How to Rapidly Modify and Download Huge 3D Meshes

Progressive compression is a multi-resolution scheme for sending large 3D meshes in a network environment. However, if 3D meshes are deformed, it takes too much time to compress them from scratch. Wavelet-based fast recompression speeds up the compression time by up to 33 times.

Article | Special Issue

Imagine that users are downloading a huge 3D model for 3D printing on a website. If users are not satisfied with the model and want to change it, they have to wait a long time again to download the modified model. Large data on the external device would be burdensome for the users due to the unavoidable time for data transmission, including the required compression and decompression process.

Prof. Sunghee Choi and her student Yeonghun Kim in the School of Computing at the Korea Advanced Institute of Science and Technology (KAIST) developed a wavelet-based progressive compression system, named “Recompressor”, to greatly reduce the total downloading time, even in the case of frequent deformation of 3D large mesh models. Their work supports the connectivity-preserving deformation of huge 3D triangular meshes with up to 10 million faces, boosting the additional compression speed by up to 33 times.

Their ‘propagator’ is the core part that can imitate the original compression process. Given displacement vectors between older and newer version of mesh, the propagator generates a new compressed file corresponding to the new mesh, overwhelmingly faster than before using its propagation algorithm. Surprisingly, the result file is not an approximation but is exactly same as the one through the original compression process.

The key idea is mesh wavelets, which can be defined as having detailed signal representation information for the mesh. Prof. Choi’s group exploits the following two main properties of mesh wavelets: 1) they are highly independent to vertex movements and 2) they have detailed information which lead to higher mesh compression ratio. By utilizing the theoretical supports, they successfully designed a propagation algorithm and overcame the trade-off between compression ratio and speed. Moreover, their system provides two important advantages for huge meshes: 1) the system supports a lossless way keeping all the details of mesh after compression and 2) the system makes meshes easy to visualize due to the smaller data size of lower resolutions.

This research was published in Computer-Aided Design in August, 2020 under the title of “Wavelet-based progressive fast recompression for large deformed mesh” (DOI: 10.1016/j.cad.2020.102859).