Primitive vision based on sensing change

Abstract

An instructive way to discover a new avenue of research is to think about old work with new eyes. If, with a mind full of modern trends and recent theories, one reads through old literature with a discerning eye, it is surprising what crops up. With a bit of luck, you suddenly realise something that the old ones missed. There has been an enormous outburst of research on spatial vision in the past 25 years or so, starting with the controversies about the existence or otherwise of elements in the mammalian visual cortex tuned to particular spatial or temporal frequencies, and arguments about whether visual processing neurons should be considered as matching templates (detecting edges and bars) or as Fourier analysers. This period has also seen the rise of white noise analysis, in which the stimulus is randomly distributed in space and time so that all possible combinations of stimulus pattern have some chance of appearing. Pattern perception has also been intensively studied as a problem in human psychophysics, as a problem in visual behaviour of lower animals, and as a way of giving object recognition to computers. Show more