Comparing Keystoning Correction With Large Format & 35mm

[Andrew O'Neill]

 

[Ben 4]

Great demonstration!

 

[Alan Edward Klein]

Correcting these things in Lightroom are a snap. But you do lose edges of the picture. So you have to shoot wider to account for that. Fortunately, I mainly shoot landscapes, not buildings, so it's not much of an issue.

 

[bernard_L]

@Andrew O'Neil

Please take a second look at the "8x10" and "35mm(corrected)" images side by side. The 35mm image has parallel verticals, sure. But is is vertically compressed compared with the "8x10".

To confirm this, I cropped identical (as best I could) portions of each image, scaled (same factor in both directions, of course) the 8x10 image so that its width matches the 35mm, and put them side by side.

35mm left; 8x10 right

The vertical compression of the "corrected" 35mm image confirms my suspicion that the perspective correction in PS (or whatever you used) merely stretches horizontally the top of the image to make verticals parallel.

This ignores the vertical foreshortening. Imagine you take a picture of a tall business building, the center of the pictures being, say, on the 6th floor. With a view camera, full perspective control, lensboard and film vertical, all windows of the building will have the same size on film. With a 35mm camera just looking up, the 9th floor windows will be vertically foreshortened relative to those of the ground floor. And that is not corrected by stretching apart the verticals.

This shortcoming has mild consequences in the video, because the aiming tilt is not very pronounced. But one should be aware of the difference between a true perspective correction and just making the verticals parallel.

To end on a constructive note. How to do a proper correction in post? I use a panorama software, PTGUI (shareware) of Hugin (freeware). Panorama?? Yes, with just a single image. And use the "align" tool to re-define the location of the horizon.

And, to end on a positive note, despite my nitpicking. thank you for your informative and stimulating videos!

 

[Rayt]

Keystone correction of course isn’t a substitute for T/S lenses or LF movements. Unless you actually shoot the same scene with both methods as in the video and compare side by side the viewer would need to look hard to see the flaws. I shoot with in camera perspective control and it works very well.

 

[wiltw]

@Andrew O'Neil

Please take a second look at the "8x10" and "35mm(corrected)" images side by side. The 35mm image has parallel verticals, sure. But is is vertically compressed compared with the "8x10".
...The vertical compression of the "corrected" 35mm image confirms my suspicion that the perspective correction in PS (or whatever you used) merely stretches horizontally the top of the image to make verticals parallel.

This ignores the vertical foreshortening.

I have long known that I have to resize a photo in ONE dimension differently from the other axis, in order to restore the result of postprocessing correction of converging lines. The amount of correction does vary, but often falls in the range of 120% to 130% in one direction. I often will try to restore the relative size if the compression is too obvious, but many times I do not try to correct when it is not too apparent.

 

[wiltw]

I just shot a paper cutter grid with a P&S set to WA, and this is what resulted...

1. the incoming RAW image depicted converging verticals
2. using Perspective Correction tool of Paintshop Pro, the squares of the grid came out as rectangles (measuring 20.55mmV x 19.5mmH on my 27" monitor)
   note also the alteration of the aspect ratio of the shot, which is 4:3 before perspective correction and ends up closer to square.
3. If resized the horizontal dimension by 120%, while leaving the vertical dimension at 100%, the aspect ratio of the shot was restored (somewhat)
   but the measurement of the squares shows them to be rectangles with more width than height

Here are the three stages described above

Pardon the barrel distortion caused by the P&S lens at WA setting!

 

[Alan Edward Klein]

@Andrew O'Neil

Please take a second look at the "8x10" and "35mm(corrected)" images side by side. The 35mm image has parallel verticals, sure. But is is vertically compressed compared with the "8x10".

To confirm this, I cropped identical (as best I could) portions of each image, scaled (same factor in both directions, of course) the 8x10 image so that its width matches the 35mm, and put them side by side.

35mm left; 8x10 right

View attachment 357718

The vertical compression of the "corrected" 35mm image confirms my suspicion that the perspective correction in PS (or whatever you used) merely stretches horizontally the top of the image to make verticals parallel.

This ignores the vertical foreshortening. Imagine you take a picture of a tall business building, the center of the pictures being, say, on the 6th floor. With a view camera, full perspective control, lensboard and film vertical, all windows of the building will have the same size on film. With a 35mm camera just looking up, the 9th floor windows will be vertically foreshortened relative to those of the ground floor. And that is not corrected by stretching apart the verticals.

This shortcoming has mild consequences in the video, because the aiming tilt is not very pronounced. But one should be aware of the difference between a true perspective correction and just making the verticals parallel.

To end on a constructive note. How to do a proper correction in post? I use a panorama software, PTGUI (shareware) of Hugin (freeware). Panorama?? Yes, with just a single image. And use the "align" tool to re-define the location of the horizon.

And, to end on a positive note, despite my nitpicking. thank you for your informative and stimulating videos!

I tried correcting the 35mm original with Lightroom using only Transform auto, aspect and rotate and got this. Is this better, the same,?

 

[bernard_L]

1. using Perspective Correction tool of Paintshop Pro, the squares of the grid came out as rectangles (measuring 20.55mmV x 19.5mmH on my 27" monitor)
   note also the alteration of the aspect ratio of the shot, which is 4:3 before perspective correction and ends up closer to square.
2. If resized the horizontal dimension by 120%, while leaving the vertical dimension at 100%, the aspect ratio of the shot was restored (somewhat)
   but (...snip ...)

Kudos for performing an actual experiment. As you realized with a P&S the true distortion of the lens interferes with the "distortions" of projection.

There is more to perspective correction than fixing the convergence of verticals plus a scale factor. Details aside , you can understand that the vertical scale correction varies for individual squares (or windows) depending on distance and slant angle (foreshortening).

 

[bernard_L]

I tried correcting the 35mm original with Lightroom using only Transform auto, aspect and rotate and got this. Is this better, the same,?

Hard to tell because:

• In the example shown by Andrew in his video the perspective effects are moderate.
• Once the average value for the vertical scale factor has been applied, only second order effects remain.
• Images grabbed from the video are not very sharp.

My point was not that the proposed 35mm correction was terribly bad but to raise a flag about the correct method. The difference will be significant in situations like 24mm (equiv) FL staring up at a building at 45degrees.

 

[Sirius Glass]

For Keystoning I use the f/2.8 28mm Nikon lens on my Nikon AF cameras or the moderate corrections available with my 4'x5" Pacemaker Speed Graphic. However for my Hasselblad, I have not prespective correction other than tilting the easel since I do not have enough need for it to justify carrying a tripod and buying a Hasselblad ArcBody or FlexBody camera. <<Grrrr …>> Still I appreciate and need to look at keystone corrections with LF cameras. I may need to think about buying a Hasselblad ArcBody [would need a new set of Rodenstock lenses with more capability]or FlexBody camera [using my lenses with less capabilty and less flexibility].

 

[ic-racer]

Correcting Converging Verticals With the Enlarger

I did this print a while back. I was using a 210mm Symmar-S on my 8x10 camera. That lens does not offer any movement on 8x10 format, so I had to tip the camera up to make this image. Back in the darkroom I setup the enlarger like below, and made the second image. One nice thing about this is...

www.photrio.com

 

[bernard_L]

Nice! A purely analog, and geometrically correct, way to obtain a posteriori the same result as in-camera.
Further down in the thread that you quote, a remark by Ian Grant, on the aesthetic rather than technical level.

Perhaps though total correction can look false, we don't see it naturally like that although our brains mentally correct.
First I'm not talking about the example here. You can find plenty of examples of images of churches, looking up and corrected that exaggerate the height, so it's a balance to find what's most natural or suits your vision.

 

[bernard_L]

Brunelleschi and Alberti had it figured out by the middle of the quattrocento. No Internet, no Photoshop.

quoted from: https://www.silviaminguzzi.com/anotherperspective/linear.htm

No amount of left-right stretching of the top of the mage is gong to make all the squares... square, as would be the case if using a view camera with film and lensboard parallel to the plane of the tiling (it's a floor, not a wall, but that does not matter in this discussion).

 

[Jim Jones]

My old DeJur Versatile Professional enlarger has a tilting film stage for correcting perspective. The film stage in many enlargers can be shimmed to accomplish this. When I have to tilt a camera up to photograph a tall building, I also take one shot with the camera back vertical. I can compare this latter shot when correcting perspective in digital printing.