3D point cloud has been widely applied in virtual reality and augmented reality. A complex 3D scene always needs a large number of point cloud to represent and demands a lot of space to store. Thus, point cloud compression becomes a crucial issue to research. An end-to-end optimized lossy geometry compression model for 3D point clouds geometric information is proposed. The model solves the problem of missing large area point cloud by using autoencoder and deep convolution generative adversarial network algorithm and reconstructs high-quality point cloud data. In order to improve the reconstruction quality of the high point cloud and ensure that the compression efficiency does not decrease, a deconvolution jump connection structure is proposed in this paper. This structure transmits the information of each layer of the decoder to the last output layer for feature fusion, and finally, the decoder can reconstruct a high-quality point cloud at lower bits rate. Experimental results show that compared with MPEG G-PCC (Octree) standard, the proposed model can decrease BD-BR by 62.01% and increase BD-PSNR by 4.497 1 dB on the MVUB dataset, and obtain higher visual quality.
在评价实验中使用MVUB的体素化点云数据集。原始点云和重建点云可能有不同数量的点。例如,在本文模型中,重建点云比原始点云多出大量点,而在MPEG G-PCC的方法中则相反。点到面对称PSNR(point to plane symmetric peak signal to noise ratio,PPS-PSNR)[20]比标准PSNR更加客观和准确。因此,本文测量每个占用体素的码率,并使用PPS-PSNR评估点云帧质量,PPS-PSNR计算公式如下
CAOC, PREDAM, ZAHARIAT. 3D point cloud compression: A survey [C]//The 24th International Conference on 3D Web Technology. New York: ACM, 2019:1-9 . DOI:10.1145/3329714.3338130 .
[2]
YANGY X, CHENX M, HANY H. Dadu-CD: Fast and efficient processing-in-memory accelerator for collision detection [C]//2020 57th ACM/IEEE Design Automation Conference. New York: IEEE Press, 2020: 1-6. DOI:10.1109/DAC18072.2020.9218709 .
[3]
SCHWARZS, PREDAM, BARONCINIV, et al. Emerging MPEG standards for point cloud compression[J]. IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 2019, 9(1): 133-148. DOI:10.1109/JETCAS.2018.2885981 .
LIJ Y, HUQ W, AIM Y. Point cloud registration based on one-point RANSAC and scale-annealing biweight estimation [J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(11): 9716-9729. DOI:10.1109/TGRS.2020.3045456 .
[6]
LIUH J, LIAOK, LINC Y, et al. Pseudo-LiDAR point cloud interpolation based on 3D motion representation and spatial supervision [J]. IEEE Transactions on Intelligent Transportation Systems, 2021,6048: 1-11. DOI:10.1109/TITS.2021.3056048 .
[7]
ZHOUH, FANGZ J, GAOY B, et al. Feature fusion network based on attention mechanism for 3D semantic segmentation of point clouds [J]. Pattern Recognition Letters, 2020, 133: 327-333. DOI:10.1016/j.patrec.2020.03.021 .
[8]
HUANGT X, LIUY. 3D point cloud geometry compression on deep learning [C]// Proceedings of the 27th ACM International Conference on Multimedia. New York: ACM, 2019: 890-898. DOI:10.1145/3343031.3351061 .
[9]
WENX Z, WANGX, HOUJ H, et al. Lossy geometry compression of 3D point cloud data via an adaptive octree-guided network [C]// 2020 IEEE International Conference on Multimedia and Expo (ICME). New York: IEEE Press, 2020: 1-6. DOI:10.1109/ICME46284.2020.9102866 .
[10]
MATSUZAKIK, TASAKAK. Binary representation for 3D point cloud compression based on deep auto-encoder[C]// 2019 IEEE 8th Global Conference on Consumer Electronics (GCCE). New York: IEEE Press, 2019: 489-490. DOI:10.1109/GCCE46687.2019.9015550 .
[11]
GUARDAA F R, RODRIGUESN M M, PEREIRAF. Deep learning-based point cloud coding: A behavior and performance study [C]// 2019 8th European Workshop on Visual Information Processing (EUVIP). New York: IEEE Press, 2019: 34-39. DOI:10.1109/EUVIP47703.2019.8946211 .
[12]
TUC X, TAKEUCHIE, CARBALLOA, et al. Point cloud compression for 3D LiDAR sensor using recurrent neural network with residual blocks [C]// 2019 International Conference on Robotics and Automation (ICRA). New York: IEEE Press, 2019: 3274-3280. DOI:10.1109/ICRA.2019.8794264 .
[13]
QUACHM, VALENZISEG, DUFAUXF. Learning convolutional transforms for lossy point cloud geometry compression [C]// 2019 IEEE International Conference on Image Processing (ICIP). New York: IEEE Press, 2019: 4320-4324. DOI:10.1109/ICIP.2019.8803413 .
[14]
YUANC K, LIK, LAIY K, et al. 3D face reprentation and reconstruction with multi-scale graph convolutional autoencoders [C]// 2019 IEEE International Conference on Multimedia and Expo (ICME). New York: IEEE Press, 2019: 1558-1563. DOI:10.1109/ICME.2019.00269 .
[15]
BALLÉJ, MINNEND, SINGHS, et al. Variational Image Compression with a Scale Hyperprior [DB/OL].[2021-10-12].
[16]
MINNEND, BALLÉJ, TODERICIG. Joint autoregressive and hierarchical priors for learned image compression [C]// Proceedings of the 32nd International Conference on Neural Information Processing Systems (NIPS'18). Red Hook:Curran Associates Inc, 2018:10794-10803.
[17]
LINT Y, GOYALP, GIRSHICKR, et al. Focal loss for dense object detection [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2):318-327. DOI:10.1109/TPAMI.2018.2858826 .
[18]
WUZ R, SONGS R, KHOSLAA, et al. 3D ShapeNets: A deep representation for volumetric shapes [C]// 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE Press, 2015:1912-1920. DOI:10.1109/CVPR.2015.7298801 .
[19]
LOOPC, CAIQ, ESCOLANOS. O, et al. Microsoft Voxelized Upper Bodies——A Voxelized Point Cloud Dataset [R/OL].[2021-03-03].
[20]
TIAND, OCHIMIZUH, FENGC, et al. Geometric distortion metrics for point cloud compression [C]// 2017 IEEE International Conference on Image Processing (ICIP). New York: IEEE Press, 2017: 3460-3464. DOI:10.1109/ICIP.2017.8296925 .
[21]
PATEUXS. Tools for Proposal Evaluations [R/OL]. [2021-02-03].