well, image recognition is certainly a very complex undertaking. (nothing that i would consider myself . . . unless i was being paid well for it, of course.)
but i would say, as a place to begin, you would want to scan through the image pixel by pixel and store the locations of any contrast points . . . like when a black pixel is right next to a white pixel, store the location and then you can come back later and try to correlate contrast points with geometrical shapes.
if youre looking specifically for a circle (and theres nothing else in the way) you can search horizontally (or vertically) for the longest distance between contrast points. this would give you the diameter as well as the coordinates of the origin.
but itll be much, much, much, much tougher if there as any noise at all in the picture.
(and sorry, i dont know any links - google it.)