Tilt-Shift Lenses vs. Photoshop Lightroom

Hehe, man, I admit I'm disappointed the result wasn't, "oh, wow! the Nikon PC lens is WAY better than Lightroom correction!" After photographing the Washington capitol building last night with a normal 35mm, I got all excited and ordered the 24mm PC lens (despite the huge hit to my bank account). However, while editing the 35mm photos and trying out Nikon's NX Studio thing, I noticed the "perspective control" slider, and to my embarrassment... the resulting effect looked totally natural, aside from messing up my composition a bit. I mostly photograph people, small animals, and other not-very-tall stuff, so I've never looked for a perspective control option in a post-processing program before, and didn't even consider it as an alternative to buying a new lens. I do prefer doing things via practical/mechanical means instead of editing, but, yeah, I'm gonna have a hard time justifying keeping that 24mm PC lens unless it's absolutely awesome now, hehehhhhh...

I still debate whether I should go the Canon tilt-shift route or stick with Lightroom/DxO ViewPoint.  My wide angle is the Canon 17-40mm.  I worked with an architectural photographer who used the canon 17mm t/s and raved about it.  But with correction, I was getting very similar results at 17mm and my software corrections.  Tis a conundrum!  

I grew up admiring modern American architecture, typically as photographed by Ezra Stoller.  As a biologist, I never bothered with tilt-shift (and was too cheap to abandon my Minolta camera and lenses).  Now, as a tourist and doing some photos of surfers with Olympus, I've been a happy user of Lightroom's tilt-shift, which has joined my checklist of things to do to make photos presentable.  Even part-way corrections can benefit a photo.

The small-sensor Olympus is a great tourist camera, compact and with image stabilization that makes hand-held shots in dim lighting feasible.  Putting a tilt-shift with an adapter on the little camera would probably be a travesty; better to get a Nikon or Canon body with a bigger sensor, then spend a few days in a big city with a tilt-shift.

Hey Todd, nice article.

I worked with a Sinar 4x5 35 years ago in school and quite a bit for three years in my professional work. As noted, shifting the plane of focus is something you can’t do in 35mm and it does have its uses. However, the biggest disadvantage of view cameras for field work is the pure bulk of the equipment and time to operate it. Hardly a grab-n-go proposition!

I use Transform on a regular basis for my architectural exterior and interior shoots. A few things I haven’t seen mentioned:

1. One big advantage of the digital process is the built-in lens corrections within Camera Raw. Believe it or not, the Nikkor 8–15mm ready-made profile can correct an image taken at 8 mm fisheye! Try it sometime.
2. Digital geometry correction is a two-step process, enabling the lens correction and using transform yields the best results.
3. Enabling “Remove Chromatic Aberration” is extremely helpful when pushing pixels around, especially for the geometric corrections you’ve described.
4. Bridge can do it too! Don’t think ill of me, but Bridge is an excellent workflow—especially in an enterprise environment. Transform is in the main header bar (like Lightroom, it used to be in Lens Correction). Same Camera Raw as Lightroom and Photoshop.

For architectural exteriors, I use a Nikkor 14–24mm lens on a D800 and raise it using a Lastolite Extending Handle, which elevates the camera to about 14 feet off the ground, then trigger the camera with a radio remote (all purchased from B&H, of course). I can usually get a high resolution image in the 2x3 ratio, but in some cases need to crop to a 3x4 ratio to retain subject matter.

As an architectural photographer for more than 40 years, 25 of which were spent behind a 4x5 camera, I welcome the flexibility of utilizing Photoshop to correct perspective distortion.  While TS lenses are cool, you have to purchase a lot of different lenses to capture architecture from differing vantage points.  With my 4x5, I used 65mm, 75mm, 90mm, 135mm, 150mm and 210/360mm lenses depending upon how I wanted to represent the building or interior.  I now shoot - on a Nikon D800 - with a 16-35mm zoom and simply correct perspective in Photoshop.  My clients aren't pixel peepers -nor am I.  

Great article, Todd. For myself, as a tourist shooting archy, I tend to the wide angle, allowing for the future post-processing loss of some detail. I don't really need to see the layer of cement holding the structure together.

View cameras (mentioned at the head of the article) offer another advantage not available from PS or the lenses mentioned: they can select a "plane of sharpest focus" that is independent of the "plane of no parallax". Software does a lousy job of correcting focus and the lenses don't have enough freedoms of motion to do it all. I recommend that photographers read about view cameras or talk to some one who is familiar with their operation and theory of use. It's a great way to improve photography with standard equipment since you become better aware of the trade offs made by framing, focusing, and camera position and orientation decisions.

I've done comprehensive tests to compare using a tilt/shift vs. using software.

What I found was that architectural details are not treated the same, and software corrects distortion unevenly from the center to the edges.  This can lead to irregularly shaped objects such as windows, not being a perfect rectangle a they usually are.  It's definitely visible without pixel peeping.  



I still think the best way to shoot architecture to get perfect proportions is to use a view camera, something I've started to do.

Dear Todd, there's one major mistake in your article: you write "Keystoning occurs when vertical lines converge as the LENS is tilted above or below the horizontal plane" or "Tilting a LENS vertically emphasizes this distortion". I understand that it's quicker than explaining that it is the non-parallelism between the sensor/the film and a specific plane of the subject the real reason for keystoning, but it leads to a substantial misunderstanding: people may think that mounting a T/S lens will prevent them from keystoning, while it is obvious that no lens may obtain parallel lines if the sensor and the plane of the subject are not parallel. By the way, you have horizontal keystoning to, when you turn the camera - therefore the sensor - to the right or to the left instead of up and down.

Best,

saverio

There were two reasons why I bought a couple of tilt-shifts. 

One was because the previous own was a pro who'd decided to quit and sold all his gear - and when he bought it, he must have been suffering from an extremely severe case of GAS, because both the lenses I bought (and the one I didn't bother with) were virtually untouched.  So I have two virtually brand new Nikon t-s lenses (the 24 & the 85), for way less than the normal price of either of them.

The other was because I do a lot of what might best be described as architectural photography.  And I'm fed up to the back teeth with losing too much of the picture, during the process of correcting verticals.  It is FAR better to capture the image you want to end up with, in camera, before ANY PP starts.  Then if there's any need to crop, it's going to be minor and you won't lose key parts of the subject matter.

I am delighted to learn there's little or no discernable difference in image quality.  But there's a hell of a difference in framing the shot - and I am thrilled to bits with the contribution the t-s lenses make to that part of the process.

BTW - if anyone's looking for a program to correct verticals, DxO make a great program for that - ViewPoint 3.  It's generally possible to let it do the corrections on auto.  Occasionally it slips up, because it can't grab enough info to run the program on auto, but then it's very user friendly (once you get the hang of it) doing it manually in ViewPoint.  And you can also tweak horizontals, to correct the perspective, if you want to.

I have been struggling with this problem for 50 years.  I used to use a Curtigon Schneider perspective correcting lens with my Pentax Spotmatic which helped some.  I also tried tilting the easel in the darkroom but it only did a partial job and only if I was square on.  I guess an angled tilt would work but it was difficult to achieve.  Digital correction works much better for two reasons.  (1) you are rarely square on when you take the picture as you were at Borough Hall or the courthouse around the corner.  You are usually at a slight or even great angle.  Thus there is both horizontal and vertical distortion.  This can be easily corrected in photoshop by using the distortion command.  (2) correction changes the proportions of the subject.  This is obvious in you pictures of the courthouse but you do not address it.  The central semi-circular block has totally different proportions in the shift lens and digitally corrected pictures.  After making a digital correction to the converging lines at the top, you need to elongate the subject by stretching it up or down.  Alternatively, make the correction by widening the top and narrowing the bottom.  Incidentally the cropping loss is frequently of insignificant matter or sky which can be cloned in.

You should also note that the distortion only occurs with a wide angle lens.  With a 55 mm the perspective in the picture will match the perspective we perceive in eyesight.  With a longer focal length the distortion is actually in the opposite direction.

I am sure that an optical solution would be better to a digital one just as I am sure that film is better than digital and that a view camera is better than a 35MM.  But given the limitations of all these, digital correction seems to work best.

The comments about creating software corrections by starting with excess resolution are very true, which is why I use a 36mp camera. It’s very important to let the correction software reduce the total pixels, and not to up-sample in any of the software one is using. It’s also important to do the corrections off the raw file if you can, and to do it with the correct method, which involves computed aerial projection. Avoid fast, simple methods like distorting only along one side, as you completely distort the proportions of the subject. Drawing lines on key edges (DXO, or the simple method in Capture One) almost never delivers the exactness I require. I don’t know which method you used in your examples, but the photos digitally corrected show some proportion distortion. Aerial projection is the most “accurate” or similar to using a shifted lens.

In Camera Raw, that means going to the lens corrections tab and the manual sub-tab. In this tab, you can drag sliders, but exact work is best done by putting numbers in the Vertical and/or Horizontal number boxes. The Rotation slider/number box allows exactness to one-tenth of a degree, pretty phenomenal. Capture One has even finer number inputs for its aerial projection approach. Enable the grid display and be patient with these adjustments; it takes some time to figure out a complex correction, especially because it is almost impossible to level a tripod to perfection. I find if I set up carefully, I will almost always still need some small bit of rotation with a vertical correction. I use horizontal correction much less, but it is very useful when needed.

The wider the lens, and the more it is tilted, the greater the correction needed, and the greater the digital artifacts created at the corners with software correction. There is a limit to what is acceptable, and if you find yourself at a 50 or 60 setting in the Camera Raw vertical correction box, results at the corners are going to suffer. No question, delivering the same angle of view with a good T/S lens will be superior to shooting a wider capture and then correcting in software. But digital corrections make any lens a functional shift lens, and that includes the focal lengths not available in T/S. Anything longer than 28mm or 35mm adjusted digitally makes results that are virtually indistinguishable from an equivalent T/S lens. Use fixed focal length lenses to avoid the distortions introduced by zooms.

A few thoughts (from an architectural photographer):

In Lightroom (and other software like Capture One) you can get the look of the corrected image even closer to the shifted image by using the aspect slider. The shifted image is taller and stretching the software corrected image vertically will align this.

Getting a hold of the narrower field of view in the digitally corrected photos is no problem if a wider angle lens is used. Nikon’s or Sigma’s 14-24 mm lenses are great examples. And in real world scenarios somebody who wouldn’t have a tilt-shift lens might rather have one of those.

In the end the needed final resolution can even out the sharpness differences. If you shoot with a current DSLR like the Nikon D810 or D850 you have plenty of pixels to work with and most clients will rarely need that much resolution (even though they think they do ...). So once you scale down to, say 24 megapixels there won’t be any issues with a digitally corrected  image.

Lastly, even though many declare it a matter of taste whether images are keystone corrected or not, in architectural photography a main reason to correct is that especially contemporary architecture often features slanted lines and only in a corrected image will you be able to truly discern them from real verticals ...

As an architectural shooter, I find that for interiors, quite often 17mm TS is too wide and yields too much depth/distortion. My approach is to shoot with a 24mm TS, but in 2-3 shifts to capture the entire space. Then I stitch them together in post. This gives me a very wide field of view, but with a more compressed look than I would get with a 17. I think that TS is a better approach than pixel interpolation.

Oh, by the way, the Tilt-Shift lenses that I've been using are also much sharper in the corners that normal wide angle lenses (in general).

I'm a landscape and seascape photographer so I don't see buildings in my photos very often.  However, I use the my Tilt Shift lenses in the Tilt position to get a deeper field of focus.  While I'll leave it to you to research on the Web, the Scheimpflug principle allows us to change the plane of focus thus giving us deep focus.  It takes quite a bit of practice and some patience with the math but without doing focus stacking, I don't know of any way to do it in Lightroom, Photoshop, or any other piece of software.  Do you?

I have both Canon and Nikon tilt-shift gear.  I just got the Nikon 19 tilt shift and so I'm anxious to try it out.  I had been using the 24 tilt-shift but it was limited to a single axis and the new 19 works more like the Canon tilt shift gear.

I require perspective control for mountain landscapes and I also need very large format images (380+ megapixels typically).  My solution is to shoot heavily overlapping images with a high quality prime lens, usually my Nikkor 105 on a D800e, and stitch them with with a program such as PTGUI.  I can adjust the projection and apparent eye height in that program and, if needed, do a further shift in photoshop.  The extra pixels compensate for any loss and I can shoot as wide as I wish.  It takes many images and an hour of computer time to grind through it, but it is worth it if the product is a 4 x 8 foot 300 dpi image.

Good article Todd. In comparing tilt-shift type lenses vs digital processing, I'm wondering if you've tried the comparison with any of the dedicated perspective correction plug-ins. For example, I have DxO Viewpoint2. I haven't compared it to the correction functions of Photoshop or Lightroom or to a tilt-shift lens.