A Structure for Generalized Space-Time Processing: The Feedforward Visual Case

摘要

Traditional interpretation of early visual image processing in a retinal level has focused exclusively on spatial aspects of receptive fields (RFs). We have learned recently that RFs are spatiotemporal entities and this characterization is crucial in understanding and modelling circuits in early visual processing ([2]). We present a generalization of the layered computation concept to describe visual information processing in the joint space-time domain.