Independent Component Analysis (ICA) was applied to a large set of calibrated natural images. A series of properties of the IC-filters were calculated and their distributions compared with published measurements in simple cells in Macaque cortex. The distributions of spatial frequency bandwidth, orientation bandwidth, length of the receptive fields, and aspect ratio match quite well. The distributions of the peak of the spatial frequency response are different: whereas simple cells have a broad distribution (covering different spatial scales), ICA yields filters with peaks close to the highest spatial frequency allowed by the sampling lattice. First results of applying ICA to natural video sequences indicate that this discrepancy is resolved by taking time into account.