Strange "distortion" using the Cron 28

[colorflow]

Advertisement (gone after registration)

Jack:

Really appreciate it. Never thought that rectilinear correction would increase keystoning. One final question. Is there a rule of thumb on the relationship between the amount of upward tilt and keystoning based on focal lengths?

 

Thanks again,

Alan

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[colorflow]

... Now if he kept it perfectly level on all four corners lets say than most likely he would have to much foreground. .

 

Guy, good point, that's probably the reason I unconsciously tilted it up. It is not really critical, just curious. At first I thought it was barrel distortion...

 

Thanks,

Alan

[AlanG]

Alan G. - I am just surprised that a very small angling could cause such apparent tilt/shift effect on such distant objects. Thought it would take more angling to produce such results, and generally on closer buildings. Yet the closer building on the right show no such effect? Thanks for your post work

 

Alan

 

The distance of the objects is irrelevant. If you angle up one degree, all objects will be converging one degree. The same is true if your camera is tilted just one degree. It is surprising how we are capable of noticing slight crookedness in photos. With architecture you usually don't want to be off more than about .2 degrees before it becomes noticeable to critical eyes.

 

This effect is the same with all lenses irrespective of focal length and is simply a result of "perspective." That is your vantage point in relation to the subject. All a lens does is show more or less in a frame and in no way does it alter perspective. A perspective control lens simply lets one shoot at a different angle in relation to the subject and then by shifting it crops in to a given part of that frame. You could get the same reult by shooting with a wider lens and cropping. When you shoot a distant object with a longer lens, you are cropped in so tightly that with a 1 degree angling up, there is very little difference in size between the width of the top and bottom of a house for instance. But it is still there. It is just less obvious than when you see a large size change over an image.

 

As I said in my first post, you can angle up and then tilt the camera enough to make one side of a building parallel to the edge of the image frame. There is no limit to how much you can do this. But this compounds the tilting problem on the other side. So if one side is correct, just use free transform to pull the one corner straight.

 

Of course one can use a combination of rotation and perspective correction to fix this in software. But I find it is easiest to use a free trnasform and pull the necessary corners until the image is aligned to my reference lines. (I know that Photoshop works well for this but I don't use that software - still using Picture Publisher 10, as it is easier and faster for my work.) I use DXO for most of my conversions and it has various geometry tools that make this type of correction a snap as part of the raw conversion process thus usually eliminating the need to do it with Picture Pubisher too.

[colorflow]

The distance of the objects is irrelevant. If you angle up one degree, all objects will be converging one degree. .....

.

 

Thanks Alan, I started realizing that as I read all the comments in this thread. But as Jack pointed out above, isn't keystoning much less noticeable with longer lenses?

 

Alan

[ho_co]

Very clearly put, Jack & Alan!

 

One other thing to mention:

 

This perspective exaggeration is the same phenomenon we see when we stand on railroad tracks and see them converge at infinity.

 

That is, things further away appear smaller; since the tops of buildings are further from us than the bottoms (being at the other end of the hypotenuse of the right triangle), they are imaged smaller. Slightly different geometry but the same effect with the tracks: Where we are standing, it's clear that they are 4 ft 8.5 in apart; but at a greater distance that same separation subtends a smaller angle of our vision, and the tracks seem to converge.

 

I once read a user's condemnation of a particular digicam, saying that the lens distorted badly, making the tops of buildings seem to converge when he shot up from ground level. (Sigh.)

 

--HC

[ho_co]

But as Jack pointed out above, isn't keystoning much less noticeable with longer lenses?

Yes, because you're seeing so much less with the long lens.

 

A 600 mm lens has a field of view of 4.13° on 35mm film.

A 24 mm lens has a field of view of 84.06° on 35mm film.

 

How much can you tell about an object when you're only looking at the part of it that fits within a 4-degree circle?

 

A building 100 meters away will appear exactly the same way with both lenses if the camera isn't moved between the two pictures; the only difference is the amount of the frame the building occupies.

 

With the wide lens you can see enough of the subject to tell that the top seems narrower than the bottom, but with the tele the field of view is so narrow that you don't notice it.

 

Take a house from your image and blow it up to 1000% and you simply can't see enough of it (on normal computer screens) to be aware of the perspective.

 

This is in no way a 'distortion,' but a proper rendering of perspective by the lens.

 

Since we aren't as aware of it in the world as we are in a two-dimensional representation, it is often referred to as "perspective exaggeration" when it seems to get out of control. Perspective is 'exaggerated' only in the sense that we are looking at a much wider part of the world than we normally see.

 

And that brings us back to the first paragraph of this post.

 

I hope that helps fill in just a bit.

 

By the way: Take a look at the image you posted. All the horizontal sections (driveways, walkways, piers, whatever they are) that front the houses run downhill to the right. Level them (by rotating as others have done) and everything looks more or less right. If you want to, touch that up by making the building walls parallel and you'll be happy.

 

--HC

[Jack_Flesher]

Advertisement (gone after registration)

Jack:

Really appreciate it. Never thought that rectilinear correction would increase keystoning. One final question. Is there a rule of thumb on the relationship between the amount of upward tilt and keystoning based on focal lengths?

 

Thanks again,

Alan

 

There obviously is a direct mathematical relationship, but it is non-linear and really isn't relevant in practice: you already know that it is of concern only when the camera is other than perfectly level. It is why cameras that can adjust for it have grid lines on their focusing surface. Obviously cameras like your M do not have that ability, so like with many other issues with the M, you use experience to compensate. At the end of the day if you are shooting an Myou and have to point it up at a 45 degree angle to get everything you want in the shot, then that's what you do -- convergence or not... Then call it an artistic interpretation

 

The short version is, longer lenses from farther away will always present less of a problem since the camera's -- and hence the sensor's -- angles of tilt off of level are reduced.

 

Cheers,

 

Jack

[colorflow]

Yes, because you're seeing so much less with the long lens.

 

 

A building 100 meters away will appear exactly the same way with both lenses if the camera isn't moved between the two pictures; the only difference is the amount of the frame the building occupies.

 

With the wide lens you can see enough of the subject to tell that the top seems narrower than the bottom, but with the tele the field of view is so narrow that you don't notice it.

 

 

--HC

 

What you said is obviously true if perspective is the only thing that's in play here. However, seems to me the optics of a wider lens exaggerates perspective more than longer ones. I suspect if I had used a 90mm or even a 50mm with the same camera position (I would still see the entire buildings) but less or no tilt. Probably the rectilinear effect Jack was talking about.

 

Alan

[ho_co]

Alan--

Jack, Alan and I are all saying the same thing.

 

Perspective is determined simply by viewpoint, i.e. shooting from one particular distance and angle will give you exactly the same perspective no matter what lens you use.

 

When you use a wide lens, blow up the image more and stand closer to it. If you duplicate to your eye what your camera saw, you'll see the same thing as well.

 

In other words, if you shot with a lens of 110° field of view, stand close enough to the image so that it subtends 110° to your eye and you will not see any 'distortion' or 'perspective elongation' or whatever you want to call it.

 

If your shooting lens had a field of view of 6°, then back up enough from the image so that it subtends only 6° of your vision and you will not see any 'distortion' or 'perspective exaggeration' or whatever you want to call it. (You can also do that with the first image: Just crop from that image with 110° diagonal a portion representing only 6°, as I said previously.)

 

It's when you view images at different magnifications from the same distance, keeping your field of view the same, that you see what you think is a problem.

 

Jack has said that twice and Alan and I at least once. You asked

But as Jack pointed out above, isn't keystoning much less noticeable with longer lenses?

and we have all answered 'yes.' It is less noticeable, but it is the same amount and isn't reduced at all no matter what focal length you use. It is determined solely by angle and distance.

 

(FWIW, it was the differences in distance from Saddam Hussein in various photos that started all the rumors about 'body doubles.' Well, the perspective differences plus a slow news season. You'll notice how quickly the reports of body doubles disappeared once there was something more interesting to report, like an invasion. )

 

I suspect if I had used a 90mm or even a 50mm with the same camera position (I would still see the entire buildings) but less or no tilt.

The apparently vacant house on the right is already almost too large for the frame. I don't think you could have got it in with a 50mm, and certainly not with a 90.

but less or no tilt.

Again, that's what this thread is about. If you don't tilt the camera, you don't stretch perspective.

 

--HC

[Guest guy_mancuso]

Guy, good point, that's probably the reason I unconsciously tilted it up. It is not really critical, just curious. At first I thought it was barrel distortion...

 

Thanks,

Alan

 

 

Great answers guys, let me go back and say this is a real common issue. Folks tend to point up all the time to get the correct composition. If you stand next to a 30 ft building and your 6 ft obviously your going to point up , now if you got on a scissor lift and had the camera perfectly level and raised the lift to get the compositon you like, my bet would be right in the middle of the building at 15 ft. This would give you absolute straight lines, well here is the issue your 6ft and the correct spot is 15 ft so in effect you 9 ft away on the ground from a perfectly composed and straight image. So naturally you would point up. Again you have two choices to get this image a scissor lift or a shift lens and at 6ft perfectly level you just shift up. Okay not always practical to do either one. So the only third choice you have is fix it in PS.

 

Now having said all that . This is a perfect example of what i just said . In this case. I had my DMR tethered to a labtop. Now to avoid movement on the scissor lift I placed the camera on the lift perfectly level . Than me on the ground with laptop controlled the camera and lifted the lift up to a height around 15 ft until it was the composition i wanted and shooting tethered to see the final image. So here basically i took the lift approach but i also gained a nice high viewpoint and there is no PS work here

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[colorflow]

Alan--

Jack, Alan and I are all saying the same thing.

 

....

Jack has said that twice and Alan and I at least once. You asked

and we have all answered 'yes.' It is less noticeable, but it is the same amount and isn't reduced at all no matter what focal length you use. It is determined solely by angle and distance.

 

(

--HC

 

HC:

 

I think we've beat this perspective question to death. I understand and agree with everything you and Jack said. Theoretically keystoning is determined solely by angle and distance, but barrel distortion, rectilinear design, and perhaps other focal lens related design characteristics also play a part. What I was conjecturing was not about perspective, but about if the differences in optical designs of wide verses longer lenses affect the appearance of keystoning. Of course I won't see the larger houses on the right, I was talking about the appearance of the distant houses across the water. Please don't rehash this anymore. I will test this myself with a slightly tilted 28, 50, 90 if I have time..

 

Alan

[AlanG]

Yes keystoning is much less noticeable with longer lenses. That is because the narrow field of view means there will be less difference in subject size (after tilting) from top to bottom, between the middle of the frame and the edges. Whereas with wider lenses the difference in size can be greater.

 

Well you asked for a formula so here it is. (I studied this in college but it is taxing my memory to its limit at this point.)

 

1/f = 1/o + 1/u Where f= focal length, o=object distance and u=image distance. (The object distance is the distance between the rear nodal point and the subject. The image distance is the distance between the rear nodal point and the sensor.) So as the object distance increases by angling the camera, the image distance decreases and you have a smaller image on that part of the sensor. This is why lines converge and this is probably the most basic optical principal in photography. It is why the "Schiempflug" rule works when you move a lens so that it is no longer parallel to the sensor.

 

To picture this you have to understand the concept in perspective of a "vanishing point." If you shoot a grid perpendicular to the camera, then all lines will be parallel and there will be no vanishing point. If you angle the camera upwards, all lines will aim at a specific point somewhere above the grid. (Very far above if the camera isn't tilted very much.)

 

So these lines stop being depicted as parallel and now converge. The important thing to understand is that the angle of their convergences is the greatest at the edge of the frame and least at the middle of the frame. So if you tilt your camera upwards 1 degree, it isn't as if every vertical line on the left side of center will tilt one degree to the right and every line on the right of center will tilt one degree to the left. No, only the lines at the very edge of the frame will show that one degree of tilt. The rest of the lines will show somewhat less tilt until you get to the line in the center which will be perpendicular. Whereas if you rotate a camera on axis one degree below the horizon, everything will be tilted (rotated)exactly one degree.

 

Now I'm not sure if I cleared things up or added to any confusion.

[colorflow]

... The important thing to understand is that the angle of their convergences is the greatest at the edge of the frame and least at the middle of the frame. So if you tilt your camera upwards 1 degree, it isn't as if every vertical line on the left side of center will tilt one degree to the right and every line on the right of center will tilt one degree to the left. No, only the lines at the very edge of the frame will show that one degree of tilt. .

 

 

Thanks Alan, and that is why I was surprised to see this even with buildings across the water closer to the center.

 

Alan

[AlanG]

Thanks Alan, and that is why I was surprised to see this even with buildings across the water closer to the center.

 

Alan

 

You also had your camera rotated on axis a little. Basically, you tilted up a little and then rotated your camera until the building on the right side looked vertical. This made everything else tilt even more than either situation would do on its own.

 

Also, I haven't done any calculations or measurements to support this, but logically, it seems to me that if a lens produces a field of view of 90 degrees horizontally, that is the only time where there will be a 1 to 1 relationship at the edge of the frame between the angle that the camera is tilted upwards and the angle of keystoning at the edges. e.g. you tilt up 10 degrees, objects at the edges will tilt inwards 10 degrees. So with even wider lenses, there will be more and with longer lenses less. Pretty much the effect you've been observing.

 

If I get time, maybe I'll break out a protractor and check this.

[colorflow]

.... So with even wider lenses, there will be more and with longer lenses less. Pretty much the effect you've been observing.

 

 

Yes, that's what I've been trying to get to - wider lenses produce more keystoning towards the edges than longer lenses. I'm trying to do the geometry which is a little hairy to keep all the angles straight...

 

Alan

[ho_co]

Yes, that's what I've been trying to get to - wider lenses produce more keystoning towards the edges than longer lenses. I'm trying to do the geometry which is a little hairy to keep all the angles straight...

 

Alan--wider lenses don't produce more keystoning than longer lenses. It is just that the keystoning is more noticeably with wides.

 

Please try to understand what everyone has already said: This is specifically and only an issue of perspective and has nothing to do with distortion. (Yes, all lenses have some distortion but what this thread is about is that you rotated and tilted the camera and continue to say that perspective is related to distortion.)

 

I will test this myself with a slightly tilted 28, 50, 90 if I have time.

 

Please do that. And read up on optics or read what others have posted here.

 

And get it thru your head that focal length doesn't cause variances in keystoning, even though keystoning is more evident in wide angle shots.

 

--HC

[AlanG]

 

And get it thru your head that focal length doesn't cause variances in keystoning, even though keystoning is more evident in wide angle shots.

 

--HC

 

I think it does because the different areas of the image hit the sensor from a more extreme angle with wide angle lenses than they do with longer lenses. So a crop from a wide may match a tele shot, but the edges of the wide angle shot will have greater keystoning because the angle of keystoning varies from center (perpendicular) to edge (most extreme.)

 

Just consider that if you have a 2000 mm lens and photograph at a 1 degree angle up, the effect will be almost impossible to see. Yet if you have a 15mm lens and tilt up one degree, the objects near the edges will clearly converge. But the objects in the very center will not show noticeable convergence.

 

The fact that the wide angle lens has a greater angle of view does allow for this even though the perspective is the same. It's the vanishing point aspect of perspective where all converging lines must meet at the same point.

[robsteve]

 

I only spent about 5 minutes total on this and am not what I would consider "done" with it. If I have time, I may try and spiff it further tomorrow.

 

Cheers

 

Jack

 

Jack:

 

I think the color on the original version was probably more true for an east coast early or late sun. Probably what he would have got if he shot it on slide film.

 

Robert

[Advertisement]

Hallo,

Look here: Leica M

[colorflow]

I think it does because the different areas of the image hit the sensor from a more extreme angle with wide angle lenses than they do with longer lenses. So a crop from a wide may match a tele shot, but the edges of the wide angle shot will have greater keystoning because the angle of keystoning varies from center (perpendicular) to edge (most extreme.)

 

Just consider that if you have a 2000 mm lens and photograph at a 1 degree angle up, the effect will be almost impossible to see. Yet if you have a 15mm lens and tilt up one degree, the objects near the edges will clearly converge. But the objects in the very center will not show noticeable convergence.

 

The fact that the wide angle lens has a greater angle of view does allow for this even though the perspective is the same. It's the vanishing point aspect of perspective where all converging lines must meet at the same point.

 

Alan, I just laid out the geometry of the optics and clearly the wider the angle the more keystoning at the edges. I am struggling to work out the quantitative relationship between the tilt and keystoning angle.

 

Thanks for understanding my question and initial gut feel conjecture.

 

Alan

[colorflow]

Jack:

 

I think the color on the original version was probably more true for an east coast early or late sun. Probably what he would have got if he shot it on slide film.

 

Robert

 

Thanks Rob. i posted this not for the subject or composition, just curious about the keystoning which we have beaten to death. As for color, I think you are right but I am a poor judge of that since I can't tell blue from green.

 

Alan