Towards retina-like Bayer filtering with spikes

Martin Pham

Abstract: The goal of this project is to demosaick RAW images using a spike-encoded Bayer filter. We use a spiking neural network approach to the demosaicking by constructing a retina-like network using the Brian2 neural simulator. The retina network models different neuronal equations describing the action potential dynamics. We compare CPU, GPU and discuss challenges to neuromorphic emulation and realization.