Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PRISM: Color-Stratified Point Cloud Sampling

Hansol Lim, Minhyeok Im, Jongseong Brad Choi

TL;DR

PRISM tackles RGB-LiDAR downsampling by exploiting color diversity as a sampling driver. It treats color as the stratification domain and enforces a per-bin capacity, selecting a global $k^*$ to achieve a target compression via $S(k) = \sum_b \min(n_b, k)$ with per-bin counts $n_b$. The method preserves rare colors and texture-rich regions, yielding near $r_{target} \approx 1\%$ compression with higher color-entropy gains across diverse scenes, as demonstrated on Toronto-3D, ETH3D, and Paris-CARLA. While geometric fidelity can lag behind purely spatial samplers, PRISM offers predictable compression and photometric preservation, making it attractive for radiance-field learning and downstream photorealistic reconstruction.

Abstract

We present PRISM, a novel color-guided stratified sampling method for RGB-LiDAR point clouds. Our approach is motivated by the observation that unique scene features often exhibit chromatic diversity while repetitive, redundant features are homogeneous in color. Conventional downsampling methods (Random Sampling, Voxel Grid, Normal Space Sampling) enforce spatial uniformity while ignoring this photometric content. In contrast, PRISM allocates sampling density proportional to chormatic diversity. By treating RGB color space as the stratification domain and imposing a maximum capacity k per color bin, the method preserves texture-rich regions with high color variation while substantially reducing visually homogeneous surfaces. This shifts the sampling space from spatial coverage to visual complexity to produce sparser point clouds that retain essential features for 3D reconstruction tasks.

PRISM: Color-Stratified Point Cloud Sampling

TL;DR

PRISM tackles RGB-LiDAR downsampling by exploiting color diversity as a sampling driver. It treats color as the stratification domain and enforces a per-bin capacity, selecting a global to achieve a target compression via with per-bin counts . The method preserves rare colors and texture-rich regions, yielding near compression with higher color-entropy gains across diverse scenes, as demonstrated on Toronto-3D, ETH3D, and Paris-CARLA. While geometric fidelity can lag behind purely spatial samplers, PRISM offers predictable compression and photometric preservation, making it attractive for radiance-field learning and downstream photorealistic reconstruction.

Abstract

We present PRISM, a novel color-guided stratified sampling method for RGB-LiDAR point clouds. Our approach is motivated by the observation that unique scene features often exhibit chromatic diversity while repetitive, redundant features are homogeneous in color. Conventional downsampling methods (Random Sampling, Voxel Grid, Normal Space Sampling) enforce spatial uniformity while ignoring this photometric content. In contrast, PRISM allocates sampling density proportional to chormatic diversity. By treating RGB color space as the stratification domain and imposing a maximum capacity k per color bin, the method preserves texture-rich regions with high color variation while substantially reducing visually homogeneous surfaces. This shifts the sampling space from spatial coverage to visual complexity to produce sparser point clouds that retain essential features for 3D reconstruction tasks.
Paper Structure (16 sections, 11 equations, 4 figures, 2 tables)

This paper contains 16 sections, 11 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Camera-LiDAR Integration
  5. Evaluation Metrics
  6. Methodology
  7. Overview
  8. Color-based Stratification
  9. Stratified Sampling with Bin Limit
  10. Global k Selection
  11. Results and Analysis
  12. Quantitative Results
  13. Qualitative Analysis
  14. Ablation Study
  15. Conclusion
  16. ...and 1 more sections

Figures (4)

  • Figure 1: ETH3D courtyard scene. The scene contains high color variations in small objects present in the scene alongside building walls with same color.
  • Figure 2: Color distribution comparison of PRISM and baseline methods on the courtyard scene. PRISM preserves chromatic diversity closely aligned with the input distribution. All of the baselines were sampled close to 1%
  • Figure 3: Point cloud comparison of PRISM and baseline methods. PRISM retains higher density in texture-rich regions with chromatic variance.
  • Figure 4: Effect of bin capacity $k$ on point cloud density. As $k$ increases, compression ratio grows while maintaining color-guided sampling characteristics.