Research On The Implicit Shape Registration For Dynamic Three-dimension Model

The research on the dynamic three-dimension model is a novel field which describe the dynamic models (human, animal, car, machine, clothing and plant) in the real world with three-dimension geometry models. With the help of dynamic three-dimension models, we can notably improve the ability of computer graphics for exhibiting the real world. Shape Registration is finding a coherent mapping between two or more coordinate systems such that corresponding points of models in the different systems are mapped to each other. It has become a key ingredient in many mesh operations, such as acquisition, reconstruction, and editing. The registration depends on many factors: 1) method how to represent shapes, 2) nature of the transformation to move the points from the source towards the target. 3) dissimilarity measure. We present an automatic implicit approach for pair-wise non-rigid registration of partially overlapped 3D shapes. Shapes of interest are implicitly embedded in a three-dimensional vector distance space. A novel global rigid alignment algorithm is proposed, which could calculate pure global rigid transformation for both static objects and similar dynamic objects with different orientation and translation. Based on IFFD deformation, a local non-rigid registration is proposed, which can produce a smooth and continuous transformation. Our method could consistently preserve shape topology during the deformation, produce smooth and continuous local registration fields, and correctly establish dense one-to-one correspondences. In addition, it naturally deals with noise time-varying points set even with missing data, since it directly make use of implicit surface representation. Experimental results on several synthetic and real-world datasets demonstrate the robustness of the proposed algorithm.