Hi, sorry I did not notice there were responses here.
Steve, there is no software correction of any kind applied here. The upward shifting of the sensor in the second image allows the camera to be tilted lower for the second image, bringing the more distorted upper periphery closer to the middle of the field where parallel lines are less affected by the distortion. I assure you, there is no trickery here. I had hoped to get out over the weekend to shoot a few more examples, but I had other engagements.
I thought calicojack's explanation was a good one.
---------- Post added 10-20-2014 at 02:24 PM ----------
Originally posted by vonBaloney Yes I think it is confusing because it is presented like the second image was just a sensor shift different than the first, but obviously the sensor was not parallel in the first image (and it looks like it could have been made so without using sensor shift). Right?
I thought this was pretty clear:
"re-composed to match the composition".
If you merely shift the sensor up and do not recompose, the distortion is still fixed, but you would be looking at much more sky, obviously.