To solve the problems that the color of colonic polyps is similar to the background color, and the different sizes of polyps affect the segmentation effect, this paper proposes an improved image segmentation network for colonic polyps with a double U-shaped structure. The proposed model is based on the DoubleU-Net architecture. Firstly, the spatial attention block is integrated into the skip connection of the U-shaped structure to extract the correlation information of spatial features. Secondly, the channel attention block is incorporated into the skip connection to clearly express the dependency of useful channels and suppress the features unrelated to the polyp segmentation task. Finally, at the input of the encoder with the second U-shaped structure, the selective kernel block is introduced to adaptively select different receptive fields and improve the target segmentation accuracy. The experimental results show that the proposed method is better than the existing methods in terms of objective indices and visual effects. The research results can provide new reference for the early detection and surgical planning of colonic polyps.