Adaptive Event Address Map Denoising for Event Cameras

Event cameras provide high temporal resolution and dynamic range have application prospects in areas such as visual navigation, robotics, image reconstruction. output an asynchronous stream of address events that can be highly polluted by noise due to lighting motion conditions. In order better analyze process with different spatiotemporal correlation, we introduce a method using event map (EAM) representation classify into background noise, near-edge recessive noise. addition, since closer correlation real events, they are difficult suppress current denoising methods. Accordingly, propose adaptive EAM adjusting the time window form measuring number after pre-denoising performing median filtering opening. The adapt scenes motions generate fill directly used for subsequent process. Alternatively, resolution, restore from EAM. To suitably evaluate performance on real-world data, average local variance amount proposed achieves improvements 24.94% 72.79%, respectively, compared conventional denoising.