Point Cloud Inpainting Based on Delaunay Triangulation

Point cloud is a point-based representation of three-dimensional data used to describe the geometry of objects or scenes However, due to its large data volume, a point cloud needs to be compressed to reduce the data volume. After point cloud compression, some points may be lost, causing the loss of...

Full description

Saved in:
Bibliographic Details
Published in2023 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) pp. 1525 - 1529
Main Authors Liu, Yu-Lin, Chou, He-Sheng, Lee, Ming-Zhan, Chan, Mei-Ling, Lin, Ting-Lan, Chen, Chiung-An, Chen, Shin-Lun
Format Conference Proceeding
LanguageEnglish
Published IEEE 31.10.2023
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:Point cloud is a point-based representation of three-dimensional data used to describe the geometry of objects or scenes However, due to its large data volume, a point cloud needs to be compressed to reduce the data volume. After point cloud compression, some points may be lost, causing the loss of edge details and the appearance of point cloud holes and outliers, leading to a significant subjective quality decrease and affecting point cloud development in practical applications. In this paper, an effective point cloud repair method is proposed. First, use the mean filter to remove outliers, so that the subsequent point-filling algorithm can avoid errors caused by outliers. Secondly, the point cloud is analyzed by the Delaunay triangulation method. Perform statistical analysis on the areas of all triangles. If an individual triangle has an area larger than the average, it is marked. After comparing the triangles, the coordinates of all the holes in the point cloud can be found, so as to identify the holes in the point cloud. In the hole-filling strategy, the interpolation method is used to average the coordinates and color information of the three vertices of the triangle to obtain the filling point, that is, to complete the restoration of the point cloud data. When evaluating the results, in addition to PSNR, the GraphSim index proposed by Q. Yang was also referred to [1]. This indicator is closer to the subjective evaluation of people and better reflects the subjective evaluation of people. According to the experimental results, the GraphSim of the proposed method can reach 0.363, which significantly improves the point cloud data after V-PCC compression [2].
ISSN:2640-0103
DOI:10.1109/APSIPAASC58517.2023.10317549