An Improved Approach for Visualizing Dynamic Vision Sensor and its Video Denoising

Abstract

Dynamic vision sensor (DVS) is an event-based camera capturing the changes of vision with high speed and low storage consumption. To better understand what DVS captures, we need to visualize the events. Existing methods have realized visualization. To optimize the vision experience, this paper proposes a framework to visualize events with rich information, high speed and less noise. Firstly, we propose an improved visualization approach using overlapped events based on human vision system. Secondly, we propose a video denoising method using shared dictionaries. In our experiments, the proposed method realizes the expected purpose on the whole video.