Binary Representation for 3D Point Cloud Compression based on Deep Auto-Encoder

Deep learning on a three dimensional (3D) point cloud has attracted a great deal of attention. In this paper, we propose a binarization method for latent representation in a 3D point cloud auto-encoder. The proposed method is based on z-score normalization and deterministic binarization. In the experiment, the proposed method increased data compression efficiency 32-fold, while slightly decreasing the generalization ability of the point cloud reconstruction.