基于事件相机的可视化及降噪算法

针对事件相机（Event Camera）输出的异步事件流信息不利于人眼观察、难以衔接应用任务且存在大量噪声的问题，介绍一种可视化及降噪算法。结合事件流能够反映场景中物体运动边缘信息的特点，利用物体运动边缘的时间和空间连续性进行降噪处理，进而利用事件数量和时间阈值双限制的方式累积事件得到事件“帧”，达到可视化、便于应用的目的。在真实数据集实验中，降噪算法可以有效处理背景噪声，在运动起始或缓慢时保存更多细节边缘事件信息，提升有效角点检测数量，可视化算法在保证帧率的同时，降低事件数量方差，提高事件“帧”信息的均匀性。实验结果证明了可视化及降噪算法的有效性。 

Visualization and noise reduction algorithm based on event camera

To overcome the problem that the asynchronous event stream generated by the event camera is hard to observe, utilize and there is a lot of noise, we introduce an improved visualization and noise reduction algorithm for the event camera. Because the event stream reacts to the object movement, the proposed algorithm gets valid events by filtering the noise with the time and space continuity of moving edge. To easily observe and apply, events are accumulated with a double limitation of the events number and the time threshold. In the real dataset experiment, the noise reduction algorithm can effectively deal with the background activity noise and save the detail edge information when the movement begins or moves slowly, increasing the number of corner detections. The visualization algorithm reduces the variance of events number while ensuring the frame rate, and improves the information uniformity of the "event frame". The experimental results show the effectiveness of the proposed method in terms of noise reduction and visualization.