Orientation Representation and Efficiency Trade-off of a Biological Inspired Computational Vision Model

摘要

Biological evolution endows human vision perception with an "optimal" or "near optimal" structure while facing a large variety of visual stimuli in different environment. Mathematical principles behind the sophisticated neural computing network facilitate these circuits to accomplish computing tasks sufficiently as well as at a relatively low energy consumption level. In other words, human visual pathway, from retina to visual cortex has met the requirement of "No More Than Needed" (NMTN). Therefore, properties of this "nature product" might cast a light on the machine vision. In this work, we propose a biological inspired computational vision model which represents one of the fundamental visual information - orientation. We also analyze the efficiency trade-off of this model.