Parallel...or Not

[Pelorus]

As a newish tech cam user I've really enjoyed the learning process - getting the camera set up right, movements...

I've got one small challenge that's really irritating me however: Whilst I can get the camera level and upright by using the spirit levels on the camera, I sometimes struggle to get it parallel to the scene. This is particularly so with architecture, and the results are frustrating. Because I've got the camera set up level in both planes, I sort of feel like I've got a right to expect the verticals and horizontals will all line up...and sometimes they don't.

The issue seems to be the rotation of the camera about the vertical axis. Because the camera is so small compared to the scene it is very difficult to get it exactly parallel all of the time. It's just really difficult to judge. I've taken to setting up using the spirit levels and then sighting across the top of the camera at a (hopefully) horizontal line, assuming that any divergence between the top of the camera and he line is from left or right rotation of the camera. Sometimes it works...and sometimes it doesn't.

Any tips or tricks would be gratefully received.

Mike

 

[Pelorus]

Here's a rather extreme example - it was pretty dark where the camera was. It's made worse by the fact that the shop is on a slope and I have a sneaking suspicion that nothing in the building is square. However the display window shelf is more horizontal than this picture suggests.

 

[Geoff]

You aren't square to the plane of the building. A few tricks, then:

- align top of camera with something in your picture plane
- find out what the building is aligned to, and aim for that: for example, sometimes I'll look down the top of the camera from the side and align that line with some point in the far distance that I know is parallel with the building face.
- use slight rotations with a geared setup, to get one right
- LR and C1 have pretty good correction tools.

Of these, the second one works best for me. Correction tools are easy and sometimes necessary.

 

[tjv]

What back are you using? If it's one with good (CMOS) live view, then enable the grid lines in the display and use a good geared head. For CCD backs, it's easy enough to line the scene up and take a test shot, then adjust from there. This is one of the reasons I really enjoy using a ground glass back for focusing and framing though, as I feel more in tune with the process–it's a very subjective thing–and don't enjoy looking too much at the LCD screen for framing, only exposure confirmation. With the GG things are pretty easy to setup quickly, without resorting to multiple test frames.

If you're using a pancake camera you might also frame up the shot then use your distometer to measure the distance to the extreme edges of the frame and rotate / move the camera until you've got everything straight on the horozontal plane, then the geared head to adjust yaw and verticals. It'd be touch and go to get it completely accurate I guess, but better than guestimation.

 

[GrahamWelland]

I understand the frustration as I found that trusting levels vs what I see can be a problem. I will use the grid view either in live view or review. When shooting flat plane objects like building & door/Windows it isn't always possible to be in the ideal place but I will use shift and/or rise as far as possible but sometimes you just can't be exactly perpendicular to the subject.

The other thing, if you don't mind bending a few pixels from raw, is to shoot a little wider and leverage the truly excellent C1 keystone correction tool. Only you will know

 

[Pelorus]

Thanks Geoff, Graham and tjv. I don't have live view worth anything...it's a Leaf Aptus-II 7 which I really like.

The problems are multiple: It's quite hard to see everything in the dark, particularly with specs; I find the view on the screen frustrating for small scale corrections like that; and I'm not much chop at geometry at the best of times.

But as Graham says... (and it could still tolerate a tiny bit more correction)

 

[AreBee]

Mike,

I've taken to setting up using the spirit levels and then sighting across the top of the camera at a (hopefully) horizontal line, assuming that any divergence between the top of the camera and he line is from left or right rotation of the camera... Any tips or tricks would be gratefully received.

1. A company logo is normally positioned level, even on sloping buildings.

2. The accuracy of your method of lining up the camera body top edge with a parallel feature will be maximised by locating your eye such that the camera body width is seen to equal the length of the longest horizontal feature (ideally equal to the image horizontal field of view).

3. The farther a feature is from the camera, the closer it will approach the horizon. Hence, with distance clouds move 'down' an image; with distance railroad tracks move 'up' an image. The converse is also true. If the camera is level and no rise/fall adjustment is made then the horizon is located at the image mid-height.

If the window pane bottom edge is horizontal then the left side is farther from the camera than the right side because it has moved 'up', closer to the horizon, causing the line to slope. In order to shorten the distance on the left (or, if you prefer, increase the distance on the right), rotate the camera about its vertical axis in a clockwise direction, when viewed from above.

4. Notice that the window pane top edge slopes down from right to left and the bottom edge slopes up from right to left. If two parallel edges are seen to slope in opposite direction then the horizon is located between them. If the slope value is identical then the horizon is equidistant from each edge.

The slope value of the window pane top and bottom edge is unequal - the bottom edge slopes more than the top edge. This indicates that the horizon is closer to the top edge than the bottom edge. If you review an image for the purpose of checking parallelism of camera and subject, consider horizontal features farthest from the horizon, because feature slope will be greatest, and hence the accuracy of error correction maximised.

Hope this helps.

 

[jlm]

Your example indicates, as you stated, that regardless of level and plumb correctly set, your sensor plane is not parallel to the building facade
A rotation of the camera about the vertical axis should have solved this
Had you a view camera with swings in the back you could have corrected that way
Similar problem shooting flat art where i use a pointing laser in a small case against the target aimed at the lens where i have a reflecting glass. When the dot reflects exactly back on the beam the lens axis is dead square to the image plane
The laser is an enlarger alignment tool purchased at b&h
versalab alignment tool: $140