Research on Image Segmentation Method Based on U-net Improvement

Image segmentation is a crucial task in the domain of computer vision, aiming to partition an image into multiple connected regions with similar characteristics. This paper focuses on extracting tooth contours automatically from dental medical images through an improved U-net network to enhance dentists’ working efficiency and assist physicians in more accurate diagnosis and treatment. The proposed improvement approach involves a series of modifications based on the U-net, such as introducing the CBAM attention module and integrating the ability of Contextual Transformer (CoT) to incorporate global context. This enhances the performance of U-net in image segmentation tasks, allowing the network to capture and analyze the complex features in medical images more delicately and comprehensively, achieving precise and efficient automatic segmentation of medical images.