Abstract

Linear subband image transforms, composed of banks of linear filters tuned to various spatiotemporal frequencies, have been proposed as models of simple cells in striate cortex. Energy measures, composed of the squared sum of the outputs of quadrature pairs of linear filters, have been proposed as models of complex cells. However, given that neurons have a limited dynamic range, the squaring nonlinearity is unreasonable. Thus, we (and others) have proposed that automatic gain control (AGC), mediated by nonlinear (multiplicative) feedback of the contrast signal, plays an important role in striate physiology.

© 1989 Optical Society of America