Normal-Based Global Motion Estimation for LiDAR Point Cloud Lossless Geometry Compression

Point clouds generated by Light Detection And Ranging (LiDAR) are essential for automatic driving. Compression of the LiDAR point cloud faces the challenge of accurate global motion estimation due to the different characteristics of objects in the scene. In this paper, we propose an adaptive global motion estimation algorithm based on normals. First, preprocessing is performed on the point cloud to eliminate points unsuitable for providing global motion. Then, the corresponding point is found according to the different angles of the plane where the object is located. Finally, the global motion estimation is performed according to the correspondence. Experimental results show that the proposed method outperforms the Inter-EM of G-PCC. It can provide an average of 0.24% saving in coding bits on the G-PCC standard dataset and 0.63% saving on the KITTI dataset.