There is no "conversion" between a raster image and a vector graphic. Strictly speaking, a true "conversion" would be a grid array of vector rectangles, which would have no practical advantage whatsoever over a raster image. Generating a vector version of a raster image is not a "conversion" in the sense of converting a TIF to a PNG. There is no "translation" of a raster image into a vector graphic in the sense of translating a Corel Draw file into an Illustrator file.
There is only redrawing what is represented by a raster image as paths, in the same sense that you can redraw an oil painting with a pencil. Autotrace routines (like Illustrator's so-called LiveTrace feature) try to automate that re-drawing by detecting regions of similarly-colored pixels. There is no shape-detection intelligence involved. The autotrace routine doesn't know that the eye's pupil is round or that a football is symmetrical (thus, the "distortion" you are taking about).
So...it's probably easier to just make a redo using the pen tool in Ps on a layer above the original, "tracing" it, then scaling the finished vector however you'd like...
I do know that Ben Willmore gave a lesson on this very subject on creativeLive...but I was so spaced by that time that I really didn't catch half of what he was talking about...maybe if you contact him through his website, he can help. He's a very cool super guru. http://photos.digitalmastery.com/