Independent component analysis of natural image sequences yields spatio-temporal ®lters similar to simple cells in primary visual cortex

Simple cells in the primary visual cortex process incoming visual information with receptive ¢elds localized in space and time, bandpass in spatial and temporal frequency, tuned in orientation, and commonly selective for the direction of movement. It is shown that performing independent component analysis (ICA) on video sequences of natural scenes produces results with qualitatively similar spatio-temporal properties. Whereas the independent components of video resemble moving edges or bars, the independent component ¢lters, i.e. the analogues of receptive ¢elds, resemble moving sinusoids windowed by steady Gaussian envelopes. Contrary to earlier ICA results on static images, which gave only ¢lters at the ¢nest possible spatial scale, the spatio-temporal analysis yields ¢lters at a range of spatial and temporal scales. Filters centred at low spatial frequencies are generally tuned to faster movement than those at high spatial frequencies.