KDDI Research, Inc. Print

KDDI Research members received Top 3% Paper Recognition at the ICASSP 2023

June 16, 2023
KDDI Research, Inc.

A Research paper authored by KDDI Research, Inc. on accurate denoising technology for three-dimensional (3D) point cloud data (Note 1) was certified as one of the Top 3% Paper Recognition (Note 2) at the ICASSP 2023 (Note 3), which is the premier international conference on signal processing. This is the achievement of a joint research project with the laboratory organized by Prof. Antonio Ortega of the University of Southern California.




The aim of KDDI Research, Inc. is to develop telecommunications that give users the sense that “they are actually there” by transmitting and displaying 3D data (such as photorealistic human figures) captured from physical space to each other. A 3D reproduction technology that acquires virtual 3D data from physical space is one of the key elemental technologies needed to achieve such ultra-realistic XR experiences. However, 3D data acquired from 3D reproduction technology are often contaminated with noise caused by the measurement errors of scanning devices. Thus, we proposed an accurate denoising method for a point cloud, which is widely used as a format for representing 3D data.
In the proposed method, a graph (Note 4) is constructed first from a noisy point cloud. Then, the noise is eliminated by attenuating the high-frequency signal components on a graph (parts where the coordinate signals change intensively on a graph). Although this graph-based approach is conventional, we propose a novel graph construction method that can provide accurate denoising results by taking the surface of a point cloud into consideration to reduce redundant connections on the graph. Specifically, a graph faithfully representing the geometric structure of its point cloud can be constructed by eliminating the connections which are not consistent with the surfaces. 
The paper was highly evaluated for the novelty of its graph construction method that is significantly different from conventional approaches and its effectiveness for a range of data was demonstrated by experiments. By applying the proposed method to a noisy point cloud, it is expected to improve the visual quality and enhance the user experience in XR applications.


【Paper information】
Ryosuke Watanabe, Keisuke Nonaka, Eduardo Pavez, Tatsuya Kobayashi and Antonio Ortega, "Graph Wavelet-Based Point Cloud Geometric Denoising with Surface-Consistent Non-Negative Kernel Regression," 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023, pp. 1-5,
Graph Wavelet-Based Point Cloud Geometric Denoising with Surface-Consistent Non-Negative Kernel Regression


This work was supported by "Strategic Information and Communications R&D Promotion Programme (SCOPE)" of Ministry of Internal Affairs and Communications, Grant no. JPJ000595.


【KDDI Research's Initiatives】
KDDI and KDDI Research formulated "KDDI Accelerate 5.0" (Japanese only), based on the concept for a next-generation society envisioned for 2030, and summarized how it can be accomplished in the "Beyond 5G/6G Whitepaper" in terms of the "Future Image" and "Technologies" required to create an ecosystem that spurs innovation.
To develop a new lifestyle, the two companies will continue to invest in R&D on seven different technologies and on an orchestration technology that links them together. The results presented in this study correspond to one of the seven technologies, "XR".




(Note 1)A data format that represents a 3D object as a set of points composed of position and color values.
(Note 2)Papers rated in the top 3% of all the accepted papers were certified.
(Note 3)IEEE ICASSP(2023 IEEE International Conference on Acoustics, Speech, and Signal Processing)is a flagship international conference on signal processing.
(Note 4)A data structure consisting of a large number of points (3D point cloud) with position and color values and connections (edges) between highly similar points.


※The information contained in the articles is current at the time of publication.Products, service fees, service content and specifications, contact information, and other details are subject to change without notice.