Abstract: The captured data of a time-of-flight(TOF) camera can't be directly used due to its low image resolution and high noise level.Therefore,this paper presents an algorithm for 3D reconstruction of non-rigid shapes from one TOF camera.First,the captured data from different views are segmented and denoised.Second,the constructed meshes are roughly aligned with rigid alignment.Feature frames are selected to reduce motion blur and redundancy.Then the feature frames are further aligned with global non-rigid alignment to deal with misalignment of static objects with non-rigid errors and human bodies under low-frequency non-rigid deformation.Finally,a qualified mesh is generated using Poisson surface reconstruction.Our experiments show that the algorithm can improve the quality of captured depth data and extend the typical application of TOF cameras.