This paper proposes an algorithm for obtaining an event-based video from a noisy input video given by physics-based Monte Carlo path tracing of synthetic 3D scenes. Since the dynamic vision sensor (DVS) detects temporal brightness changes as events, the problem of efficiently rendering event-based video boils down to detecting the changes from noisy brightness values. To this end, we extend a denoising method based on a weighted local regression (WLR) to detect the brightness changes rather than applying denoising to each video frame. Specifically, we regress a WLR model only on frames where an event is detected, which significantly reduces the computational cost of the regression. We show that our efficient method is robust to noisy video frames obtained from a few path-traced samples and performs comparably to or even better than an approach that denoises every frame.

翻译：本文提出一种从物理模拟的合成三维场景的蒙特卡洛路径追踪产生的含噪输入视频中获取事件视频的算法。由于动态视觉传感器（DVS）将时域亮度变化检测为事件，高效渲染事件视频的核心问题可归结为从含噪亮度值中检测变化。为此，我们扩展了一种基于加权局部回归（WLR）的去噪方法，直接检测亮度变化而非对每帧视频进行单独去噪。具体地，我们仅在检测到事件发生的帧上拟合WLR模型，这显著降低了回归计算成本。研究表明，该高效方法对来自少量路径追踪样本的含噪视频帧具有鲁棒性，其性能与对每帧进行去噪的方法相当甚至更优。