LiDAR Point Cloud Compression Improvement through Sampling Operations

Chang Duan, Chen Liao, Shun Wang, Quan Zhang, Chao Zhang, Ce Zhu

Abstract: LiDAR point cloud data has unique characteristics, and its compression method is also different from that in image/video compression. Most existing lossy encoding methods often lead to both a significant reduction in the number of point and a loss of accuracy. Specifically, the decrease in total number is due to down sampling operations during the encoding process. In fact, how to take advantage of sampling operations to further improve the coding efficiency of LiDAR point cloud data has not yet been discussed. This article proposes the equidistant segmented sampling method on preprocessing. After decoding, up sampling is performed to increase the total number of points. We conduct R-PCC method on the KITTI dataset, D1PSNR is used to measure encoding distortion. Our method achieved a 36.32% BD-Rate improvement compared to conventional R-PCC.