Receptive fields and profiles, and wavelet analysis
pp. 45-111
in: , Elements of neurogeometry, Berlin, Springer, 2017Abstract
This chapter discusses receptive fields and receptive profiles of visual neurons, starting with the photoreceptors and the ganglion cells of the retina, and proceeding via the neurons of the lateral geniculate nucleus to those of V1. It explains how they act on the optical signal as filters, and broaches the problem of their linearity or nonlinearity. In the linear case, the effect they have on the signal falls to a large extent under the rule of what is known as "wavelet analysis' in signal processing. The chapter gives an outline of this fundamental notion, then discusses how receptive profiles can be interpreted within the framework of information theory as a means of optimizing the processing of natural images, which have very particular statistical properties. To a certain extent, the geometric formatting of the signal by means of a certain kind of wavelet (essentially, partial derivatives of Gaussians) is a way of optimizing the compression of natural images.