It is believed that vision is adapted to natural inputs. We investigate the relationship between natural time-varying image properties and visual sensitivities to spatial and motion signals. We hypothesize that the visual system is adapted to improve efficiency of visual representation through spatiotemporal decorrelation of input signals. The statistical properties of natural images, in particular, the power spectrum and the image velocity distribution are analyzed systematically and are used to predict the visual sensitivity. We show that the image power spectrum is a product of a spatial scale-invariant term and the velocity distribution and consequently, the contrast sensitivity to achieve the optimal coding is a function of the product of the same two terms. Some of the predictions are compared quantitatively with experimental data from neurophysiology and psychophysics. The agreements are good. We conclude that the spatiotemporal contrast sensitivity is organized to decorrelate the spatial and motion signals thus improve the efficiency of information processing in natural environment.

This paper outlines a dynamic theory of development and adap(cid:173) tation in neural networks with feedback connections. Given in(cid:173) put ensemble, the connections change in strength according to an associative learning rule and approach a stable state where the neuronal outputs are decorrelated . We apply this theory to pri(cid:173) mary visual cortex and examine the implications of the dynamical decorrelation of the activities of orientation selective cells by the intracortical connections. The theory gives a unified and quantita(cid:173) tive explanation of the psychophysical experiments on orientation contrast and orientation adaptation. Using only one parameter, we achieve good agreements between the theoretical predictions and the experimental data.