Correcting perspective

[Ed Richards]

(Assume that I have enough pixels so that after processing I have all the resolution I need.)

Is there any difference in an image made by shifting a lens to get a correct rendition of a interior, say a church, and tilting the camera and correcting in photoshop? My assumption is that if the camera is in the same place, you will get the same view once you correct for the back not being square. Say I am using my 47mm XL on 4x5, which does not allow much movement, and I tilt the camera to get more of the nave and less of the pews.

[Henry Ambrose]

Use up all the camera movements you have, then frame to get the rest if you must. Keep in mind that you'll loose some image area when you transform in PS, so frame loose. But then that causes its own problems when things are tight - i.e. you can't frame it very loose.

In this particular case you might also consider positioning the camera higher than normal to gain the higher areas and loose some seating. Where you place the camera does not have to correspond with the place where a person would or could normally view the room.

10 or 12 feet is not too high. (depending on the room) You can tie your tripod onto a tall step ladder if need be.

[Jeff Conrad]

I pretty much agree with Henry—do as much as you can with the limited movements, and correct what you must with PS. If it’s feasible to attain, an elevated camera position can help, but of course the perspective (in the sense of point of view) will be different. For example, when shooting the exterior of a building that has window sills from an elevated position, you’ll see the bottoms of the sills that are above the camera position, and the tops of the sills that are below the camera position.

As Henry mentioned, you need to discard a fair amount of the image when you correct in software. And if preserving the original shapes is important, the required correction may be more than you think—you need two corrections:


One to make vertical lines parallel, which is fairly straightforward and intuitive.
One to restore the original aspect ratio, which is less intuitive (at least for me).

If you’re careful, both operations can be done in one transform in PS, but I need to give it quite a bit of thought. There’s a good article (http://www.rodenstock-photo.com/mediabase/original/Entzerren_am_Computer_A4_e_Druck_7860.pdf) (PDF) on how to do this on the Rodenstock Web site. As luck would have it, the image is of the interior of a church.

[toyotadesigner]

You can tilt the camera and take two landscape shots, overlapping 25%. Then throw them into Hugin (Mac, Linux, Windoze) and stitch them vertically. Depending on the size and resolution of the original images this process can force your machine down to the knees and take hours of cpu time, but if it is an important image in a project I am using this approach.

But be warned: Hugin is not a click and drag application. If you want perfect results, you should spend some time with smaller images to understand the stitching and perspective control of Hugin. You might have to add custom points, which is a tedious task.

[Bob Salomon]

This will address your question directly and explain and show the differences.

http://www.rodenstock-photo.com/mediabase/original/Entzerren_am_Computer_A4_e_Druck_7860.pdf

[toyotadesigner]

The perspective correction in Photoshop is totally different from a correction in Hugin.

[Kirk Gittings]

This will address your question directly and explain and show the differences.

http://www.rodenstock-photo.com/mediabase/original/Entzerren_am_Computer_A4_e_Druck_7860.pdf

Perfect Bob! I was going to post a lengthy explanation but that PDF says it all.

[Bob Salomon]

Perfect Bob! I was going to post a lengthy explanation but that PDF says it all.

At Photokina the author was doing live comparisons during the show. Very informative!

[anglophone1]

This will address your question directly and explain and show the differences.

http://www.rodenstock-photo.com/mediabase/original/Entzerren_am_Computer_A4_e_Druck_7860.pdf

Really good, thanks for this!

[Brian Ellis]

Is there ANY difference? Yes. Is there a meaningful difference. Maybe, maybe not.

Here are three images of an old building (sorry they're not LF, they're being shown only to illustrate a technique). I thought the building was an interesting example of the way buildings used to be designed with a concern for aesthetics and how lmodern remodeling can mess the building up. A car was getting ready to park in front of the building and ruin the photograph so I quickly aimed the camera up and in the general direction of the building and clicked the shutter.

The first photograph on the left is the result. With no time to get the camera aligned even horizontally correct, and hand-holding the camera, the result isn't suitable for any purpose.

The second is the photograph of the building after making corrections in Photoshop. I can see the "stretching" effect of the vertical correction near the top and this effect likely would render the photograph unacceptable to a professional architectural photographer. But for many purposes I think it's fine from a perspective standpoint (and could have been even better if I had started with a better original).

The third photograph is an extreme crop of the part of the building that really interested me. From a perspective standpoint I think it too works fine for most purposes.

If I had a view camera, tripod, and about 15 minutes to set up and make the photograph, or if I had a tilt-shift lens on my digital camera, I could have gotten things perfect. But then I couldn't have made the photograph at all under the circumstances.

The point is that while corrections in Photoshop or a similar program don't allow one to make a photograph that's exactly like the photograph that could have been made with a view camera (or a tilt shift lens on a smaller camera), those adjustments can be used to make a photograph that's satisfactory for many purposes.

[Ed Richards]

Thanks Bob! This is excellent. I understand the transform issues and why they waste a lot of resolution. As long as have enough pixels going in, that should not be a problem.

I usually do not have the option of tall tripods and ladders when I do this shooting, but I am sometimes able to shoot from the choir loft. That solves the shift problem, but the images look fairly peculiar because it is an unnatural perspective for a viewer.

[tgtaylor]

My oh my - it looks like the view camera is still needed by the professional photographer.

Thomas

[Frank Petronio]

As long as you have overkill - crop and resolution-wise - you'll be fine. If you make huge moves then watch the transitions between skies and buildings, trees, areas of local contrast because that is where the artifacts and loss of quality might first begin showing up.

FWIW I try to do all the moves in one operation. The few steps you make, the less digital rot you get in your images. For example, don't do one move, then go and do another tweak. Instead undo and go back in your History steps and redo the first move perfectly....

You want to do all your big global overall changes right off and save the file as your working master. Transform, crop, adjust curves (black, white, midtones, grey balance all in one step), maybe hue/saturation, then image resize down to your working resolution. Save and done. Then attack the local stuff, spot the dirt, fix the blemishes, make the boobs bigger, take out the neck wrinkles, Liquify away the double chins, etc....

[Jim Jones]

The article that Jeff and Bob linked to is a thorough explanation, but perhaps too thorough. When planning on correcting perspective digitally, take one reference photo with the camera back vertical as well as the important one with the camera tilted up. Correct convergence by a grid in the image editor. Then adjust vertical expansion until the edited image has the same proportions as the reference photo. This takes most of the fun out of the procedure for mathematicians, but is fast.

[Henry Ambrose]

That is a good article to keep in mind when doing this kind of thing. But you're gonna do what it takes or you'll not have the picture you want.

Ed has a 47mm lens on a 4x5 camera so he's as wide as it gets. There is really nothing left, camera equipment wise, to make the picture he desires. If he wants more its happening in the computer. The good news is that a high res scan of 4x5 film gives lots of leeway to do the correction needed. Its not perfect but its what will work.