The Analysis And Improvement Of Triangle Mesh Correspondence And Deformation Algorithms

Computer animation is one of the most interesting fields in mesh related domain. The procedure for generation of 3D mesh animation follows: 1) generate original mesh data by 3D scanner or just a camera, and carry out pretreatment, such as filling hole in meshes, do simplification, subdivision and optimization for meshes; 2) edit key frames for animation; 3) do interpolation for key frames. Here, we can use skeleton method to help simplify the generation of key frame. But it is not always an easy job to configure so many parameters in the skeleton. After a long time hard working, we get some accumulation of mesh deformation series. When facing a similar mesh, we can use mesh deformation method, which can reuse original deformation by transferring any deformation of a source triangle mesh onto a different target mesh. However, it is still time consuming because you must find the correspondence of meshes, which may include the selection for correspondence feature vertices in both meshes. And then, lots of numerical computing steps will be done, which may also be time consuming. Therefore, we try to find out an effective way to solve mesh correspondence and deformation transfer, which means less user intervention and less computation.In this paper, we propose a correspondence method based on triangle meshes, and carry out the deformation transfer by solving a gradient equation. Our method is better than the traditional method in many aspects: First of all, candidates for feature vertices can be generated automatically, so it will be not so tired to select the correspondence vertices. Secondly, intermediate deformation method has been proposed to map both meshes into a near space to make the correspondence easier. Then it comes the space hash grid structure, which help to make the searching for the nearest point or face in a constant time. Finally, we try our best to avoid unnecessary computing and simplify the solving of sparse matrix for Laplacian-based optimization method and the gradient equation for deformation transfer. Experiments demonstrate how this approach can easily produce a correct correspondence between different meshes and generate a good result by deformation transfer.