Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Octree-based Learned Point Cloud Geometry Compression: A Lossy Perspective

Kaiyu Zheng, Wei Gao, Huiming Zheng

Abstract

Octree-based context learning has recently become a leading method in point cloud compression. However, its potential on lossy compression remains undiscovered. The traditional lossy compression paradigm using lossless octree representation with quantization step adjustment may result in severe distortions due to massive missing points in quantization. Therefore, we analyze data characteristics of different point clouds and propose lossy approaches specifically. For object point clouds that suffer from quantization step adjustment, we propose a new leaf nodes lossy compression method, which achieves lossy compression by performing bit-wise coding and binary prediction on leaf nodes. For LiDAR point clouds, we explore variable rate approaches and propose a simple but effective rate control method. Experimental results demonstrate that the proposed leaf nodes lossy compression method significantly outperforms the previous octree-based method on object point clouds, and the proposed rate control method achieves about 1% bit error without finetuning on LiDAR point clouds.

Octree-based Learned Point Cloud Geometry Compression: A Lossy Perspective

Abstract

Octree-based context learning has recently become a leading method in point cloud compression. However, its potential on lossy compression remains undiscovered. The traditional lossy compression paradigm using lossless octree representation with quantization step adjustment may result in severe distortions due to massive missing points in quantization. Therefore, we analyze data characteristics of different point clouds and propose lossy approaches specifically. For object point clouds that suffer from quantization step adjustment, we propose a new leaf nodes lossy compression method, which achieves lossy compression by performing bit-wise coding and binary prediction on leaf nodes. For LiDAR point clouds, we explore variable rate approaches and propose a simple but effective rate control method. Experimental results demonstrate that the proposed leaf nodes lossy compression method significantly outperforms the previous octree-based method on object point clouds, and the proposed rate control method achieves about 1% bit error without finetuning on LiDAR point clouds.

Paper Structure

This paper contains 15 sections, 5 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Method
  4. Experiments
  5. Conclusion
  6. References

Figures (4)

  • Figure 1: Data Analysis. (a) The ratio of remaining points when setting different $qs$. (b) Histogram of octree occupancy distribution in depth level 16.2 and 17. (c) Laplacian fitting results of the context model output in depth level 16.2 and 17.
  • Figure 2: Proposed leaf nodes lossy compression framework. (a) The process of compression, the number in the octree represents occupancy, and the number in the coding module represents the decoding order. (b) Details of leaf nodes compression, leaf lossless steps $s$ is set to 2 in the figure. (c) Structure of backbone context model. For non-leaf coder, $K$ is set to 255. For leaf bit-wise coder, $K$ is set to 2. For leaf bit predictor, $K$ is set to 8.
  • Figure 3: Performance comparison using rate-distortion curve on MPEG 8i and MVUB dataset.
  • Figure 4: Performance comparison using rate-distortion curve on Ford dataset with variable rate. The square points are results when using the truncating way.