A compact representation for 3D animation using octrees and affine transformation

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT AUTHOR'S REQUEST.] We present a new and compact 3D representation for non-rigid objects using motion vectors between two consecutive frames. Our method relies on an Octree to recursively partition the object into smaller parts. Each part is then assigned a small number of motion parameters that can accurately represent that portion of the object. The partitioning continues as long as the respective motion parameters are insufficiently accurate to describe the object. Our method employs an affine transformation as the motion vectors. A technique using adaptive thresholding, singular value decomposition for dealing with singularities, and a quantization and arithmetic coding further enhance our proposed method by increasing the compression while maintaining very good signal-noise ratio. Besides the work we have done for synthetic data (animation), we also challenge a much more difficult problem - the motion representation for real data (cloud of points), where the correspondence is unknown. We applied Iterative Closest Points (ICP) algorithm for computing a pseudo correspondence, combined with an Octree structure to deal with the non-rigidity that ICP can not capture. About the motion vectors, we still use the affine transformation as we did for animation data. Even though our result for this part is not strong, we give a detailed analysis for the failure and proposed several possible solutions.