TOP活動実績1998年

(9) 側抑制ニューラルネットワークを用いた幾何学的視覚モデル
    BPES'98,第13回生体・生理工学シンポジウム論文集pp.405-408,1998-9.

In this paper, we revised the model of T.Yamanoi et al. on visual perception based on two-layer lateral inhibition by introcucing lateral inhibitory neural networks. The visual field is assumed to be a tangential development of the output function. A metric of the visual filed is calculated from local transformation of the output function. A differential geometrical model of visual perception has been made on the basis of two layers of lateral inhibition output from optic nerves in the retina. The output can be approximated closely by a difference of Gaussian function. We derived this output directly from two-dimensional input stimulus on the lateral inhibitory neural networks. Some illusions are explained by the model.
PREVIOUS << >> NEXT