That's a picture of "VFI" atomic radii, which is based on data related to the length of the covalent bonds that the elements form when they're part of a molecule. The VFI method of determining atomic radii works fairly well for elements that actually take part in covalent bonding, but is not a useful way to determine the atomic radius for noble gases, since they only form very very weak covalent bonds. The stronger a covalent bond, the shorter it will be. Strong, short covalent bonds normally go hand-in-hand with short atomic radii, which is why the VFI method of measuring radius works well for elements other than noble gasses. Any attempt to determine the atomic radius of a noble gas with the VFI method will be skewed by the fact that noble gasses only form very very weak (and therefore very long) covalent bonds.
Note that this was partly explained in the text at the bottom of your picture: "These data are based on interatomic distances in the structures of the elements."
Pete's chart, and probably whatever hypervalent_iodine wanted to link to, suffer from the same problem. Although Pete's chart doesn't actually say how the radius values were calculated, I would bet money that it is also VFI radii, or some similar method that depends on bond length.
Here is another link from the same source as your graphic:
http://www.crystalmaker.com/support/tutorials/crystalmaker/atomicradii/index.html
Go down to the bottom and look at the actual numbers for atomic radii. You'll see that the radii decrees all the as way across the row from left to right, with no discrepancy for the noble gases.
Edit: If anyone here believes that the noble gasses really have a smaller atomic radius (by which I mean the average distance from the nucleus to the valence electrons), I defy you to provide a coherent explanation of why that would be, especially considering Slater's rules...
