ISO/IEC MPEG-I — Point Cloud Compression

Video-based (V-PCC) & Geometry-based (G-PCC)

The development of MPEG point cloud compression standards was initially driven by the distinction between dense and sparse point clouds, each requiring specialized compression techniques. Two primary standards have emerged: Video-based Point Cloud Compression (V-PCC) for dense point clouds and Geometry-based Point Cloud Compression (G-PCC) for sparse point clouds. Advancements in G-PCC have led to three additional branches: Enhanced G-PCC (E-G-PCC), Geometry for Solid (GeS), and Low Complexity, Low Latency LiDAR Coding (L3C2), each targeting specific application needs. These standards provide tailored solutions for various use cases, including real-time streaming and high-quality visualization. Overall, MPEG's strategies ensure that point cloud compression remains effective and relevant across diverse industries such as media, gaming, autonomous driving, and robotics.

See standards family Which one should I use? V-PCC (IS), G-PCC (IS), L3C2 (FDIS), E-G-PCC (under development)

Dense vs sparse point clouds

The basic split

Dense point clouds typically sample surfaces uniformly and in high detail, which makes them well suited for 2D atlas projection and video-based compression (V-PCC). This approach efficiently exploits temporal coherence and hardware video decoders, and is particularly effective for humans, objects, and other dense dynamic scenes. In contrast, sparse or irregular point clouds, such as LiDAR scans or mapping data, do not project well into atlases and would waste bandwidth on empty pixels. For these, geometry-based compression (G-PCC) is more appropriate, as it codes points directly in 3D using hierarchical data structures. This basic split between dense/projectable and sparse/irregular content motivated the original two-track MPEG standardization.

DenseSparseLiDAR

The advanced split of G-PCC

The original dense vs. sparse dichotomy is only a first-order view of the PCC landscape. Within the G-PCC framework, MPEG has progressively introduced specialized branches to address more complex scenarios. Enhanced G-PCC (E-G-PCC) extends the baseline with advanced temporal prediction, significantly improving coding of dynamic and time-evolving point clouds. GeS (Geometry for Solid) targets dense or solid objects where the surface behaves more like a continuous manifold than a sparse scatter, bridging some of the performance gap with V-PCC while staying purely geometry-based. Finally, L3C2 is a low-complexity, low-latency branch designed around the acquisition patterns of spinning LiDAR, enabling efficient real-time processing for autonomous driving and robotics. Together, these branches form an advanced split of G-PCC that maps directly onto content categories such as dynamic, dense/solid, or LiDAR-driven point clouds.

Categories of MPEG PCC test contents — Illustrative categories used in test content selection.

Standards family

Core standards

V-PCC — ISO/IEC 23090-5 (with V3C) — Published IS 2023
G-PCC — ISO/IEC 23090-9 — Published IS 2023

V-PCC leverages video codecs via atlases; G-PCC operates natively in 3D.

New branches (under the G-PCC framework)

E-G-PCC — enhanced temporal prediction for dynamics (23090-38, under development)
GeS — Geometry for Solid, dense content focused (pre-WD track)
L3C2 — Low Complexity, Low Latency LiDAR Coding (23090-30, FDIS 2025)

These address dynamic content, dense solids, and spinning LiDAR respectively.

V-PCC — Video-based Point Cloud Compression

How it works

Segments the point cloud into surface patches, parameterizes and packs them into atlases, and compresses geometry/attributes with a video codec. HW decoders can be reused for high throughput.

AtlasesVideo codecsHW-friendly

Where it shines

Dense, temporally coherent content (humans, objects, telepresence) and delivery over existing video pipelines.

G-PCC & branches — Geometry-based compression

G-PCC

Compresses directly in 3D using octree occupancy and prediction/transform tools for geometry and attributes. Versatile across static/dynamic, dense/sparse, including LiDAR.

E-G-PCC

Adds temporal prediction across frames and optimizations that improve dynamic content compression and static efficiency; published as a new part for clarity and compatibility.

GeS

Geometry for Solid targets dense/solid point clouds to approach V-PCC-like performance without relying on the video coding ecosystem.

L3C2

Tailored for spinning LiDAR with known acquisition order; very low latency/complexity using OneChain attribute coding and joint geometry/attribute processing.

Applications & requirements

Applications and requirements across PCC standards — Where each standard fits best (dynamic, dense, LiDAR, latency, etc.).

Suitability snapshots

V-PCC vs G-PCC (by content category)

Standard	Suitability
V-PCC	Solid / Dense
G-PCC	Sparse / Scant / LiDAR

Branches of G-PCC

Standard	Suitability
E-G-PCC	Sparse / Scant / LiDAR
GeS	Solid / Dense
L3C2	Spinning LiDAR

Suitability summarizes expected performance by content class.

Use cases & recommendations

Use case	Recommended standard
Mobile visualization of a dense static object (gaming asset)	GeS
Sparse dynamic point cloud for autonomous driving	G-PCC / E-G-PCC
Dense dynamic point cloud for real-time VR simulation	V-PCC
Streaming a dense static point cloud over a video-based network	V-PCC
Spinning LiDAR for low-latency robotics detection	L3C2

Use case	Recommended standard
Archiving sparse static cultural-heritage scans	G-PCC / E-G-PCC
Real-time traffic monitoring with dynamic spinning LiDAR	L3C2
High-quality dense point clouds for cinematic rendering	V-PCC
Remote structural analysis with sparse dynamics	E-G-PCC
Urban planning with dense static city blocks	GeS

Illustrative mapping; actual choice depends on bitrate, latency, and infrastructure constraints.

Official resources

Standards

Also see MPEG pages for V3C/V-PCC, G-PCC, and L3C2 for news and updates.

Public introductions & material

FAQ

V-PCC vs G-PCC — quick difference?

V-PCC projects to 2D atlases and uses video codecs (easy HW reuse, strong for dense surfaces); G-PCC stays in 3D (versatile across densities and LiDAR).

Do I need both?

Many systems implement both and choose per content/workflow. Branches (E-G-PCC, GeS, L3C2) further tailor to dynamics, dense solids, and spinning LiDAR.