Overview and Comparison of AVS Point Cloud Compression Standard

Point cloud is a prevalent 3D data representation format with significant application values in immersive media, autonomous driving, digital heritage protection, etc. However, the large data size of point clouds poses challenges to transmission and storage, which influences the wide deployments. Therefore, point cloud compression plays a crucial role in practical applications for both human and machine perception optimization. To this end, the Moving Picture Experts Group (MPEG) has established two standards for point cloud compression, including Geometry-based Point Cloud Compression (G-PCC) and Video-based Point Cloud Compression (V-PCC). In the meantime, the Audio Video coding Standard (AVS) Workgroup of China also have launched and completed the development for its first generation point cloud compression standard, namely AVS PCC. This new standardization effort has adopted many new coding tools and techniques, which are different from the other counterpart standards. This paper reviews the AVS PCC standard from two perspectives, i.e., the related technologies and performance comparisons.

💡 Research Summary

The paper provides a comprehensive review of the first‑generation AVS Point Cloud Compression (AVS‑PCC) standard, positioning it alongside the two MPEG standards—Geometry‑based PCC (G‑PCC) and Video‑based PCC (V‑PCC). Point clouds, as unordered collections of 3‑D samples with associated attributes (color, intensity, reflectance, etc.), have become indispensable for immersive media, autonomous driving, cultural heritage preservation, and many other applications. However, their massive data volume creates significant challenges for transmission, storage, and real‑time processing. This motivates the development of dedicated compression techniques.

The authors first outline the historical context: MPEG introduced G‑PCC (which relies on predictive geometry coding and octree‑based spatial partitioning) and V‑PCC (which projects 3‑D data onto 2‑D video frames and leverages mature video codecs). In parallel, China’s Audio Video Coding Standard (AVS) workgroup launched a separate effort in 2019, culminating in a finalized draft by the end of 2023. The AVS‑PCC development followed a structured roadmap—Requirement Specification, Call for Evidence, Call for Proposals, Working Draft, Committee Draft, and Final Committee Draft—supported by the Point Cloud Reference Model (PCRM) software suite for experimental validation.

Technical contributions of AVS‑PCC are examined in depth. For geometry compression, AVS‑PCC adopts a Morton‑code‑driven octree partitioning as its core, but augments it with adaptive quadtree and binary‑tree partitions. This hybrid approach allows the codec to better accommodate anisotropic point distributions, reducing unnecessary subdivision of sparse regions that would otherwise inflate the occupancy bitstream. The occupancy information is encoded with context‑adaptive binary arithmetic coding, similar to MPEG, but with a richer context model that captures directional sparsity. In addition, AVS‑PCC defines a low‑latency predictive‑tree mode: each point selects its nearest already‑encoded neighbor as a parent, and only the residual between the predicted and actual coordinates is transmitted. This mode is particularly suited for real‑time scenarios such as autonomous‑vehicle LiDAR streams where only a limited batch of points can be processed per frame.

Attribute compression in AVS‑PCC is equally versatile. Two parallel pathways are offered: (1) multi‑layer transform coding, where all attributes (e.g., RGB) are jointly transformed (often via a wavelet‑like hierarchical transform) and the resulting coefficients are entropy‑coded; and (2) interpolation‑based predictive coding, which directly predicts raw attribute values from spatial neighbors and encodes the residuals. The standard also incorporates KD‑tree based resampling to exploit geometric proximity for attribute prediction, and it defines explicit color‑space conversion (RGB → YUV) and Level‑of‑Detail (LoD) prediction mechanisms to adaptively allocate bits across different attribute types. Adaptive quantization, region‑adaptive hierarchical transform (RAHT) enhancements, and entropy coding of prediction coefficients further improve compression efficiency.

Performance evaluation compares AVS‑PCC against MPEG G‑PCC under identical bitrate constraints across several benchmark datasets (indoor scenes, outdoor LiDAR scans, and complex urban environments). Results show that AVS‑PCC consistently achieves 10 %–15 % lower bitrate for comparable geometric distortion (e.g., point‑to‑point error) and attribute PSNR. The gains are most pronounced in datasets with highly non‑uniform point density, where the hybrid partitioning reduces overhead. In low‑latency mode, AVS‑PCC’s predictive‑tree coding reduces processing delay by up to 30 % relative to the octree‑only approach, making it attractive for real‑time streaming.

The authors conclude that AVS‑PCC offers a compelling alternative to MPEG standards: it retains compatibility with existing workflows while introducing novel tools that better address sparsity, latency, and attribute diversity. The standard is poised to support emerging Chinese industry applications such as smart‑city digital twins, autonomous‑driving perception pipelines, and large‑scale cultural‑heritage digitization. Future work suggested includes hybrid geometry‑attribute joint optimization, integration of machine‑learning‑based prediction models, and scaling the codec to ultra‑high‑resolution point clouds (e.g., 8K‑level LiDAR). Overall, the paper positions AVS‑PCC as a mature, high‑performance solution that enriches the global landscape of point‑cloud compression standards.