2026-09-30 千葉大学
図1 イベントカメラシミュレーション結果の比較。
＜関連情報＞
- https://www.chiba-u.ac.jp/news/research-collab/post_721.html
- https://ieeexplore.ieee.org/document/11694352
イベント適応型時間洗練によるパストレーシングベースのイベントカメラシミュレーション Path-Tracing-Based Event Camera Simulation via Event-Adaptive Time Refinement
Yuichiro Manabe; Tatsuya Yatagawa; Shigeo Morishima; Hiroyuki Kubo
IEEE Transactions on Visualization and Computer Graphics Published:17 September 2026
DOI:https://doi.org/10.1109/TVCG.2026.3726999
Abstract
Event cameras asynchronously detect brightness changes as events rather than capturing full image frames, making them attractive for a wide range of applications. This has created a growing demand for simulation methods that generate realistic event streams directly from 3D scenes. However, accurate simulation remains challenging because existing approaches often rely on image-frame inputs, and physically based rendering for such inputs is often delegated to external path tracers. Unfortunately, such external rendering engines are black boxes with limited opportunities for event-specific optimization. In addition, conventional path tracing uses fixed-interval temporal sampling over the entire scene, so brightness changes can only be detected at globally synchronized discrete time points. To address these limitations, we propose a path-tracing-based method for high-temporal-resolution event camera simulation from 3D computer graphics scenes. Our method combines path tracing with the bisection method for precise event timing estimation and leverages statistical hypothesis testing to speed up the event search process. We implement the method using hardware-accelerated path tracing, exploiting motion-blur-based time interpolation and stream compaction for efficient rendering. Our results show that substantial acceleration is achieved when statistical search pruning and stream compaction are used together, providing a practical framework for high-fidelity simulation of event streams.
