基于注意力融合多尺度特征的解压缩点云质量增强方法

钟芯; 唐春明; 彭凌西

doi:10.12263/DZXB.20240914

您当前的位置：

首页 >

文章列表页 >

基于注意力融合多尺度特征的解压缩点云质量增强方法

学术论文 | 更新时间：2025-12-27

- 基于注意力融合多尺度特征的解压缩点云质量增强方法
- A Method for Enhancing the Quality of Decompressed Point Clouds Based on Attention-Fused Multi-Scale Features
- 电子学报 2025年53卷第8期页码：2794-2804
- 作者机构：
  
  1.广州大学计算机科学与网络工程学院，广东广州510000
  2.广州大学数学与信息科学学院，广东广州510000
  3.广州大学机械与电气工程学院，广东广州510000
- 作者简介：
  
  [ "钟芯女，1999年出生于广西壮族自治区梧州市.主要研究方向为机器学习、计算机视觉.E-mail: 2112206163@gzhu.edu.cn" ]
  [ "唐春明男，1972年9月出生于湖南省怀化市.教授，博士生导师.主要研究方向为密码学及其应用.E-mail: ctang@gzhu.edu.cn" ]
  [ "彭凌西男，1978年出生.教授，博士生导师.主要研究方向为人工智能技术及应用、网络安全." ]
- 基金信息：
- DOI：10.12263/DZXB.20240914
  中图分类号： TP391
- 收稿：2024-10-08，
  
  录用：2025-02-28，
  
  纸质出版：2025-08-25
- 稿件说明：
移动端阅览
钟芯, 唐春明, 彭凌西. 基于注意力融合多尺度特征的解压缩点云质量增强方法[J]. 电子学报, 2025, 53(08): 2794-2804.

ZHONG Xin, TANG Chun-ming, PENG Ling-xi. A Method for Enhancing the Quality of Decompressed Point Clouds Based on Attention-Fused Multi-Scale Features[J]. Acta Electronica Sinica, 2025, 53(08): 2794-2804.
钟芯, 唐春明, 彭凌西. 基于注意力融合多尺度特征的解压缩点云质量增强方法[J]. 电子学报, 2025, 53(08): 2794-2804. DOI：10.12263/DZXB.20240914

ZHONG Xin, TANG Chun-ming, PENG Ling-xi. A Method for Enhancing the Quality of Decompressed Point Clouds Based on Attention-Fused Multi-Scale Features[J]. Acta Electronica Sinica, 2025, 53(08): 2794-2804. DOI：10.12263/DZXB.20240914

摘要

基于几何的点云压缩算法（Geometry-based Point Cloud Compression，G-PCC）可以实现显著的点云压缩效率，但在低比特率场景下解压缩点云会产生严重的几何压缩伪影，并对整体视觉体验产生负面影响.为解决这一问题，本文提出了一种基于注意力融合多尺度特征的解压缩点云几何质量增强方法.具体地，该方法设计了多尺度输入模块对解压缩点云进行下采样操作，得到不同尺度的点云数据.接着，多尺度的点云被并行输入到离散卷积网络中提取从局部到全局的多尺度特征信息.最后，本文设计了跨尺度注意力特征融合模块来对多尺度特征进行融合，以增强特征的完整性和准确性.实验结果表明，本文所提出的方法在公开数据集上的平均峰值信噪比达到了67.968 4 dB，相较于标准压缩算法G-PCC提高了1.629 4 dB，主客观实验结果均表明本文方法能进一步提高解压缩点云的质量.

Abstract

Geometry-based point cloud compression (G-PCC) can achieve significant point cloud compression efficiency

but decompressing point clouds in low bit rate scenarios produces severe geometric compression artifacts and negatively affects the overall visual experience. To address this problem

this paper proposes a geometric quality enhancement method for decompressed point clouds based on attentional fusion of multiscale features. Specifically

the method designs a multi-scale input module to perform downsampling operations on the decompressed point cloud to obtain point cloud data at different scales. Then

the multi-scale point clouds are input in parallel into a discrete convolutional network to extract multi-scale feature information from local to global. Finally

a cross-scale attentional feature fusion module is designed in this paper to fuse the multi-scale features to enhance the completeness and accuracy of the features. The experimental results show that the proposed method achieves an average peak signal-to-noise ratio of 67.968 4 dB on the publicly available dataset

which is an improvement of 1.629 4 dB compared to the standard compression algorithm G-PCC

and the subjective and objective experimental results show that the method can further improve the quality of decompressed point clouds.

关键词

Keywords

references

WANG Q , KIM M K . Applications of 3D point cloud data in the construction industry: A fifteen-year review from 2004 to 2018 [J ] . Advanced Engineering Informatics , 2019 , 39 : 306 - 319 .

GUO K W , XU F , YU T , et al . Real-time geometry, albedo and motion reconstruction using a single RGBD camera [J ] . ACM Transactions on Graphics , 2017 , 36 ( 4 ): 1 .

PERRA C MURGIA F , GIUSTO D . An analysis of 3D point cloud reconstruction from light field images [C ] // 2016 Sixth International Conferenceon Image Processing Theory, Tools and Applications (IPTA) . Piscataway : IEEE , 2016 : 1 - 6 .

KRIVOKUCA M , CHOUP A , SAVIL P . 8 i Voxelized surface light field (8 iVSLF) dataset [EB/OL ] . ( 2021-01-19 )[20 24 - 10 -08 ] . https://mpeg-pcc.org/index.php/pcc-content-database/8i-voxelized-surface-light-field-8ivslf-dataset/ https://mpeg-pcc.org/index.php/pcc-content-database/8i-voxelized-surface-light-field-8ivslf-dataset/ .

V-PCC Codec Description . Document ISO/IEC JTC1/SC29/WG11 N19332 [S/OL ] . ( 2020-04-01 )[ 2024-10-08 ] . https://www.iso.org/committee/45316.html https://www.iso.org/committee/45316.html .

G-PCC Codec Description . Document ISO/IEC JTC1/SC29/WG11 N19331 [S/OL ] . ( 2020-04-01 )[ 2024-10-08 ] . https://www.iso.org/committee/45316.html https://www.iso.org/committee/45316.html .

ZHU W J , MA Z , XU Y L , et al . View-dependent dynamic point cloud compression [J ] . IEEE Transactions on Circuits and Systems for Video Technology , 2021 , 31 ( 2 ): 765 - 781 .

CHOY C , GWAK J , SAVARESE S . 4D spatio-temporal ConvNets: Minkowski convolutional neural networks [C ] // 2019 IEEE/CVF ConferenceonComputer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2019 : 3070 - 3079 .

JIA W , LI L , LI Z , et al . Deep learning geometry compression artifacts removal for video-based point cloud compression [J ] . International Journal of Computer Vision , 2021 , 129 ( 11 ): 2947 - 2964 .

XING J R , YUAN H , CHEN C , et al . Wiener filter-based point cloud adaptive denoising for video-based point cloud compression [C ] // Proceedings of the 1st International Workshop on Advances in Point Cloud Compression, Processing and Analysis . New York : ACM , 2022 : 21 - 25 .

AKHTAR A , GAO W , LI L , et al . Video-based point cloud compression artifact removal [J ] . IEEE Transactions on Multimedia , 2021 , 24 : 2866 - 2876 .

AKHTAR A , LI Z , AUWERA G V , et al . PU-dense: Sparse tensor-based point cloud geometry upsampling [J ] . IEEE Transactions on Image Processing , 2022 , 31 : 4133 - 4148 .

FAN X Q , LI G , LI D Q , et al . Deep geometry post-processing for decompressed point clouds [C ] // 2022 IEEE International Conference on Multimedia and Expo (ICME) . Piscataway : IEEE , 2022 : 1 - 6 .

LEE M K , KIM Y H . Distance weighted refining segmentation method for visual quality improvement in V-PCC [C ] // 2023 IEEE International Conference on Visual Communications and Image Processing (VCIP) . Piscataway : IEEE , 2023 : 1 - 4 .

SHENG X H , LI L , LIU D , et al . Attribute artifacts removal for geometry-based point cloud compression [J ] . IEEE Transactions on Image Processing , 2022 , 31 : 3399 - 3413 .

王韦韦 . 三维点云数据压缩与质量增强技术研究 [D ] . 济南 : 山东大学 , 2021 .

WANG W W . Research on 3D Point Cloud Data Compression and Quality Enhancement Technology [D ] . Jinan : Shandong University , 2021 . (in Chinese)

DING D D , ZHANG J Z , WANG J Q , et al . CARNet: Compression artifact reduction for point cloud attribu-te [EB/OL ] . ( 2022-09-17 )[ 2025-05-12 ] . https://arxiv.org/abs/2209.08276v1 https://arxiv.org/abs/2209.08276v1 .

LIU Q W , DING K , ZHANG C X , et al . A point cloud matching algorithm based on multiscale point pair features [C ] // 2023 IEEE International Conference on Real-time Computing and Robotics (RCAR) . Piscataway : IEEE , 2023 : 953 - 958 .

ZHANG R J , XUE Y Q , WANG J , et al . MSFA-net: A multiscale feature aggregation network for semantic segmentation of historical building point clouds [J ] . Buildings , 2024 , 14 ( 5 ): 1285 .

YE T , LIU A , YAN X P , et al . An efficient 3D point cloud-based place recognition approach for underground tunnels using convolution and self-attention mechani-sm [J ] . Journal of Field Robotics , 2025 , 42 ( 4 ): 1537 - 1549 .

WU Y , HU X D , ZHANG Y , et al . SACF-net: Skip-attention based correspondence filtering network for point cloud registration [J ] . IEEE Transactions on Circuits and Systems for Video Technology , 2023 , 33 ( 8 ): 3585 - 3595 .

LI Z L , BAO J W , LIU Y , et al . Complement decoded point cloud with coordinate adjustment for video-based point cloud compression [J ] . Signal, Image and Video Processing , 2024 , 19 ( 1 ): 48 .

MOENNING C , DODGSON N A . Fast Marching Farthest Point Sampling [R ] . Cambridge : University of Cambridge, Computer Laboratory , 2003 .

NIE D , LAN R , WANG L , et al . Pyramid architecture for multi-scale processing in point cloud segmentation [C ] // 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE , 2022 : 17263 - 17273 .

QIC R , YIL , SUH , et al . PointNet++: Deep hierarchical feature learning on point sets in a metric space [C ] // Proceedings of the 31st International Conference on Neural Information Processing Systems . California : NIPS , 2017 : 5150 - 5114 .

D’ EONE , HARRISONB , MYERST , et al . 8i voxelized full bodies: A voxelized point cloud dataset, document WG 11 M 40059 /WG 1 M 74006 [EB/OL ] . ( 2024-06-19 )[20 24 - 10 -08 ] . https://doi.org/10.48550/arXiv.2407.05677 https://doi.org/10.48550/arXiv.2407.05677 .

XU Y , LU Y , WEN Z Y . Owlii dynamic human mesh sequence dataset [EB/OL ] . ( 2021-01-05 )[ 2024-10-08 ] . https://mpeg-pcc.org/index.php/pcc-content-database/owlii-dynamic-human-textured-mesh-sequence-dataset/ https://mpeg-pcc.org/index.php/pcc-content-database/owlii-dynamic-human-textured-mesh-sequence-dataset/ .

CODINGM D G . G-PCC test model v14: ISO/IEC JTC1/SC29/WG7 output document N00094 [S/OL ] . ( 2021 - 01- 05 )[ 2024-10-08 ] . https://www.iso.org/committee/45316.html https://www.iso.org/committee/45316.html .

WATANABE S . Common test conditions for PCC: ISO/IEC JTC 1/SC29 WG 11 N 19084 [EB/OL ] . ( 2020-01-01 )[2 024 - 10 -08 ] . https://datatracker.ietf.org/group/iso-iec-jtc1-sc29-wg11/about/ https://datatracker.ietf.org/group/iso-iec-jtc1-sc29-wg11/about/ .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

面向时序异常检测的可变视距多向扫描方法

基于稀疏平滑自蒸馏的差分隐私深度学习方法

基于非一般类算子融合方法及硬件架构设计

基于深度压缩感知的联合信源信道编码方法研究