Understanding movements better to correct perspective.

[dikaiosune01]

I have a nagging that this has been asked before. I do apologize and ask that you bear with me. I freely acknowledge that my thread searching skills are not up to snuff; as I gave up after three variations of searches.

I've been shooting 4x5 for about 3 years now. However to this day, I don't fully understand how movements can correct perspective. I've done multiple wall tests and building tests; and usually after an hour of head scratching I give up without shooting a single shot because I couldn't get the result I was going for.

This was the exercise I was going for:
When shooting a building from ground level, the resulting image will look like a trapezoid, with the top tapered. This is a result of angling the camera upwards, shooting up at the building. My exercise is to straighten the building into a proper rectangle.
If there are lamp posts in the foreground or other buildings in various planes in front or behind the building, is it possible to straighten those secondary subjects as well?
Are the limitations of movements limited by the image circle? Movements? Or both? And I need to better choose my subject that are within the limitations of my equipment.

Thank you

Shen Hao 4x5 135mm and 180mm lenses with image circles large enough for 5x7

[Tin Can]

I am no expert, but it does seem to me that perspective control is limited to one plane. Yes that plane may be at any angle, but it is not elastic. As soon as you introduce another plane, focus and perspective change.

[Struan Gray]

If you keep both the back of the camera and the lens panel vertical, everything that is vertical on front of the camera (buildings, lamposts, signs etc) will end up parallel in the photo.

The classic way to show this is to take two photographs of the same scene. First, by tilting the camera up to include the top of the building. Second, by levelling the camera and using shift to move the image on the ground glass until you get the framing you want.

Most people start by shifting the lensboard up. When you run out of up-shift on the lensboard, some cameras allow you to shift the camera back down. When you've maxed out both of those, you can get even more shift by pointing the camera up a bit using the tripod head, and then returning the lensboard and camera back to vertical by tilting them on the rail/baseboard.

All that really matters is the relative position and orientation of the lensboard and the camera back. How you get there is up to you.

Somewhere you will run into the limits of your lens' coverage. Either you get a black horizon (common with most modern lenses) or you get a fuzzy image which cannot be sharpened by stopping down. In the latter case, it's a question of taste where the limit lies.

If you want to practice, it is easier to see the effects with a normal lens, since you avoid vignetting and the hot-spot which results if you use a wide angle lens onto a plain piece of ground glass. In fact, one of the little-appreciated joys of LF is that you can make a normal lens look like a wide angle by shifting to the edge of it's image circle.

Finally, since you're in Hong Kong, you can see the opposite effect too: go to a high building, or a lookout on the mid-levels and point your camera down at a group of tall buildings. Now you can see the tops splay outwards rather than inwards, which can be corrected by levelling the camera and shifting the lensboard down.

Note that you can force the image geometry so that, say, a set of windows on a high building remain rectangular. You cannot avoid the change in perspective as you move from the ground floor to the top of the building - you will see the ground floor windows and the rooms inside head-on, but on the upper floors you'll be looking at the underside of the lintel, and seeing the ceiling of the rooms inside.


There are similar effects for sideways shifts, but they are used less often because the perspective effects of lines of houses receding into the distance are expected by the viewer. They can be useful if you have a restricted viewpoint off to one side and want to show a head-on view, or if you just want to adjust the image geometry to give a particular impression. I have used it on occasion to give a more graphic, two-dimensional feel to a photograph of woodland by adjusting the vanishing points of the perspective.

[Ed Bray]

Whilst I am not a long term user of Large Format, I am pretty sure that only the Camera Back needs to be vertical to ensure that buildings remain with the correct perspective, whilst tilting of the Lens Standard only effects the plane of focus within the image.

[polyglot]

Think of it like this: the further away an item is from the camera, in a direction perpendicular to the film plane, the smaller it is. Simple perspective.

If you point the camera horizontally then every part of a vertical building wall will be the exact same distance from the film along that horizontal axis. Therefore the top and bottom parts of the building are rendered as equal sizes and it has no taper. You can test that trivially with any camera; no need for a view camera.

Now if you point the camera upwards, the top of the building will be further along that film-perpendicular axis than the bottom of the building, so the top of the building is rendered smaller. It's tapered.

You can think of perspective correction using shift as just cropping. Shoot a really wideangle shot of your building with a level camera, then crop out all the foreground dirt. That's what you're doing when you shift the lens up (or the film-back down) - you're recording just the top part of that larger image. The walls and streetlamps are all vertical because the film is vertical. That's all there is to it.

[rbultman]

Get the Kodak book Photography with Large-Format Cameras. There is a section on movements that is very will illustrated, better than the treatment in Adams' The Camara, IMO. That can be had used for less than $20. Even better is Simmons' Using the View Camera. Used copies of this book are insanely expensive (IMO).

To repeat others slightly:
1) Start with the bed horizontal and level and the front and rear standards vertical. Remove any swing, shift, or tilt.
2) Adjust the front rise until the subject is properly framed vertically. If you run out of rise or lens, meaning the corners are no longer sharp due to exceeding the movement allowed by the lens image circle, remove any front rise and tilt the camera bed until the subject is framed vertically. (If your camera also had rear rise/fall, you could adjust that until the image was properly frame vertically.)
3) If you tilted the image bed in order to vertically frame the subject, the verticals will no longer be parallel - this is your trapezoid problem, caused by the rear standard not being parallel to the building. Tilt the front and rear standards until they are vertical. This should correct the trapezoid problem and any focus issues. If the secondary vertical objects are also parallel, originally, to the vertical lines of the building, they should also be corrected now.

In short, try to keep the front and rear standards parallel to the plane you wish to keep in proper perspective.

[TheToadMen]

Maybe this will help:
http://en.wikipedia.org/wiki/View_camera#Movements

http://www.bhphotovideo.com/find/Product_Resources/largeformat2.jsp

http://www.largeformatphotography.info/how-to-focus.html

http://www.toyoview.com/LargeFrmtTech/lgformat.html

http://paulturounetblog.files.wordpress.com/2007/08/view-camera-movements.pdf

You can also make a search for: "The Scheimpflug Principle".

Bert from Holland
http://thetoadmen.blogspot.nl

[Bruce Watson]

When shooting a building from ground level, the resulting image will look like a trapezoid, with the top tapered. This is a result of angling the camera upwards, ...

Not exactly. It's the result of the film plane being tilted in relation to the plane of the building.

Start with the film plane being level and plumb. Use a level to be sure. This will in turn typically result in lots of front rise.

Often you'll have to tilt the camera base, then level and plumb the film plane, then use front rise -- all this to get enough front rise to capture the view you want. This in turn can introduce some vignetting as you get close to the edge of the lens' image circle. It can also introduce some vignetting from the bellows scrunching up -- something to watch for, especially with short lenses. Just sayin'.

All this said, some tapering down to the top of buildings is often desirable. If you nail the verticals so they are exactly parallel, they can often look unnatural in the final print. How much tapering is artist's decision of course.

Trees OTOH, tend to want to be exactly parallel in my experience, unless it's clear that you are really looking up at them, as in you are close to them. Probably has to do with the vast experience people have with how a forest actually looks. Similarly to how people don't tolerate distortion in human faces maybe.

[Struan Gray]

Whilst I am not a long term user of Large Format, I am pretty sure that only the Camera Back needs to be vertical to ensure that buildings remain with the correct perspective, whilst tilting of the Lens Standard only effects the plane of focus within the image.

True. I was going to add a bit about putting the plane of focus where you want it, but I'd already waffled on long enough :-)

Short form: keeping the lens board vertical usually puts the plane of focus where you expect it. Mostly relevant when using the tilt-the-rail and reset the standards to vertical trick.

[goodfood]

As you live in Hong Kong, the street is narrow and the buildings are very tall. The lens you in hand are 135mm and 180mm, those are standard lens for 4X5. You need a much wide angle lens such as 90mm, 75mm, or even 65mm with the coverage more than 4X5 film to have the movement. Also when you use those wide angle lens, you need a bag bellow on your camera. Check with your local camera club or dchome.net have some people using large format.

[austin granger]

I can't really offer any better advise than has already been given, but the first thing I thought of when I read your question was the Kodak book mentioned by rbultman. It was tremendously helpful to me in understanding movements.

[Emmanuel BIGLER]

From Ed Bray:

I am pretty sure that only the Camera Back needs to be vertical to ensure that buildings remain with the correct perspective,

From Struan Gray:
True. I was going to add

Well, this good ol' rule is valid only for quasi-symmetric lens formulae. I mean: any of our beloved LF lenses except telephotos. But usually nobody will tilt & shift a telephoto lens to capture a tall building anywhere in the world ;)

The strange effect was discovered a few years ago by a friend on the French LF forum http://www.galerie-photo.info, who was doing architecture shots with a digital SLR camera and one of those wide-angle retrofocus tilt+shit lenses available for small format cameras. He discovered that even if the sensor = the back of the camera is kept perfectly parallel to the vertical lines you want to see vertically projected in the images (e.g. the rectangular windows and edges of a rectangular building), you might get into some kind of trapezoid distorsion when you tilt the lens axis up.

In this situation the image is not sharp since it does not comply with Scheimpflug's rule, the "lens panel" not being parallel to the building and sensor plane.

The reason is extremely subtle and directly related to the fact that for those retrofocus lenses, the pupils are not located in the nodal planes (= principal planes). The effect hardly ever showed-up before, simply because there are no retrofocus lenses available for the large format (with the exception of an improbable LF Flektogon aus Jena, an absolute rarity once mentioned here).

The effect will exist even in a distorsion-free optical system, for example a perfect lens element like the "theoretical thin aplanatic lens element" available in the Oslo lens design software. Simply add a diaphragm in front of the element and look for a spot-diagram of a grid-shaped extended object. Keep the object and image planes parallel with respect to each other, but tilt the "lens plane": the image will be of course blurred but also strangely distorted. The effect naturally vanishes if the diaphragm is located just at the centre of the perfect thin lens element. This also applies for all quasi-symmetrical lenses like all wide-angle and standard LF lenses for which the pupils are located very close to the nodal points.

I have made a short summary (in French) of this effect in this article
http://www.galerie-photo.com/decentrement-bascules-scheimpflug-petit-moyen-format.html

Simply look at figures 15 and 16, at the end or the article, appendix #1, showing Oslo-edu simulations of the image of a square grid through a theoretical 'perfect thin aplanatic lens element'.
To perform the simulation yourself, you can dowload the Oslo command files from here
file "retrofoc-simul-oslo.zip" to be downloaded from here http://cjoint.com/?CJbjVn1ZbET
Once loaded, plot the optical system itself (lens->drawing->system) and use from the main menu the "Source" feature and select "pixelated object" choose a grid and click OK.
With command file retrofoc5.len available from the zip collection juste pointed above, you'll see the distorsion in the image although both object planes and image planes are parallel.

The only difference between figure 15 and figure 16 are that in figure 16, the diaphragm is located at some distance ahead of the lens. In both figures the object plane and the image plane are perfectly parallel. According to the good ol' rule of LF photography, in both cases no distorsion should occur. Yes of course in figure 16 the image is blurred since the set-up does not obey Scheimpflug's rule. But in fact distorsion appears on figure 16, where the diaphragm is not located at the nodal point of the thin lens element: therefore we get an optical system which is like a retrofocus or a telephoto; the entrance pupil simply being the diaphragm itself, is not located at the nodal point. This simulates what happens in a thick compound lens like a wide-angle retrofocus lens. We already knew that panoramic stitching requires to rotate around the entrance pupil and nowhere else, this is another manifestation of special effects in perspective and projection for defocused images that happen in a lens for which the pupils are not located in the principal or nodal planes.

As a conclusion, yes Ed and Struan you are right ;) ...

... since here we do not use retrofocus wide angle lenses (Vade Retro, Satanas), and that all 35 mm photography, be it film or digital, is strictly banned here!

[Struan Gray]

As a conclusion, yes Ed and Struan you are right ;) ...

Thank goodness for that.

Part of me is groping for a vaguely-remembered exercise in optics design in which the aperture is moved from behind a single meniscus to in-front. In this case, the lens itself is asymmetric, and the movement of the aperture trades various aberrations such as field curvature and distortion for others like coma and astigmatism. I can't help wondering is something similar is going on here.

In any case, one of the joys of movements on LF cameras with normal LF lenses is that you can separate the effects of rise and tilt. When I'm taking a picture I tend to work through three stages to get the image on the ground glass how I want it. First framing, then geometry, then plane of focus. There is some feedback and adjustment between the three aspects, for example, adjusting the geometry through movements can make for small changes to the framing, but I find it useful to treat them pseudo-individually in series.

[rbultman]

I stumbled across this video this morning which explains movements.

https://www.youtube.com/watch?v=0JU-eHpk97Y

The video is short. It explains each movement and then the effect of each movement in the ground glass.

He also mentions a PDF that describes camera movements but you have to send him an email in order to get it.

Regards,
Rob

[Emmanuel BIGLER]

From Struan:
In this case, the lens itself is asymmetric, and the movement of the aperture trades various aberrations such as ... I can't help wondering is something similar is going on here.

Well, the strange distorsion that you can simulate with Oslo exists even when computed with a theoretical perfect thin aplanatic lens, something which is only a mathematical entity that takes bunches of rays on input in a simulation and dispatches them on output without any aberrations.
Of course in a real lens, the exact location of the iris has a strong influence on several aberrations. The Wollaston meniscus will exhibit its best image quality only of the iris is located at the proper distance in front of the lens element. Something totally counter-intuitive.
And if we go back to the end of the XIXst century, the Old Masters have recognised very early the advantages of quasi-symmetric designs that we still use today with the view camera ; types of lenses but that cannot be used with reflex cameras for short focal lengths.
In such a quasi-symmetrical design, the iris is located very close to the centre of symmetry of the glass assembly; the pupils are located very close to the nodal points; hence fans of rays entering throught the entrance pupil will on exit exhibit a fan of rays with the same angles (property of nodal points); and the whole lens will be free from the very special effect mentioned above: you just need to keep your film plane parallel to the building and you are allowed a somewhat loose tolerance on tilting the lens itself.
So let's enjoy our symmetrical designs, and let retrofocus lenses live with their "built-in" distorsion when used to make an image between parallel planes but with an optical axis slanted and a blurred image ;)

I apologise for this digression that could appear as complex and very confusing, but as soon as we accept what happens in our quasi-symmetical lenses, as far as perspective rendition is concerned, we can forget about the glass and consider only what happens in a pinhole camera, i.e. central perspective like discovered in Renaissance times.
We just have to consider the entrance pupil of our lens, this is our equivalent pinhole defining alignments in object space, and imagine that exit rays come out of the exit pupil like if we had the rear part of a pinhole camera with its pinhole located at the centre of the exit pupil. So in a sense our view camera, as far as the geometry of the image projection is concerned, is like a pinhole camera, for which the entrance and the exit would be separated by some distance that has no influence on what is projected, i;e. exactly what the artists have been drawing since the XV-st century.

So basically
- setting the lens position in object space and setting the film back defines the kind of projection we would have in a pinhole camera, therefore commands perspective rendition but does not guarantee any sharpness;
- setting the lens plane in position and tilt commands image sharpness, has a very weak influence on the projected geometry if the lesn is quasi-symmetrical.
Sharpness is something that we do not care for in pinhole photography.

The problem is that different view camera controls inter-mix the 2 effects at the same time.
For example Saint Ansel's tilting of the back to take a picture of the Mittens in Monument Valley solves the problem of image sharpness for the foreground, but brings-in some trapezoid distorsion. This is hard to notice in the image; nobody cares.
In order to keep the film plane vertical you would have to tilt the lens panel in order to get the foreground sharp (according to Scheimpflug) and eventually you would get short of image circle ...

In architecture shots you can proceed in a very counter-intuitive way. Instead of framing the composition like with a small hand-held camera, and afterwards tilting to get vertical lines projected parallel on the ground glass, you could first set the camera horizontal and the back vertical, and only after that, define your image frame by moving the front standard or the rear standard vertically. This might not be practical if you are too close to very tall building.
Then one should remember that what the camera sees is defined by the position of its entrance pupil: if the image is not good in terms of perspective rendition, may be you have to move the camera and experiment another view point ... but for a 100-story building, carrying a ladder and climbing on it is not be very helpful to get a better perspective. And tilting the camera in all directions will not change the position of the entrance pupil beyond some very minor adjustments.

[Struan Gray]

Thanks Emmanuel.

dikaiosune01: in case this has you worried and confused, here are my three main ways of proceeding:

1) in Scotland, where I am often taking landscapes where everything is far away. I just point the camera at the scene, focus, and shoot.

2) mostly in Scotland, if I'm taking pictures of vegetation patterns on the ground in front of me: I just point the camera, focus roughly, lean the lensboard forward to lay the plane of focus along the ground, and shoot.

3) in my local woodland: I keep the camera horizontal. Frame with shift and rise/fall. Focus roughly. Place the plane of focus with tilt and swing. Shoot.

The last case corresponds to what an architectural photographer would do: it leads to a photograph which looks more like a classical painting or architectural drawing, in which the perspective follows academic rules. I find it works well for woodland and tree photographs too.

[Emmanuel BIGLER]

Thanks, Struan for this useful and efficient summary. I totally agree with you.
Yes, we do not need to take all our LF pictures like if we wanted to properly record the frontage of a classical building in compliance with classical rules.
Like Eugène Atget did in his famous views of the old Paris : he preferred to be short of image circle, but would never allow any slanted verticals.
http://www.gingkopress.com/i/atget-paris_i3.jpg

Regarding Struan's point #2, the advantage of tilting the lens panel if you have axis tilts, is that once focused this tilt (when applied subsequently) only slightly defocus the image.
But for a landscape where you do not care for perfectly corrected projection, you can also tilt the rear standard back and refocus.

In old Europe, sometimes, even the most scrupulous LF photographer (who would always insist on perfectly levelling the camera and would feel dishonoured to let any kid of trapezoid distorsion in his images), will have hard times to decide for the proper camera movements and precise corrections to apply.
For example in Champagne, France, the old city of Troyes with its magnificent half-timbered houses has some challenging subjects to offer ;) (http://cjoint.com/?CIqt5X6NvIS)
And in Alsace, in Strasbourg, the Saint Guillaume Church with its slanted frontage wall is waiting to challenge your skills! (http://upload.wikimedia.org/wikipedia/commons/thumb/d/dc/Absolute_Eglise_St_Guillaume_01.jpg/280px-Absolute_Eglise_St_Guillaume_01.jpg)