Reminds me a little of the SIGGRAPH paper on hybrid images, where the distance at which you view an image determines, for example, whether you see a smiling or frowning person:

http://cvcl.mit.edu/hybridimage.htm