It's been some time but I got stuck in finding some tools to exactly match the zoomed in footage to the original 1024x1024 (recreated) still image. The best match is to take the HD 1920x1080 image, resize it to 890x569, then place it at (67, 428) on the raw image. This gives the best measure of matching which is called PSNR (with a value of 28.19 dB). I wanted an objective way of doing this that didn't rely on "eyeballing" what looks like points lining up.
By comparison, Mick's approximation would be resize to 878x583 and place at (74, 419), which I'm sure was a rough approximation by hand.
One of the difficulties I realized is that the sky is moving, so software can't match the whole image accurately. You need to crop out only the ground to get a valid match.
The reason I wanted to be precise, is to more accurately measure speed. Although you can measure speed by "fly lengths", the problem is the pixels are not square, and unless the fly is going in a perfect horizontal line, it could affect the speed measurement.
* The best matching was for these parameters, but the value itself is approximate, since I also matched the brightness and contrast roughly. Another analyst would probably get a much lower figure since the zoomed image has been brightened, and this measure is sensitive to changes in brightness. I could use a better measure like SSIM to avoid this issue.