Abstract: This paper presents a series of new methods for human mandible reconstruction,based on the characteristics of CT data.Our work involves medical image segmentation and surface reconstruction algorithms.Firstly,a new segmentation method namely “no-seed region segmentation” is employed to extract the planar contours. Then,a bow “reference curve” is created to show the shape trend of mandible.The “normalized position” of the contours on the curve is used to solve the correspondence and branching problem in surface reconstruction.And influence from previous layer is considered to help fully produce smooth surface.Finally,tiling is implemented by “mating points at mult-i order convex hulls”,and the iteration is done for several steps.