Found it here. Briefly it is something like this..

MIT's Computer Science, AI laboratory and other colleagues have been trying to find out what is the smallest amount of needed information to represent an image? It is said that the results/discovery could lead to great advances in the automated identification of online images and, ultimately, provide a basis for computers to see like humans do. They found out that smallest amount to be 32x32 resolution of any image. This work is going to be presented in June in Alaska at a conference on Computer Vision and Pattern Recognition(CVPR '08). They say it's doing well for most regular objects like cars, trees, humans etc and added they have much more to do.

Interested people could read more about this here.