Route 1 would give you a much more dynamic approach relative to #2. You could divide the image up in an editor (tedious), or programmatically (harder but more fun)