Do Tilt shift lenses correct perspective or distort

A forum to ask questions, post setups, and generally discuss anything having to do with photomacrography and photomicroscopy.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

austrokiwi1
Posts: 350
Joined: Sun Sep 14, 2014 10:53 am

Post by austrokiwi1 »

Whether you call it distortion or not, the shifted image is different. In the shifted image you can see the edge of the coin. In the normal photo, the edges are invisible

No the edges are not visible in the shifted image. I have used shift in this manner for around three years and have never once had the edges of a coin visible in the resultant shot.

Here is a real life example. Camera Sony A7Rii( full frame). Lens is a Mamiya 645 120mm F 4 @ 5.6. Subject a 62(/3)mm diameter medal. Lens is mounted via a mirex Tilt shift adapter with 15mm shift applied. The set up looks like this:

Image

The unedited( it has been reduced for posting here) picture follows, the black section on the right hand side of the frame is the edge of the back light . You will note the edges are not visible. Note the medal is very thick( only the rim is visible not the edge/sides) so if you were correct the edge/side would be easily visible. My understanding based on my understanding and experience experience is that with shift the only way you would see the medal coin edge would be if you used a pericentric lens( I have not used a pericentric lens):
As you can see the edges are not visible ( in a real life example)


Image


I would also note there is no a obvious distortion.

This concept of shifting causing distortion versus the un-shifted image seems to be difficult for folks to understand. I have gone at this from multiple angles on both forums. For sure I am not using some arcane definition of distortion. Maybe too many words, not enough pictures...

NO you have not given any explanation, all you appear to have done is repeated the same assertion. Each time I have asked you to explian using a diagram you come back to the same written explaination. Accordingly it is a little exaserperating to read this from you:
Maybe too many words, not enough pictures


your assertion is at substantive variance to my experience and reading. Please I am now asking for the third time; using your diagram show me where and how the distortion is occurring and as previously requested please also supply some links or names of references. I repeat what you are saying is at substantive variance to the references I have read and used for Tilt shift photography

It would help if you could tell me what type of distortion you are referring to then I can going and research it my self. Here I make an assertion: shift, as I have used it and understand it , is not a form of distortion, it is a means to change perspective and/or avoid distortion(such as that caused by paralex)
Still learning,
Cameras' Sony A7rII, OLympus OMD-EM10II
Macro lenses: Printing nikkor 105mm, Sony FE 90mm F2.8 Macro G, Schneider Kreuznach Makro Iris 50mm , 2.8, Schnieder Kreuznach APO Componon HM 40mm F2.8 , Mamiya 645 120mm F4 Macro ( used with mirex tilt shift adapter), Olympus 135mm 4.5 bellows lens, Oly 80mm bellows lens, Olympus 60mm F2.8

ChrisR
Site Admin
Posts: 8668
Joined: Sat Mar 14, 2009 3:58 am
Location: Near London, UK

Post by ChrisR »

My 2 pence - if you project a 3D object onto a 2D plane, you have a perspective. Either both are, or neither is, "distorted".
It's nonsensical to assert that one is distorted and the other isn't.
Both are rectilinear - straight lines are straight. Neither is "natural".

To me natural would need a spherical image plane - and brain to interpret it.
But you need a brain to interpret any 2D projection of a 3D object.

We can argue the toss over what constitutes "distortion" .
We can have two views of a high window, one taken with a camera lens held horizontal and the other with the camera tilted up. I can accept that one has "distorted" the shape of the window, but not that one image is a distorted version of the other image.

For a 2D subject the situation is clearer - the "shifted lens" approach removes the consideration of the point of view of the observer, whether you call that a "distorting" influence or not.
Chris R

Troels
Posts: 600
Joined: Mon Feb 15, 2016 11:06 am
Location: Denmark, Engesvang
Contact:

Post by Troels »

Allow me to add an observation.

Looking at the 3 drawings in austrokiwi's post it is obvious, that the shifted immage is completely equvalent to a normal picture with a bigger sensor, where you just crop one side of the picture away. Besides the cropping, there is no difference (if the optics are perfect).

The tilted image is a completely different story which I am not able to analyze further.

Troels
Troels Holm, biologist (retired), environmentalist, amateur photographer.
Visit my Flickr albums

ChrisR
Site Admin
Posts: 8668
Joined: Sat Mar 14, 2009 3:58 am
Location: Near London, UK

Post by ChrisR »

Troels - exactly so. When I used to do some architectural photography I routinely used a 20mm lens intending to crop the lower part of the image.
It gives very much the same result as a less wide Perspective Control lens, shifted.

Austrokiwi - I think you do see the edge of the coin, it's just so oblique a view that it's not large enough to be visible.

In the same way that you see the inside of the upper liner of a window high on a building.
If the coin were a can of beans you'd expect to see the side of the can..

If your picture is about 1:3 (72mm on the short side of a 24 x 36)
the lens - subject distance is 480mm.
How far have you moved the coin off axis?
If you've moved the coin edge to 48mm from the axis, (total 48 +63/2 mm) you'd expect to see the edge as 1/10th (first approximation, saying tan = sin..) of the coin's thickness, so perhaps 0.5 mm wide.
That would be about 1/3rd of the coin's outer raised ring width.

We aren't seeing that. Why? Have you tilted the coin?
Chris R

austrokiwi1
Posts: 350
Joined: Sun Sep 14, 2014 10:53 am

Post by austrokiwi1 »

Chris R ==> I had switched out lenses( the Oly 135/4.5 also makes a good shift lens). In switching back I did my best to set up the rig at the same magnification. Working distance, front of the lens housing( not the front element which is recessed) to the medal was just on 400mm in all likely hood probably close to the 480mm you estimated i taking into account the recessed front element. With 15mm shift ( I re-leveled the medal and the camera to make sure sensor and medal were parallel) 30mm of the medal remained under the lens (thats the lens body/housing . 33mm of the medal is out from under the lens( medal is 63mm in diameter. therefor the coin is not that far off axis that's why I haven't seen the edge of coins. I take you point some of the edge might be visible but, assuming I have understood you correctly it will not be as much as a 1/10th.
with the Oly 135mm I can get a greater Offset
Still learning,
Cameras' Sony A7rII, OLympus OMD-EM10II
Macro lenses: Printing nikkor 105mm, Sony FE 90mm F2.8 Macro G, Schneider Kreuznach Makro Iris 50mm , 2.8, Schnieder Kreuznach APO Componon HM 40mm F2.8 , Mamiya 645 120mm F4 Macro ( used with mirex tilt shift adapter), Olympus 135mm 4.5 bellows lens, Oly 80mm bellows lens, Olympus 60mm F2.8

austrokiwi1
Posts: 350
Joined: Sun Sep 14, 2014 10:53 am

Post by austrokiwi1 »

Chris R and Troels. Thanks for your comments. As I understand what you have said ( I understand it applies to all shift lenses) the lens has a field of view and an image circle that is much larger than the sensor. Shift can be understood as moving the camera to the part of the image circle that is to be sampled(. I attempted this drawing( not to scale)... does it illustrate what you are saying. It certainly matches my understanding of shift and my view that shift does not directly involve distortion ( lens faults are a separate issue)


Image
Still learning,
Cameras' Sony A7rII, OLympus OMD-EM10II
Macro lenses: Printing nikkor 105mm, Sony FE 90mm F2.8 Macro G, Schneider Kreuznach Makro Iris 50mm , 2.8, Schnieder Kreuznach APO Componon HM 40mm F2.8 , Mamiya 645 120mm F4 Macro ( used with mirex tilt shift adapter), Olympus 135mm 4.5 bellows lens, Oly 80mm bellows lens, Olympus 60mm F2.8

Troels
Posts: 600
Joined: Mon Feb 15, 2016 11:06 am
Location: Denmark, Engesvang
Contact:

Post by Troels »

Yes, that is exactly how I imagined the situation.

Troels
Troels Holm, biologist (retired), environmentalist, amateur photographer.
Visit my Flickr albums

ChrisR
Site Admin
Posts: 8668
Joined: Sat Mar 14, 2009 3:58 am
Location: Near London, UK

Post by ChrisR »

Yes :)

Ok with only 15mm shift (at the coin I assume, requiring about 5mm shift at the lens) and the distance to entrance pupil say 450mm, your edge (assuming 5mm thickness) would show about 5/(450/15) = 0.17mm, = lost in the bevel?

So, there's no distortion. The coin is close enough to a 2D subject that you have a very well behaved projection!
Chris R

ray_parkhurst
Posts: 3417
Joined: Sat Nov 20, 2010 10:40 am
Location: Santa Clara, CA, USA
Contact:

Post by ray_parkhurst »

ChrisR wrote:Yes :)

So, there's no distortion. The coin is close enough to a 2D subject that you have a very well behaved projection!
Well, sort of, but not strictly. Certainly any distortion of the shape or perspective of the coin is a second order effect and not visibly noticeable unless a direct comparison is made between the straight-on vs off-center view.
ChrisR wrote:To me natural would need a spherical image plane - and brain to interpret it.
Yes, to me as well, and this is my exact argument.

Here is what I wrote on the other forum:

"It's easier to understand this concept in terms of architectural photography. If I stand near the base of a building, and look at the building, the parallel lines of the building will appear to have parallax due to converging perspective. This is the natural look of the building from my vantage point at the base of the building. I can take a picture of the building and I will see exactly the same converging perspective. If I don't like this look, I can do one of two things to fix it:

1) I can go into photoshop and stretch the image in a trapezoidal fashion. Stretching the top of the image such that the top of the building has the same width as the bottom. This will make the vertical lines of the building look parallel. I am sure photoshop has canned alogorithms for doing this.

2) I can point the lens straight toward the building, and shift the lens or the sensor so that the building comes into view.

Both of these actions result in a similar (though not exactly the same) final image. Both actions cause the natural, converging perspective view of the building to be stretched such that the converging perspective is eliminated, and the parallel lines appear parallel. Stretching the converging perspective view is a distortion of the natural image of the building.

edited to add: as Rik pointed out in the other forum, I could also use a large format camera, and a wide angle lens, and take the picture pointed straight at the building. The building would then be present in only ~50% of the image, so I could crop the rest and end up with essentially a shifted image of the building."

austrokiwi1
Posts: 350
Joined: Sun Sep 14, 2014 10:53 am

Post by austrokiwi1 »

Ray for the fourth time (quoting the third time I asked):
I am obviously thinking in a different frame of reference( to what you a describing and what you r saying does not "click"). I repeat again a picture ( as they say its worth a thousand words And I don't want another thousand words I want understanding)
your assertion is at substantive variance to my experience and reading. Please I am now asking for the third time; using your diagram show me where and how the distortion is occurring and as previously requested please also supply some links or names of references. I repeat what you are saying is at substantive variance to the references I have read and used for Tilt shift photography

To make it clear this statement of your s makes absolutely no sense to me

"It's easier to understand this concept in terms of architectural photography. If I stand near the base of a building, and look at the building, the parallel lines of the building will appear to have parallax due to converging perspective. This is the natural look of the building from my vantage point at the base of the building. I can take a picture of the building and I will see exactly the same converging perspective. If I don't like this look, I can do one of two things to fix it:
I understand this as a correction of perspective not distorting the image. when you photograph the building tilting the sensor you get the converging lines.... because the subject is not parallel to the sensor. Using shift vertically ( I believe this is rise in large format photography terms ) I can get the top and bottom of the building in frame while ensuring the camera's sensor remains parallel to the building. If I was using a non-shifting lens I could achieve the same thing but some how getting the camera to a higher vantage point or moving far enough back so that the image circle covers the whole of the building while the sensor remains parallel with the building.. I see no Distortion in this process, I understand it as a change( correction) of perspective.


Photoshop and the like achieve the same by distorting the image but that is a completely different matter, and not applicable to shift.
Still learning,
Cameras' Sony A7rII, OLympus OMD-EM10II
Macro lenses: Printing nikkor 105mm, Sony FE 90mm F2.8 Macro G, Schneider Kreuznach Makro Iris 50mm , 2.8, Schnieder Kreuznach APO Componon HM 40mm F2.8 , Mamiya 645 120mm F4 Macro ( used with mirex tilt shift adapter), Olympus 135mm 4.5 bellows lens, Oly 80mm bellows lens, Olympus 60mm F2.8

ray_parkhurst
Posts: 3417
Joined: Sat Nov 20, 2010 10:40 am
Location: Santa Clara, CA, USA
Contact:

Post by ray_parkhurst »

Copied from the other forum:

OK, see below. I was hoping to actually take pictures rather than a diagram, but this will have to do.

As I have said, the concept is easier to understand with architecture, but I have made the diagram showing the coin situation.

I have represented the distortion of the image taken with shift (versus the natural perspective) as a lengthening of the coin, but please realize that there is more than lengthening going on between these two techniques.

Image

elf
Posts: 1416
Joined: Sun Nov 18, 2007 12:10 pm

Post by elf »

ray_parkhurst wrote: but please realize that there is more than lengthening going on between these two techniques.
What everyone is trying to tell you is this is not distortion. The images are different but that isn't how distortion is defined. Moving the focal plane isn't distorting the image, it's merely showing a different view of it. This argument reminds me of people saying a wide angle closeup of a person's face is distorted when in fact it's just a different perspective.

ChrisR
Site Admin
Posts: 8668
Joined: Sat Mar 14, 2009 3:58 am
Location: Near London, UK

Post by ChrisR »

"It's easier to understand this concept in terms of architectural photography. If I stand near the base of a building, and look at the building, the parallel lines of the building will appear to have parallax due to converging perspective. This is the natural look of the building from my vantage point at the base of the building. I can take a picture of the building and I will see exactly the same converging perspective. "

Those are where the disconnect lies.
There's nothing natural about holding a camera at some angle, chosen by brain to be pointing somewhere in the area of the subject, and accepting that 3D image projected onto a 2D plane. It's an artifice.
It's a convention, but no more conventional than holding the camera level.

Moving your eyes around the subject while looking at it, gives you different views. None of them is ever a 2D planar image.
If you ask a child what shape a square thing, painted high on a wall, actually is, he'll tell you it's square.
The only natural thing, is in the mind.

So, there's no distortion. The coin is close enough to a 2D subject that you have a very well behaved projection!
Well, sort of, but not strictly. Certainly any distortion of the shape or perspective of the coin is a second order effect and not visibly noticeable unless a direct comparison is made between the straight-on vs off-center view.
So, only the central pixel is the correct view, all the rest of the image is distorted?
It follows from that argument, that whenever you use a camera, you distort the view.
To me, saying that one photographic image is natural and the other is distorted, is wrong.

We have to be careful with "First order" and "Second order". I'm no student of such things, but there's a qualitative judgement there implying that one is acceptable and one is not.
I see no step in the perspective effect - to me it's a continuum.
But that's just my point of view :oops: .
Chris R

rjlittlefield
Site Admin
Posts: 23564
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Let me try a different tack...

Ray,

In your right-side configuration, the subject plane is rendered entirely in focus and is reproduced exactly in the image plane, except for change of scale and position. The image of any square is still a square.

But in your left-side configuration, the subject plane is rendered mostly out of focus and suffers everywhere from perspective foreshortening. Every square in the subject plane turns into some other shape in the image.

With that summary, it seems to me that the obvious labeling would be left side distorted, right side not -- exactly the opposite of your labeling.

I understand why the "natural view" is special -- because it pretty much matches what a human eye would see from the same viewpoint.

What I don't understand is why you persist in labeling that view as undistorted -- and more, the only undistorted view -- given the behavior of squares turning into something else.

Can you explain that aspect?

--Rik

ray_parkhurst
Posts: 3417
Joined: Sat Nov 20, 2010 10:40 am
Location: Santa Clara, CA, USA
Contact:

Post by ray_parkhurst »

elf wrote:What everyone is trying to tell you is this is not distortion
I know, but it is only because folks haven't quite "gotten it" yet.
ChrisR wrote:So, only the central pixel is the correct view, all the rest of the image is distorted?
Strictly speaking, yes, unless the image was taken with a "normal" lens, though this is a second order effect. Lenses in the ~50mm range (for FF) are considered "normal" because they produce the least image distortion vs what the eye would see from that vantage point. Wide angle or telephoto lenses cause distortion of the image vs the natural perspective. This is called "perspective distortion":

https://en.wikipedia.org/wiki/Perspective_distortion
https://en.wikipedia.org/wiki/Normal_lens
rjlittlefield wrote:... The image of any square is still a square
Well, I suppose a picture of a house is still a house, so by that same philosophy a picture of a square is still a square. But the apparent shape of a square, imaged at an angle, is a trapezoid. Take the image of the square at enough of an angle and it is a line, no longer recognizable as a square. Show that image to someone, and ask them what it is, and they will say it is a line. So is it really still a square?
rjlittlefield wrote:What I don't understand is why you persist in labeling that view as undistorted -- and more, the only undistorted view -- given the behavior of squares turning into something else.

Can you explain that aspect?

--Rik
The distinction is between the image being distorted vs the objects within that image appearing distorted.

In images produced by the "natural" vs "shifted" configurations above, the "natural" configuration would produce an undistorted (natural) image, with the coin appearing distorted within that image. The "shifted" configuration would produce a distorted image (as compared with the "natural" configuration) with the coin appearing undistorted within that image.

Post Reply Previous topicNext topic