A SURVEY ON POINT CLOUD UPSAMPLING

PhD Qualifying Examination


Title: "A SURVEY ON POINT CLOUD UPSAMPLING"

by

Mr. Zeyu HU


Abstract:

Acting as a preprocessing step, high quality upsampling from sparse 3D 
point clouds is critically useful for a wide range of downstream geometric 
operations such as reconstruction, rendering, meshing, and analysis. Given 
an unordered set of 3D points, the aim of this task is to generate a 
denser point set that lies on the underlying surface. Due to the 
irregularity and sparsity of the 3D point cloud, it is far from 
straightforward to generate new points that remain true to the underlying 
surfaces with complex geometric and topological structures. Although point 
cloud upsampling shares the spirit with the image super-resolution 
problem, naive adaption of image-space techniques to point sets is 
prevented by the different natures of images and point clouds.

In this survey, we broadly classify the existing point cloud upsampling 
methods into two categories: optimization-based upsampling and deep 
learning-based upsampling, and review the related techniques in each 
category. For optimization-based upsampling, we introduce several 
representative methods utilizing different shape priors. For deep 
learning-based upsampling, we discuss the state-of-the-art works, which 
roughly follow the same path and are closely related.  We conclude by 
discussing current challenges in point cloud upsampling and provide two 
interesting research directions for future work.


Date:                   Monday, 11 May 2020

Time:                   4:00pm - 6:00pm

Zoom meeting:           https://hkust.zoom.us/j/4715932509

Committee Members:      Prof. Chiew-Lan Tai (Supervisor)
                        Prof. Chi-Keung Tang (Chairperson)
                        Dr. Xiaojuan Ma
                        Dr. Pedro Sander


**** ALL are Welcome ****