I think one simple method of finding the center of a symetrical 2d body (probably any convex body)is to average the coordinates of points at the boundary. Threshold the image (it looks like it is already done somewhat) and sum all the x coordinates into an accumulator variable and divide by the total number of sampled points, this is the canter coordinate for the x axis, do the same for the y-value. any non boundary noise pixels will skew values. This method should give a decent center of your elipse (note the coordinates are in respect to the origin at the top left of the picture). The vision addonns most likely have prebuilt functions to do something like thiss but in a more complex and powerful way. This is just a quick thought on the topic but I could be wrong.