Learning simple and complex cells-like receptive fields from natural images: a plausibility proof