Lossless Point Cloud Data Compression on a GPU

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Final Year Thesis Oral Defense

Title: "Lossless Point Cloud Data Compression on a GPU"

By

REN Zhengtong

Abstract:

Point clouds are sets of 3D points generated by cameras, scanners, or 
sensors describing real-world objects such as sculptures, buildings, or 
even entire cities. Since point clouds nowadays contain millions of points 
with high precision, compression is required for their efficient storage 
and transmission. In this project, we developed a new compression scheme 
for point clouds. Using relatively simple algorithms such as differential 
encoding, minimum spanning tree, and topological sorting, our scheme can 
compress a point cloud to around 60% of its original size without any loss 
of precision. With our GPU-based parallel implementation, compression of a 
point cloud with ten million points takes just over ten seconds. 
Experimental results such as these show that our new compression scheme is 
effective and efficient especially on point clouds that are large and 
precise.


Date            : 3 May 2022 (Tuesday)

Time            : 16:00-16:40

Zoom Link:
https://hkust.zoom.us/j/98805735082?pwd=V2psTk5DckpjRW8zOW8rOG9Ca2Q0Zz09

Meeting ID      : 988 0573 5082

Passcode        : 572269

Advisor         : Prof. LUO Qiong

2nd Reader      : Prof. YI Ke