Abstract: Model skeleton is a 1D-representation of 3D models,which has been widely used in the areas of computer graphics and computer vision.In order to effectively extract high quality skeletons from the input meshes,a novel skeleton extraction algorithm is proposed which is based on the maximal inscribed sphere fitting.Firstly,the user can select one point in the main trunk of the given model and one point at the end of each branch,which can be considered as initial points for skeleton extraction;Secondly,the initial inscribed sphere can be generated using our maximal inscribed sphere extraction algorithm and its center can be selected as skeleton point.Iteratively spreading the fitting ball along the main trunk,the algorithm can also find the rest skeleton points and setting the area around the main skeletons as the impact domain.Then,from the picked point at the end of each branch,our skeleton point selection method can be adopted to find the branch skeleton points along the end point towards the main trunk.Finally,the appropriate skeleton convergence point is selected using distance-first algorithm for connecting the skeleton of main trunk and branches.Taking the models provided by AIM@SHAPE as examples,the skeleton points selected by our algorithm always locate near the model medial axes which demonstrates that our approach can efficiently extract the smooth skeletons of the given 3D mesh models.