Research on Visual Invariance Based on Dynamic Receptive Field

Abstract

As an important problem in human visual research, visual invariance has not been explained well in essence yet. Hierarchical receptive field model which could explain a part of visual invariance was considered as a fixed structure comprised of different layers of cells in visual cortex without dynamic characteristics. Aiming at this problem, a dynamic receptive field model with changeable parameters based on Gabor function was proposed to explain visual invariance. Then a rotation invariance model was presented for instance to take the experiment of image feature extraction. The result showed that the dynamic receptive field in this paper had the invariance for different types of images and could excellently explain visual invariance from the math, which also accorded with the recent neurophysiological studies on visual system.