View Single Post
  #1  
Old 31st January 2019
Bikie John's Avatar
Bikie John Bikie John is offline
Full member
 
Join Date: Nov 2008
Location: Wessex
Posts: 3,995
Thanks: 187
Thanked 652 Times in 580 Posts
Likes: 511
Liked 734 Times in 491 Posts
Correcting perspective

Prompted by Paul's (pdk42) series of photos of the Flatiron Building (thread here: http://e-group.uk.net/forum/showthread.php?t=49588) and the discussion about comparison of the human eye to a camera (http://e-group.uk.net/forum/showthread.php?t=49640) I have been thinking about correcting perspective. When we shoot big buildings we usually tilt the camera upwards, which makes the building look as though it is falling over backwards, and this is quite disturbing.

Editors give us tools to resolve the converging verticals, and we used to do it in the darkroom by tilting the printing paper, but the results never look quite natural. Paul's photos show this, the building tends to come out looking top-heavy.

I have just done a little experiment to see what the editing tools are doing, and found it quite illuminating. I took a photo of a mount-cutting mat with a square grid printed on it, and shot from an angle as if shooting upwards at a building. I left plenty of space around it to allow use of the vertical correction tools without having to crop outside the edges. Here is the full shot with no edits, just resized for the thread:



Then in Lightroom I mashed the tonal values to simplify playing with it:



Here is a crop of just the mat so we can see what is happening:



As expected, the lines going away from us (which would be the verticals on a building) are converging to a vanishing point. The transverse lines stay parallel but get closer together as we get further away.

This is the result of doing a vertical perspective adjustment in Lightroom, I really don't know why it has chopped some off as I did it from the full uncropped file:



Photoshop is a little more sensible and doesn't chop any bits off:



We can see that the transformation has restored the original shape of the subject - the boxes are all nice and square. This is neat, and probably working as it should.

I think this is at the heart of why adjusted photos look wrong. When we look at a photo, we know that the top of the building is further away so the top floors should look smaller. With these transforms they are the same height which makes them look unnaturally large and hence the building looks top-heavy.

I reckon that what we need to do is add a sort of "spongy squash" from the top down so that the top is compressed more than the bottom, with a nice even gradient. If we could do that the result would look more natural.

I have been playing quite a bit this morning and not yet found anything that will do this. Do any of our post-processing wizards have any bright ideas?

John

PS You are right, I do need to get out more
Reply With Quote
The Following User Says Thank You to Bikie John For This Useful Post:
OM USer (31st January 2019)