3D Point Cloud Compression with Recurrent Neural Network and Image Compression Methods

TL;DR

This work tackles the large data footprint of LiDAR point clouds by transforming scans into calibrated 2D representations (range, azimuth, intensity) and applying a hybrid compression approach that combines traditional image codecs with a self-supervised recurrent neural network for range. The methodology includes a lossless 3D-to-2D projection, range denoising, per-channel azimuth alignment, and distinct handling of intensity data, augmented by a new intensity-oriented evaluation metric. Results show competitive or superior geometric distortion compared with baselines, while achieving substantially lower data footprints, especially when intensity is compressed with JPEG 2000 and range with the RNN. The approach offers practical impact for AV data storage, transmission, and cloud-based processing, and the authors release their code and datasets to enable reproducibility.

Abstract

Storing and transmitting LiDAR point cloud data is essential for many AV applications, such as training data collection, remote control, cloud services or SLAM. However, due to the sparsity and unordered structure of the data, it is difficult to compress point cloud data to a low volume. Transforming the raw point cloud data into a dense 2D matrix structure is a promising way for applying compression algorithms. We propose a new lossless and calibrated 3D-to-2D transformation which allows compression algorithms to efficiently exploit spatial correlations within the 2D representation. To compress the structured representation, we use common image compression methods and also a self-supervised deep compression approach using a recurrent neural network. We also rearrange the LiDAR's intensity measurements to a dense 2D representation and propose a new metric to evaluate the compression performance of the intensity. Compared to approaches that are based on generic octree point cloud compression or based on raw point cloud data compression, our approach achieves the best quantitative and visual performance. Source code and dataset are available at https://github.com/ika-rwth-aachen/Point-Cloud-Compression.

Figures (11)

• Figure 1: Flowchart of the proposed compression method.
• Figure 2: The 2D structure of a LiDAR scan. The sequence numbers from 0 to 57983 indicate in which order the points are stored within a single message.
• Figure 3: Shifting the rows of the naive range image according to the azimuth offsets.
• Figure 4: The effect of denoising for the range image.
• Figure 5: Example of an azimuth image.