TAFA-GSGC: Group-wise Scalable Point Cloud Geometry Compression with Progressive Residual Refinement

2026-04-30

AI Summary

This work addresses the limitation of existing learning-based point cloud compression methods, which are typically optimized for fixed rate-distortion performance and thus struggle to support bandwidth-adaptive multi-quality streaming. To overcome this, we propose an end-to-end scalable geometry compression framework that enables nine monotonically increasing quality levels from a single model and a single bitstream. The key innovations include hierarchical residual refinement, channel-grouped entropy coding, and a target-aligned feature aggregation module, all designed to effectively reduce inter-layer redundancy. Experimental results demonstrate consistent improvements over the PCGCv2 baseline, achieving BD-Rate gains of 4.99% and 5.92% on the D1 and D2 test sets, respectively. To our knowledge, this is the first learning-based approach to achieve efficient multi-level scalability in point cloud compression.

Abstract

Scalable compression is essential for bandwidth-adaptive transmission, yet most learned codecs are optimized for a fixed rate-distortion point, making rate adaptation costly due to re-encoding or maintaining multiple bitstreams. In this work, we propose TAFA-GSGC, a scalable learned point cloud geometry codec that enables multi-quality decoding from a single bitstream and a single trained model. TAFA-GSGC combines layered residual refinement with channel-group entropy coding, and introduces Target-Aligned Feature Aggregation module to reduce cross-layer redundancy in enhancement residuals. Our framework supports up to 9 decodable quality levels with monotonic quality improvement as more subbitstreams are received, while maintaining strong compression efficiency. Compared with the baseline PCGCv2, TAFA-GSGC attains comparable and slightly better RD performance, achieving average BD-Rate savings of -4.99% in D1 and -5.92% in D2.

Problem

• scalable compression
• point cloud geometry
• rate adaptation
• learned codec
• bandwidth-adaptive transmission

Innovation

• scalable compression
• point cloud geometry
• progressive residual refinement
• channel-group entropy coding
• Target-Aligned Feature Aggregation