TY - GEN
T1 - PU-SSIM
T2 - 20th International Forum on Digital TV and Wireless Multimedia Communications, IFTC 2023
AU - Huang, Tiangang
AU - Wang, Xiaochuan
AU - Liu, Ruijun
AU - Li, Haisheng
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.
PY - 2024
Y1 - 2024
N2 - Point cloud data acquired through scanning typically exhibits sparse, non-uniform distribution, and a certain level of noise. Therefore, it is necessary to generate a dense and high-quality point cloud via up-sampling. In recent years, point cloud up-sampling techniques gain significant advantages due to the development of deep learning. In particular, most of current end-to-end up-sampling networks adopt point-wise constraints, e.g., Chamfer distance to train the up-sampling model. However, these point-wise constraints are inadequate to reduce residual noise, meanwhile would induce structural distortions. To further improve the capability of up-sampling networks, we propose a perception-wise constraints, namely PU-SSIM. Specifically, we adopt the typical full-reference point cloud quality metric to measure the structural similarity between the generated high-resolution point cloud and the ground truth. We managed to embed it into the up-sampling network, providing a plug-in capability. The experimental results indicate that the PU-SSIM can maintain the structural details, meanwhile reduce the residual noises. The proposed perception constraint is compatible to most mainstream methods, which would benefit the community to some extend.
AB - Point cloud data acquired through scanning typically exhibits sparse, non-uniform distribution, and a certain level of noise. Therefore, it is necessary to generate a dense and high-quality point cloud via up-sampling. In recent years, point cloud up-sampling techniques gain significant advantages due to the development of deep learning. In particular, most of current end-to-end up-sampling networks adopt point-wise constraints, e.g., Chamfer distance to train the up-sampling model. However, these point-wise constraints are inadequate to reduce residual noise, meanwhile would induce structural distortions. To further improve the capability of up-sampling networks, we propose a perception-wise constraints, namely PU-SSIM. Specifically, we adopt the typical full-reference point cloud quality metric to measure the structural similarity between the generated high-resolution point cloud and the ground truth. We managed to embed it into the up-sampling network, providing a plug-in capability. The experimental results indicate that the PU-SSIM can maintain the structural details, meanwhile reduce the residual noises. The proposed perception constraint is compatible to most mainstream methods, which would benefit the community to some extend.
KW - deep geometric learning
KW - perceptual constraint
KW - point cloud quality assessment
KW - point cloud up-sampling
UR - https://www.scopus.com/pages/publications/85198940022
U2 - 10.1007/978-981-97-3623-2_25
DO - 10.1007/978-981-97-3623-2_25
M3 - 会议稿件
AN - SCOPUS:85198940022
SN - 9789819736225
T3 - Communications in Computer and Information Science
SP - 345
EP - 358
BT - Digital Multimedia Communications - 20th International Forum on Digital TV and Wireless Multimedia Communications, IFTC 2023, Revised Selected Papers
A2 - Zhai, Guangtao
A2 - Zhou, Jun
A2 - Yang, Hua
A2 - Ye, Long
A2 - An, Ping
A2 - Yang, Xiaokang
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 21 December 2023 through 22 December 2023
ER -