As one of the most important tasks in image processing, Image registration aims to find a plausible transformation to match the corresponding data, which has many applications in computer vision, remote sensing and medical imaging. In this talk, I will first introduce the registration model, especially the diffeomorphic registration model with quasi conformal theories. Then I will talk about the recent work, the registration-based segmentation method, including topology-preserving segmentation method, convexity-preserving segmentation method and star-shape-preserving segmentation method.