Photographing rectangular objects

I’m photographing some of my large artwork (average 48" x 32"). I’m using a Nikon D90 camera with an AF-S Nikkor 18-200mm lens. It’s the only lens I use any more, and I’ve never encountered this problem before.

Rather than showing the artwork as rectangular, it’s causing the edges to become concave, making them look like the corners are pulled out. I’ve set the camera on manual as well as automatic, and made all sorts of adjustments. No matter what I do, this effect still happens. It’s probably the lens, not the camera, but I don’t know what else to change.

It’s the lens. Zoom lenses are particularly bad at this application.

If you have photoshop or Lightroom, Adobe publishes ‘lens profiles’ for most common lenses that will compensate for most or all of the distortion automatically.

They create this by shooting special grids and multiple focal lengths and apertures, then letting the software figure out the tweaks needed to straighten all the lines. That info is then encoded I to a small file that Photoshop or Lightroom can use after reading the shot details in the EXIF data of your files. Pretty slick!

The 18-200 is a very nice lens (I have one), but it’s a) not particularly sharp, and b) has significant distortion, particularly at the wide end.
Photoshop can take care of the most egregious distortion, but if you are doing copy work, the residual distortions might still be to much.
Pros use a “prime” (non-zoom) lens designed for low distortion.

The only good way to shoot such things without a top-drawer lens is to back up and zoom in - put distance to use in helping flatten the image. Zoom lenses that will show barrel distortion close-up and at wider angles will flatten out at the other end of the throw.

Shoot from a distance and use Photoshop, with a lens correction profile if you can get one, or the truly fabulous perspective correction tool to completely flatten and square the images.

Note that this is not due to a defect of the lenses, but to a fundamental aspect of geometry. A field of view is actually spherical, not flat, and so any method of turning it into a flat image is going to involve some form of distortion. It’s the same problem as converting a globe to a map.

The lens correction function really does a nice job with this though, I promise.

I’ll second that, but it does require a newer version of Photoshop, and a lens profile. It’s not clear if the OP has both.

I recently had to shoot some archival documents under very poor conditions - bad lighting, cramped room, and framed on the wall. It got down to crazy-ass stuff like me having to duck out of the way because the flash angles made me a reflective blob in the glass. So I ended up with very high-res shots, but at skewed angles and with some color balance issues.

Perspective correction, using the corners of the frames for reference, did an ASTONISHING job of flattening and squaring the document images. Absolutely no detectible distortion of the images and handwriting. A high-art shot might be more demanding, but it’s worth a try, along with specific lens-mapped correction.

True enough, but I believe you can still download a thirty day free trial of Lightroom. OP could shoot all the needed Images, get the free trial and go to town.

Yeah, try this first. Zoom lenses are more prone to distortion than primes, and this usually shows up as “barrel distortion” (where the image bows outward) at the wide end and “pincushion distortion” (where the image bows inward) at the telephoto end. Looking online, the 18-200 f/3.5-5.6 (assuming this is what you have) has an interesting distortion curve, with strong barrel distortion at 18mm, about flat at 24mm, and pincushion distortion setting in at 28mm upwards.

(ETA: Oops, sorry, that’s the Canon version of that lens. Let me see if I could find info on the Nikon version and its distortion characteristics.) If I were you, I would use something like a 50mm f/1.8 D lens which is a) inexpensive and b) shows very little linear distortion.

ETA2: Looks like the distortion characteristics are similar for the Nikon lens. Somewhere around 28mm-35mm is where the distortion switches from barrel to pincushion.

Said another way, somewhere in that range distortion is closest to zero. So shoot at that zoom setting if possible.


At what zoom level(s) have you taken the pictures?

Also, can we see the works?

You don’t have to have a lens profile to correct distortion in Photoshop. Just use Filter: Distort: Lens Correction and then Settings:Custom. I find Remove Distortion set to -3 or so is right for things I’ve shot with a wide-angle lens.

Thanks so much for your help, guys. I knew the problem was with the lens, didn’t realize that the phenomenon was “normal,” and that there was a fix for it. Turns out, I already had Lightroom on my computer . . . then I discovered that the new Photoshop actually includes a lens correction filter. So I tried both, with comparable results. They both tend to slightly overcompensate, though.

Now that I’ve finished, I’ll go back and try Mr Downtown’s suggestion.

I’ll post the works when I finish playing around with them in Photoshop.

True, but that’s a relatively blunt instrument. It can’t correct the more subtle distortions due to waviness and/or non-linear distortion. Lens profiles are a built in part of LR and Photoshop, so no reason not to take advantage.

This (though the other comments in the thread are more useful to your needs). If you look up at the intersection of your ceiling and a wall, for example, if you turn your head to the left, you’ll see that the intersection angles down and the left, towards the horizon. If you turn your head to the right, you’ll see that the intersection angles down and to the right. But right above you, the line seems straight and parallel to the ground, not angled down at all, yet both of the first two viewings were clearly angled downwards (in opposite directions). So obviously the line is curved and our brain just interprets it as straight. We know that the line is actually straight, so our brain ignores the curve to match the reality of the situation rather than the reality of optics. It’s unable to do that when looking at a 2D image with a nice rectangular border around it, so the optics win in that instance.

An alternative is to shoot several overlapping images - 6 or 9 should do it, then use photoshop to render them as a single picture.

I’ve actually tried that, with about 10-12 shots. The edges were fine, but the middles got crazy.

The correct tool for this is ptgui or the (expensive) autopano.
You need to correct for angular distortion, which is much harder in Photoshop.

A view camera is the best camera to use for this. Look at the “rise and fall” subsection for the explanation. The main downside (besides being expensive) is they typically use a 4"x5" film format, but with digital technology as it is today that may well be a thing of the past.