When one straightens two (vertical or horizontal) lines the other lines often are not straighten

DxO ViewPoint Windows topics
41

Focal plane is parallel to the sensor plane. The distance between them is the perpendicular distance between them. When I focus on a wall with the camera all the wall is in focus. So it is with this building when using a levelled camera. I just wonder if I also get that strange ‘feeling’ as with the corrected image.

I assume raised is shifted. I think the trick with this is using the possibilities of an oversized image circle of the lens.
But I don’t have that strange feeling as with this corrected image.

George

42

Indeed - shifted upwards.

Of course. The image of Bourges cathedral was taken with a Schneider-Kreuznach 72mm f/5.6 Super-Angulon, which has an image circle of 229mm, which allows a maximum shift of 44mm.

That’s because the example image is quite an extreme wide angle

43

That are know figures? I never used a tilt/shift lens so I’ve no idea.

That still doesn’t explain it.
Just a guess. One can straighten the image but the direction of the camera is still there: the middle of the image when no cropping was used. My brain says the top of the building, which is further away, must have a smaller magnification as the bottom, which is closer. By leveling this is undone. Maybe my brains can’t deal with that and just says: it’s further away so it must be bigger.
Just a thought. That’s why I would like to see an image taken with a leveled camera.

George

44

I have just shown you two.

45

Right – and if it is not as ‘expected’, but all leveled (all vertical lines parallel…), it may look wider at the top because of your assumption.
A way to handle this is to let the vertical lines slightly converge until your perception tells you “right”.

2 Likes
46

@Joanna
Post 40 :wink:

George

47

I first aligned the right side with the middle using the Force Parallel tool then exported as a TIFF. Then I used the FP tool on the new file to align the left side with the other two. The “wider at the top” and “left side leaning in” illusions are still present but all three lines are now vertical.

Image00001
Image000011920×1464 247 KB

48

Very sorry for the hesitation!

OK. The downloaded file that you linked to is a PNG - something that PhotoLab cannot handle. So I converted it to JPG, but this is definitely not the original file.

Ah, sorry for the circumstances. I must have mixed up the picture.

https://forum.dxo.com/uploads/default/original/3X/8/d/8d4afc07d4f5515dc70bdbbfb1776ed200caf07f.jpeg

Showing the grid shows that all the buildings are now properly vertical, although, due to the perspective, they look like they are wider at the top than the bottom.>Does that help?

Yes, but they are not. They only look like it. But why? What lines / proportions cause that?

Now the main building looks more “normal” but, of course, the buildings at the edge will still look a bit canted in.

But these lines still are quite skew:
https://i.imgur.com/O3wMVve.jpeg

So the next action:

This is to be expected and can only be corrected by forcing parts of the image using the ReShape tool like this…

seems to straighten the lines very much:

https://i.imgur.com/58dse4k.jpeg

Does that help?

Yes, very much, very many thanks! The buildings still look quite ugly. These (kind of) images obviously cannot be saved, so I can just throw them away. So that means I simply tilted the camera way too much.

You can always try the magic wand to automatically try and do the right thing…
… but the resulting changes mean that you will be missing an awful lot of sky…
And it still isn’t quite right.
https://i.imgur.com/J4S0GPz.jpeg

I guess the automatic works, especially with such strong distortions, only for a rather imprecise adjustment. May be it is only usable for slight adaptions. If at all.

Thank you very much for trying and showing me!

When we look up, we see converging lines, but we know that they should be parallel. Our brain mixes the two ingredients to make things look like expected…within certain bounds.

But our “brain” knows as well that we look at an image, I would think, so it knows there is nothing to mix, it just is an image. And the same image above taken with a vertical image plane would not cause such (odd) view. We or I would think it is like it should be. And why do we see other images “correctly”? And movies. And feel them as they are in reality. According to that explanation, shouldn’t we see all the images wrong, or at least many? Like those of all well-known painters?

When we walk a path, the borders of the path converge too, and our brain accepts it, because we look horizontally.

Yes, and we had the same view, the correct perspective if we saw the same path on a photo. Or not?

There are limits to what we should do with perspective corrections. Imagine shooting a recessed window from one side and then correcting the perspective in order to make the top and bottom parts parallel. Due to the recess, we’d see one side of the window frame and not the other. We’ll notice that something is wrong with that image…

Yes, that sounds plausible. So these wrong perspectives, lines, proportions let the image, building look wrong.

We also use 5" x 4" LF cameras with movements. Here’s a couple of shots of tall buildings with the film plane vertical and the front raised…
Architecture 4
Architecture 3

You can see in these pictures that the point of view is on the ground when taking the pictures, I assume. So if they were taken from a higher point they might look better? But the images look right, as I like it.

I first aligned the right side with the middle using the Force Parallel tool then exported as a TIFF. Then I used the FP tool on the new file to align the left side with the other two. The “wider at the top” and “left side leaning in” illusions are still present but all three lines are now vertical.
https://i.imgur.com/7PXOdkN.jpeg

But the line with the arrow is not, I would think. The other lines (almost) seem to be straighten but do not look straighten at all (because of the optical illusion obviously).

Thank you very much!

49

You’ve got all the necessary help.
What is your specific question now?

50

You’ve got all the necessary help.

Ah, yes, many thanks for all the necessary help!

What is your specific question now?

Thank you, at the moment I would say I am afraid there is no, cannot be a specific question or a question at all left anymore as it unfortunately appears such (kind of) image(s) are not usable, not to be saved (because of the optical ilusion, the lines, proportions being left after correction, remaining wrong, which seems uncorrectable or perhaps correctable with extreme effort). That’s really too bad.

But one question might be, how much could one tilt a cam wihtout getting wrong lines, proportions, optical illusion(s). Respectively could one recognize wrong lines while tilting? So could it be seen at which degree of tilting the lines become wrong or the optical illusion occurs?

51

Well, you should see this – and I can only guess better on the camera’s backside screen than in the viewfinder.

So could it be seen at which degree of tilting the lines become wrong or the optical illusion occurs?

@Dox – how do you want to ‘measure’ what is an acceptable degree of tilting? I don’t know, if one might get the according readout on a large format cam or when using a tilt shift lens → after it felt right.
While one takes a photo somehow, the other visualizes the image in advance and composes the content instinctively – and subjectively.

All I can say, experiment and train your gut feeling.

52

Many thanks!

Also for the link. I actually do not like converging lines that are vertical in reality, respectively I even would want to get converging lines in reality vertical, 100 % straight on the photo.

Yes, training my gut, thank you for that advice!