to deterioration.To ensure the accuracy of wound image segmentation and improve the image-based wound analysis,a U-Net model based on dense connections is proposed for wound segmentation in this paper.The dense connection mechanism employs skip connections.The outputs of different stages in the encoder are aggregated to the decoder by the dense connection mechanism.Within each layer of the decoder,the Multi-view Feature Adaptive Fusion module (MFAF) is utilized to adaptively fuse the skip connection features from each layer.The effectiveness of the proposed method is validated on public clinical datasets.Comparative experimental results show that the DSC,MIoU,HD95,Recall,Voe and Rvd metrics of the proposed method for wound segmentation are 82.84%,74.06%,2.66%,85.10%,74.31% and 73.12%,respectively.Moreover,the segmentation accuracy is effectively improved for the cases with unclear wound edges.