Huang, BowenLazzarotto, DaviEbrahimi, Touradj2024-03-222024-03-222024-03-222024-03-1810.1109/ICASSP48485.2024.10446813https://infoscience.epfl.ch/handle/20.500.14299/206652Point clouds allow for the representation of 3D multimedia content as a set of disconnected points in space. Their inher- ent irregular geometric nature poses a challenge to efficient compression, a critical operation for both storage and trans- mission. This paper proposes a VAE-inspired codec tailored for dynamic point cloud geometry compression, taking ad- vantage of a temporal autoregressive hyperprior to enhance compression performance. Specifically, features derived from adjacent point cloud frames help build a hyperprior for con- ditional entropy coding. Sparse convolutions are leveraged to reach higher computational efficiency when compared to 3D dense convolutions. Remarkably, the proposed approach achieves an average 60.2% BD-rate gain against the contem- porary V-PCC compression standard from MPEG.Point cloud compressionvariational au-toencoderinter-frame codingtext::conference output::conference paper not in proceedings