Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fast graph-based denoising for point cloud color information

Ryosuke Watanabe, Keisuke Nonaka, Eduardo Pavez, Tatsuya Kobayashi, Antonio Ortega

TL;DR

This work tackles the challenge of real-time color denoising for large-scale point clouds, which is hampered by expensive graph constructions and noise level estimation in prior methods. It introduces Fast Graph-Based Denoising (FGBD), a GPU-accelerated pipeline built on scan-line graph (SLG) graph construction, noise estimation using graph-based patches (NE-GBP), and a low-cost filter selection (FSLR) to enable real-time performance while preserving denoising accuracy. The approach replaces heavy SGWT-based steps with fast, locality-driven computations and an integer-based low-pass filter parameter, achieving 30fps on around 1 million points and PSNR comparable to state-of-the-art methods. The practical impact is substantial for live streaming and XR/telepresence applications, with potential extensions to geometry denoising and other graph-based processing tasks.

Abstract

Point clouds are utilized in various 3D applications such as cross-reality (XR) and realistic 3D displays. In some applications, e.g., for live streaming using a 3D point cloud, real-time point cloud denoising methods are required to enhance the visual quality. However, conventional high-precision denoising methods cannot be executed in real time for large-scale point clouds owing to the complexity of graph constructions with K nearest neighbors and noise level estimation. This paper proposes a fast graph-based denoising (FGBD) for a large-scale point cloud. First, high-speed graph construction is achieved by scanning a point cloud in various directions and searching adjacent neighborhoods on the scanning lines. Second, we propose a fast noise level estimation method using eigenvalues of the covariance matrix on a graph. Finally, we also propose a new low-cost filter selection method to enhance denoising accuracy to compensate for the degradation caused by the acceleration algorithms. In our experiments, we succeeded in reducing the processing time dramatically while maintaining accuracy relative to conventional denoising methods. Denoising was performed at 30fps, with frames containing approximately 1 million points.

Fast graph-based denoising for point cloud color information

TL;DR

This work tackles the challenge of real-time color denoising for large-scale point clouds, which is hampered by expensive graph constructions and noise level estimation in prior methods. It introduces Fast Graph-Based Denoising (FGBD), a GPU-accelerated pipeline built on scan-line graph (SLG) graph construction, noise estimation using graph-based patches (NE-GBP), and a low-cost filter selection (FSLR) to enable real-time performance while preserving denoising accuracy. The approach replaces heavy SGWT-based steps with fast, locality-driven computations and an integer-based low-pass filter parameter, achieving 30fps on around 1 million points and PSNR comparable to state-of-the-art methods. The practical impact is substantial for live streaming and XR/telepresence applications, with potential extensions to geometry denoising and other graph-based processing tasks.

Abstract

Point clouds are utilized in various 3D applications such as cross-reality (XR) and realistic 3D displays. In some applications, e.g., for live streaming using a 3D point cloud, real-time point cloud denoising methods are required to enhance the visual quality. However, conventional high-precision denoising methods cannot be executed in real time for large-scale point clouds owing to the complexity of graph constructions with K nearest neighbors and noise level estimation. This paper proposes a fast graph-based denoising (FGBD) for a large-scale point cloud. First, high-speed graph construction is achieved by scanning a point cloud in various directions and searching adjacent neighborhoods on the scanning lines. Second, we propose a fast noise level estimation method using eigenvalues of the covariance matrix on a graph. Finally, we also propose a new low-cost filter selection method to enhance denoising accuracy to compensate for the degradation caused by the acceleration algorithms. In our experiments, we succeeded in reducing the processing time dramatically while maintaining accuracy relative to conventional denoising methods. Denoising was performed at 30fps, with frames containing approximately 1 million points.
Paper Structure (10 sections, 4 equations, 4 figures, 5 tables)

This paper contains 10 sections, 4 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. PROPOSED DENOISING METHOD
  3. Overview of fast graph-based denoising (FGBD)
  4. Graph construction based on SLG
  5. Noise estimation using graph-based patches (NE-GBP)
  6. Filter selection and low-pass filter execution
  7. EXPERIMENTS
  8. Experimental conditions
  9. Experimental results
  10. CONCLUSION

Figures (4)

  • Figure 1: Basic flow of (a) 3DPBS 3DPBS and (b) FGBD (proposed).
  • Figure 2: Neighbor search on scan-line codes (Example of 2-D).
  • Figure 3: Noise estimation using graph-based patches (NE-GBP)
  • Figure 4: The 1st frame of the "david" sequence in the MVUB with $\sigma = 30$: (a) ground truth, (b) noisy point cloud, denoising results by (c) GLR DineshGLRGTV2019, (d) CD-SGW GAC, (e) 3DPBS 3DPBS, and (f) FGBD.