Deformation-Graph-Driven And Deformation Aware Spectral Pose Transfer

Mesh deformation is often used to edit 3D models to build animation sequences,or to transfer one's animation to another in computer animation.As an example-based deformation method,deformation transfer applies the deformation of given sequences to other models to obtain new sequences,which is also of great significance to expand 3D dataset since it can generate new models.Different from deformation transfer,pose transfer clones the pose of reference model to a source model and obtains a deformed shape.It's a challenging task due to missing deforming information.In order to improve the ability of preserving shape details and reduce user interaction,this paper proposes a deformation graph driven and deformation aware spectral pose transfer method.First,deformation graph is used to drive the source model deforming under the guidance of low-frequency spectral coefficients of the reference model,then sub-model is automatically segmented according to the deformation before and after pose transfer,and the pose transfer is performed on sub-model until different scale poses are transferred.The main contributions of this paper are as follows: 1)Take advantage of the local control properties of deformation graph,a low-frequency spectral pose transfer method based on deformation graph is proposed,which not only can improve the shape maintaining ability but also reduce the number of hierarchical spectral pose transfer;2)Propose an automatic method of detection and segmentation of sub-model.This method segments sub-model automatically according to the deformation,which can reduce manual operation to a certain extent;3)Optimize the method of sub-model's pose transfer and splicing,which will not lead to the orientation change in the process of pose transfer.In sub-model splicing stage,the result of low-frequency pose transfer is deformed to obtain sub-model's pose,so as to ensure the natural transition at the junction.Experimental results show that the proposed method improves the quality of transfer results to a certain extent,reduces the number of middle-frequency pose transfer,and also reduces manual interaction.