PACE: Post-Causal Entropy Modeling for Learned LiDAR Point Cloud Compression

2026-05-02

AI Summary

This work addresses the high decoding latency and the inherent trade-off between compression performance and delay in octree-based LiDAR point cloud compression caused by causal context modeling. To this end, the authors propose PACE (Post-Causal Entropy modeling), a novel framework that reformulates context aggregation using a non-causal backbone network while preserving autoregressive dependencies only within a lightweight, stage-scalable causal predictor. This design eliminates redundant computations in the backbone and substantially reduces decoding latency. By decoupling the backbone from probability prediction, PACE enables a single model to flexibly adapt to varying latency constraints. Experimental results demonstrate that, in autoregressive mode, PACE reduces decoding latency by over 90% while achieving state-of-the-art compression efficiency, significantly outperforming existing methods in BD-BR metrics.

Abstract

LiDAR point cloud compression is vital for autonomous systems to handle massive data from high-resolution sensors. While learned entropy modeling built upon octree structures yields high compression gains, it faces two critical bottlenecks: 1) prohibitive latency, particularly during decoding, caused by causal, multi-stage context modeling; and 2) a rigid performance-latency trade-off, preventing a single model from adapting to varying constraints. These limitations stem from the tight coupling between context aggregation backbone and probability prediction. To address this, we propose PACE, a new framework that reformulates ancestral context aggregation as a non-causal backbone and confines causality to a lightweight, stage-scalable predictor, eliminating repetitive backbone executions and reducing computational overhead. The predictor supports an arbitrary number of prediction stages, supporting seamless adaptation across diverse performance-latency trade-offs without reloading parameters. Experiments demonstrate that PACE sets a new state-of-the-art in compression efficiency, achieving notable BD-BR savings and reducing decoding latency by over 90% in autoregressive mode, highly attractive for practical applications.

Problem

• LiDAR point cloud compression
• learned entropy modeling
• causal context modeling
• latency-performance trade-off
• octree-based compression

Innovation

• non-causal backbone
• stage-scalable predictor
• LiDAR point cloud compression
• learned entropy modeling
• decoding latency reduction