Every time I hear someone talking about imaging (spy) satellites, somebody invariably brings up those top secret high speed government spy satellites that can read the license plate on your car or the headlines of the newspaper you're reading. As far as I can tell, such claims have little basis in reality. While almost all data on satellites such as the advanced KH-12 flown by the National Reconnaissance Office (NRO) of the United States is classified, it's not all that hard to figure out what they should be capable of.
All an imaging satellite is, really, is a large, space based optical telescope/camera, not unlike the Hubble Space Telescope. The resolving power of any telescope can be calculated by the following formula:
θ = 115.8/D
where θ is angular resolution of the objective, expressed in seconds of arc, and D is the diameter of the objective in millimeters. This is commonly referred to as the Dawes limit of the objective, and is quite a good estimate of the theoretical resolution the objective is capable of. The Hubble telescope has an objective of 2.4 meters and this is probably pretty close to the maximum diameter that can be launched into space with currently available launch vehicles, so the NRO satellites can't be much bigger than this. The Dawes limit of a 2.4 meter objective works out to around .05 seconds of arc.
To figure out actual size of an object from its angular size at a given distance use the formula:
angular size(in degrees) = 57.3*actual size / distance
For a distance of about 650 kilometers, which is likely a typical operating altitude for imaging satellites, this works out to right around 10 centimeters. Unless the birds are equipped with larger, multi-segment mirrors or some other unknown capability, that is about the theoretical limit of their resolution. This does not necessarily mean that 10 centimeters is the smallest oject these cameras can distinguish, but that this is the minimum distance that must separate two objects for the satellite to tell them apart. So while they might be able to spot a golf ball on a putting green under exceptional circumstances, there is no way they could read a license plate.
Then there is the atmosphere. Long the bane of astronomers, the atmosphere also limits the performance of any telescope looking through it. Atmospheric turbulence (what astronomers refer to as 'seeing') commonly reduces the actual performance of astronomical objectives to around .5 seconds of arc, although i'm not sure this is as critical for taking pictures of terrestrial objects as it is for astronomical ones.
Now, if none of this is enough to convince you, consider this one undeniable fact: license plates are not usually mounted on the tops of cars.