Summary of S4-260138: New SID on Dynamic Mesh for Mobile

Document Overview

This is a New Study Item Description (SID) for Release 20 proposing a study on Dynamic Mesh for mobile applications. The work item is classified as a Study and is a child of the parent work item FS_Beyond2D (Evaluation and Characterization of Beyond 2D Video Formats and Codecs).

Background and Justification

Gaps from Rel-19 Work

• TR 26.956 studied various beyond 2D video formats including dynamic meshes, but time constraints left dynamic mesh aspects incomplete
• While TR 26.956 documented that professionally produced dynamic mesh content is of sufficient quality, further investigation is needed for emerging use cases

Key Drivers for This Study

Prosumer Content Production
- New AI-based production tools may enable volumetric video production for single persons with reduced camera counts
- Simple camera setups with cloud-based AI post-production could enable social media producers to adopt this format
- Market solutions and attainable visual quality need study

Real-time Production
- Volumetric video enhances live events (concerts, sports training)
- Real-time production and compression feasibility needs investigation
- Attainable visual quality for live scenarios requires study

MPEG V-DMC Standardization
- MPEG finalized V-DMC (ISO/IEC 23090-29) in July 2025, to be published beginning 2026
- Reference software available since October 2025
- MPEG mesh metric tools and renderer available for objective and subjective testing

Mobile Implementation
- First V-DMC implementations on mobile devices demonstrated but not documented
- Implementation aspects (pixel rate limitations, number of video decoders) need study

Format Comparison
- Dense dynamic point clouds can also represent volumetric video
- Comparative study of dynamic mesh vs. dense dynamic point cloud needed (pros/cons, quality, implementation)

AI-Generated Content
- Dynamic mesh increasingly generated via AI (from text prompts or images)
- Quality assessment beyond geometric quality needed (motion naturalness, frame transitions, visual fatigue)
- No existing objective metrics or subjective methodologies for no-reference evaluation
- Framework needed for assessing AI-generated dynamic meshes without reference models

Comprehensive Format Coverage
- With dense dynamic point clouds studied in TR 26.956 and Gaussian Splat study ongoing, studying dynamic mesh would complete coverage of three relevant volumetric video formats

Study Objectives

1. Content Generation Scenarios

Offline Prosumer Production
- Study dynamic mesh content generation for offline production in prosumer use cases (e.g., social media)
- Focus on limited camera setups (e.g., less than 5 cameras)
- Include offline production and compression aspects

Real-time Services
- Study dynamic mesh for real-time services (e.g., live concerts, live sports training in social media)
- Include real-time compression aspects
- NOTE 1: Avatar-mesh-based real-time communication is explicitly out of scope

2. Compression Technology Evaluation

• Evaluate dynamic mesh formats with existing and emerging compression technologies (e.g., MPEG V-DMC with HEVC as underlying 2D codec)
• Provide objective results using MPEG-defined objective metrics
• Generate videos for subjective testing using MPEG subjective test methodology
• Use V-PCC as baseline for comparison (similar approach to TR 26.956)

3. Implementation Aspects

• Study number of required video decoders
• Study pixel rate limitations

4. Format Comparison

• Compare dynamic mesh with other formats
• Derive recommendations on which formats suit different conditions

5. AI-Generated Content Quality Framework

• Develop and study framework for assessing quality of AI-generated dynamic meshes without reference or original model
• Decouple geometric and appearance information from textured 3D meshes

6. Reuse of Existing Results

NOTE 2: The study should:
- Reuse existing performance results from MPEG or other SDOs that fit the 3GPP evaluation framework
- Consider communication with MPEG for potential further evaluation on selected topics
- May initiate evaluation independently of MPEG if needed

7. Identify Next Steps

• Identify potential areas for normative work as next phase
• Communicate with other 3GPP WGs regarding relevant aspects as needed

Expected Output and Timeline

New Specifications

• Type: TR (Technical Report)
• Target completion: SA#114
• For information: SA#113

Impacted Existing Specifications

• TR 26.956: Complement the already published Beyond2D report with:
• More scenarios/use cases
• Additional test results
• Additional metrics

Work Item Details

• Rapporteur: Jiayi Xu (China Mobile)
• Leadership: SA4
• Other WG Involvement: None identified
• Supporting Companies: China Mobile, InterDigital, Pengcheng Laboratory, LG Electronics, CATT, Deutsche Telekom, ZTE, Nokia, Samsung, KDDI, Sony Group Corporation

Technical Scope Summary

This SID proposes a comprehensive study of dynamic mesh for mobile applications, covering production (both prosumer and real-time), compression evaluation using MPEG V-DMC, implementation aspects, format comparison with alternatives like dense dynamic point clouds, and novel quality assessment frameworks for AI-generated content without reference models. The study aims to complete the volumetric video format landscape and identify gaps for potential normative work.