Left: Fourier space
comparison. Modulation transfer function measured in a
simple cortical cell (De Valois et al, in Frontiers of Visual
Springer, 1978) compared to the appropriate Gabor elementary
Right: Direct space comparison. Measured spatial sensitivity of the same cell, compared to the same elementary signal.
From: S. Marcelja,
description of the responses of simple cortical
By the end of 1970's there were
many published accurate measurements of single cell responses from
the primary visual cortex. They suggested to me that the visual
information is represented in the particular basis set, known in
physics since the time of Heisenberg and Von Neumann. Basis vectors
are formed as a product of a Gaussian and a sine or cosine functions
and have the property that they minimise Heisenberg uncertainty.
Gabor called them 'elementary signals' and emphasised their
suitability for communication.
Today, the representation seen in the visual cortex in 1980 would be considered an example of multiresolution image decomposition. This particular basis has found many applications in signal and image processing.