Range Image-Based Implicit Neural Compression for LiDAR Point Clouds

2025-04-24

AI Summary

To address the challenges of efficient LiDAR point cloud compression and high-fidelity 3D reconstruction, this paper proposes an implicit neural representation (INR)-based compression framework tailored for range images (RIs). The method innovatively decouples each RI into a depth map and a binary mask map, and employs patch-wise INR modeling for the former and pixel-wise INR modeling for the latter. It further integrates model pruning, weight quantization, and end-to-end differentiable training. This divide-and-conquer architecture overcomes the accuracy–efficiency trade-off inherent in conventional image- or point-cloud-based compression methods applied to RIs. Evaluated on the KITTI dataset, the framework achieves a 21% reduction in 3D reconstruction error at low bitrates, a 14% improvement in object detection mAP, and a 3.8× decrease in decoding latency compared to INR baselines—outperforming all existing compression approaches in comprehensive metrics.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Learning on the Edge & Model Compression
• Data Mining & Knowledge Management: Data Compression

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

This paper presents a novel scheme to efficiently compress Light Detection and Ranging~(LiDAR) point clouds, enabling high-precision 3D scene archives, and such archives pave the way for a detailed understanding of the corresponding 3D scenes. We focus on 2D range images~(RIs) as a lightweight format for representing 3D LiDAR observations. Although conventional image compression techniques can be adapted to improve compression efficiency for RIs, their practical performance is expected to be limited due to differences in bit precision and the distinct pixel value distribution characteristics between natural images and RIs. We propose a novel implicit neural representation~(INR)--based RI compression method that effectively handles floating-point valued pixels. The proposed method divides RIs into depth and mask images and compresses them using patch-wise and pixel-wise INR architectures with model pruning and quantization, respectively. Experiments on the KITTI dataset show that the proposed method outperforms existing image, point cloud, RI, and INR-based compression methods in terms of 3D reconstruction and detection quality at low bitrates and decoding latency.

Problem

• Compress LiDAR point clouds efficiently for 3D scene archives
• Improve compression of 2D range images with neural representation
• Enhance 3D reconstruction quality at low bitrates and latency

Innovation

• Uses implicit neural representation for compression
• Divides range images into depth and mask
• Applies model pruning and quantization techniques