Free Transform with Lens Correction

[PWeareB]

Hello All,

My question is about using the Free Transform tool after applying the Lens Correction filter, when correcting converging lines in architectural photos.

After using the "vertical perspective" correction by as much as -50 to -90, buildings may look short, or "stubby."  That's when I use the Free Transform tool to drag the top or bottom edge of the image up or down to give the building a more natural-looking scale.  Usually somewhere between 102% to 105% on the scale bar does the trick, but I usually just go by feel.

So... do others use the Free Transform Tool in the same way?  If so, is there any precise formula you use based on the amount of correction you've used, or height of the building, to adjust the scale?

Thanks.

Patrick

[davescm]

Hi

Take a look at this, it may help you. It was written by Rodenstock so is pushing the advantages of Tilt & Shift lenses - but the Photoshop techniques are well written and do work:

http://www.rodenstock-photo.com/Archiv/Perspective%20Control.pdf

Dave

[D Fosse]

As far as I can tell two important facts are omitted here:

• even a tilt/shift lens is "incorrect" because of rectilinear correction. This is the familiar "corner stretching" we all know from wide angle lenses - but what it actually is, is progressive stretching away from the center. This affects all lenses to varying degrees, except fisheyes. The wider the angle, the more progressive stretching. A tilt/shift lens has an extra large projection area, so it's effectively "wider" than focal length would indicate. It's this "wider angle" you tap into with shifting.

• it's possible to mimic a tilt/shift lens very accurately by making two exposures - one horizontal shot without distortion, and one tilted to extend upwards. Then align the tilted shot to the horizontal one, keeping as much of the latter as possible. True, you will lose sharpness in the distorted top part - but you can compensate that by scaling down the horizontal part a bit before running auto-align. This way most of the transformation is "inwards" not "outwards".

[davescm]

That second technique with two shots is one I'd not thought of and like you say you will get the alignment spot on. Nice one !

Dave

[D Fosse]

That's how I did this one, and this is how a, say, 35mm tilt/shift lens would do it (originals were shot with a 50 mm). But you can still see it's stretched towards the top. If you want to keep straight verticals, I don't think there's any way to avoid that.

The real headache here is parallax error. Horizontally you can control that with a focusing rail, pulling the camera back so it rotates around the optical center of the lens. But vertically it's very difficult because the tripod head rotates at a point way below the camera. I'm working on that problem. At the moment I'm thinking of an L-bracket with a separate ball head mounted horizontally - then you can use the focusing rail on that. Phew...

(of course you can just tip the tripod head sideways, but then you get portrait shots. I want landscape).

[davescm]

There must be something commercially available (or is it a job for Trevor and his Menz Shed).

On thing I do find - in correcting, I often leave a very slight convergence on the verticals. It somehow looks "right" compared to verticals perfectly corrected by the grid. Maybe it is just me .

Dave

[D Fosse]

Yeah, that's just you   No, seriously, sometimes it pays to not overdo it. Good point.

There's always something commercially available. My focusing rail is made by Manfrotto, ingenious little device, and they seem to make a little clever add-on for every possible twisted desire. My local camera shop can probably get it - if it was possible to make sense of their website. I have a hunch the L-bracket is the way to go - after all, what I basically want to do is just turn the camera body 90 degrees.

[PWeareB]

Hi All,

Thank you, very good article. I too leave slight convergence, depending on the building. Sometimes the perfectly straight lines look awkward, especially on very tall buildings.

I like the idea in the article of placing three guide lines on the distorted original photo: at the center, at the horizon line, and one in between those two, and then as one adjusts the vertical scale with Free Transform, keeping the elements at those lines where they orignally were in the distorted version of the photo.

I have just tested it on an image and it came to a vertical adjustment of about 106%, which is a little greater than my normal eyeballing of about 102% to 105%.

Thanks!

Patrick

[Trevor.Dennis]

Dag, rigs like the Nodal Ninja are going to rotate around the lens nodal point in vertical as well as horizontal.  I've got a Chinese knock-off of the NJ but I don't like to use it, as with so much extension from the pivot points, it naturally tends to vibrate.

I found this when looking for the NJ screen shot, and had not seen it before, and it does look to have more support, but they are going to be inherently prone to vibrate unless radically over engineered.  

Gimbal heads like the Wimberly are not designed to rotate around a lens nodal point, but they are far less prone to vibration, and I suspect they could be used to eliminate parallax without being too far out of balance.  It's the balance that cancels vibration.  A tripod rig that has always made me shudder because it is so wrong from an engineering view point are the pistol grips with the pivot at the base of the grip. 

[D Fosse]

Thanks for the tip, Trevor, but yes - these do look quite a bit shaky.

The more I think about it, a simple L-bracket should work. Ideally, it should just move the thread mount from the camera base to one of the sides, aligned to the lens. Then tip the tripod head over. Problem solved...or?