Oleynik failed to take lens distortion into account. Any photographer experienced with lens correction could have set him straight.
I seem to be the only one to bother to re-create Oleynik's example. Using the same linear transforms as Oleynik, I did not get the same drastic shift in the background,
Here also is my PSD (Photoshop file) if anyone wants to check (47mb) -
https://app.box.com/shared/static/xdu7kr9fnfzi011m3yd5dd693x20wgdq.psd
Using the same difference blend technique for reference, I achieved a more accurate overlay of the background. Comparison to Oleynik's blend-
But his main mistake is using
linear transform tools to compensate for
curvilinear lens distortion. Example- of curvilinear distortion-
http://physics-animations.com/Physics/English/len_txt.htm And his Photoshop process is arbitrary.
The linear transform tools he applied in Photoshop, "scaling, rotation, distortion, perspective, shift and offset," can not correct for subtle curvilinear (barrel/pincushion) lens distortion, inherent in most lenses. Post linear transforms may only correct
linear perspective, thus his process will not yield a practical comparison for determining distance.
Here is a demonstration by a photographer-
http://photo.net/learn/fisheye/ converting a fish eye capture to a rectilinear image. A fish eye lens exhibits more extreme distortion, but the same process applies. And he shows how curvilinear correction is a separate process from linear perspective correction.
Each lens has its own unique distortion 'signature.' The most accurate process for correcting lens distortion is "lens mapping"
http://www.leova.com/vfx_library/Heckman_Set_Documentation_Dec_2006.html#Mapping (scroll down to "Mapping Lenses") a common practice for accurate 3D camera tracking and matchmoving (which I do professionally). This involves photographing a "lens distortion grid" (examples-
http://log.ericalba.org/post/6110986710/lens-distortion-grids-pdf-for-download-vfx ) with the same camera used for final captures. The grid provides a clear reference for distortion correction. Without at least some distinct horizontal and vertical lines in frame, and relatively perpendicular to camera, one may only guess at compensating for lens distortion.
Here also is a straightforward video demonstration on "Working with Lens Distortion: Removing Distortion"
In Oleynik's later examples, he
does compensate for curvilinear distortion by using a more appropriate curvilinear warp (although arbitrarily again), but by that point he has confused the evident curved distortion with a curved cyclorama within the background, while failing to notice that even the foreground objects exhibit curvature (the vertical Rover antenna also appears slightly bent). He was too focused on the background, predetermined to support his conspiracy suspicion.
And as West mentioned, if the distant background is suppose to be a cyclorama as claimed, then the distant hills should exhibit no parallax when aligned and superimposed, since a cyc has no depth. But Oleynik's comparisons show distinct parallax between distant hills. So the apparent perspective shift cant be the result of some kind of backdrop.
I also received confirmation from the dean of Kharkov University that Oelynik's bibliography is accurate on the Aulis site, except they have never offered a major or minor in Physical Metallurgy, which Oelynik may have instead received from the Kharkiv Polytechnic Institute. Oleynik never completed his PhD at Karkhov, mostly due to a tough economic period in the country.