More Tilt Shift Questions

[fike]

So as I consider how a tilt might be useful in concert with mosaic images stitching, I have been considering the towering tree scenario.  When you are at ground level and you want to get a towering tree in focus, you end up focusing somewhere several feet up the tree directly in front of you and then use a narrower aperture to increase depth of field (you may pull out your hyperfocal calculation charts or you may just use 1/3 of the distance to infinity).  At this point you hope that the focus is good enough immediately in front of you and as the tree trunk gets higher and further away.  

It seems to me that at wide angles the focus distance makes the focal plane actually an arc (or some segment of a speherical shape) with the lens at the center of the arc.  Because of this, getting the tree trunk fully in focus can be difficult.  

So, use a tilt to move the focal plane so that it leans away from the camera has it goes higher, effectively keeping the tree in the focal plane.  

Does any of this make sense?

[pfigen]

"It seems to me that at wide angles the focus distance makes the focal plane actually an arc (or some segment of a speherical shape) with the lens at the center of the arc.  Because of this, getting the tree trunk fully in focus can be difficult.  

So, use a tilt to move the focal plane so that it leans away from the camera has it goes higher, effectively keeping the tree in the focal plane."

It sounds like you're referring to curvature of field, where in a perfect world the plane of focus would be flat and parallel to the film plane but in a real world it actually curves back toward you as you move to the edges and corners of the frame. A lot of wide angles suffer from this to varying degrees - all of the Nikon wides I've ever used, the Canon 24 1.4, etc, but not my Zeiss 21mm Distagon and more pertinent to your question, none of the tilt-shift or shift only lenses I have exhibit curvature of field to a noticeable degree. The two new Canons are particularly good in that respect.

[ckimmerle]

Most wide angle lenses will arc the focal plane to a certain degree, but the amount is dependent on it's design. There are some specified "flat field" lenses, but I don't know of any in particular. The only situation that might be of concern with a curved focal plane is when photographing a flat plane on the perpendicular. In that case, the center of the frame is closest to the lens (remember, we're perpendicular), and the rest of the subject gets further away the closer it gets to the edge of the frame. If a particular lens has a high degree of curvature in the focal plane, the extreme edges may get a bit soft as the focal plane will bend towards the lens, but the subject is further away at the far edges. They go in opposite directions. Usually, stopping down a stop, or two, is all it takes. Gawd bless DOF.

As for tilt-shift lens, you are correct. All you're doing is laying (standing?) that imperfect focal plane along the plane of your trees.

This is different than curvature of field, btw.

[fike]

...

As for tilt-shift lens, you are correct. All you're doing is laying (standing?) that imperfect focal plane along the plane of your trees.

...

So I can tilt the the lens down (probably not too much) to improve focus at the top and the bottom of the tree?

[ckimmerle]

So I can tilt the the lens down (probably not too much) to improve focus at the top and the bottom of the tree?

Exactly. You'll find that it doesn't take much tilt at all, though. I rarely use more than a degree, or so when the camera is at eye level. If shooting very close to the ground and laying the focal plane horizontal, more tilt would, of course, be required. For shooting very tall trees, the amount of tilt will be primarily related. among other factors, to your distance from the subject.

There's a rule called the Scheimpflug Principal that illustrates how this works.

[fike]

...

There's a rule called the Scheimpflug Principal that illustrates how this works.

I've been reading-up on Scheimpflug and trying to make sure my understanding of the principle is consistent with real world applications of a TS lens.

[AndrewKulin]

There's an article in L-L that I found quite helpful.  The picture near at the beginnig of the article was the best I have seen to date putting the whole Scheimpflug principal into focus for me.  Later the author provides tables indicating angles of tilt against the height of the camera - essentially the closer the camera is to the ground (assuming ground is the plane you are trying to keep in focus), the greater the angle you need to tilt.

Here is the link:

http://www.luminous-landscape.com/tutorial...cusing-ts.shtml

[DanielStone]

here's a brief video that explains scheimpflug very well IMO. yes, its with polaroid film and 4x5, but the idea is still the EXACT same, at least the principle is.

helped me get it down, I still go back to it from time to time as a refresher

http://www.youtube.com/watch?v=gR4m70xr9mE

-Dan

[elf]

You may want to read through this thread before spending $1000+ on a T/S lens: http://www.tawbaware.com/forum2/viewtopic.php?t=5910

[fike]

You may want to read through this thread before spending $1000+ on a T/S lens: http://www.tawbaware.com/forum2/viewtopic.php?t=5910

That is an interesting and informative thread.  I guess there are two takeaways...one, stitchers can be flummoxed by the almost perfect alignment and two, there are easy ways around the problem.

As for me, I am thinking that I will use the shift only to move between first row and second rows of the panorama. I actually am much more interested in using tilts to make the depth of field adhere more closely to my needs. As such, I may still use a spherical pano head to rotate the camera.

[Bart_van_der_Wolf]

As for me, I am thinking that I will use the shift only to move between first row and second rows of the panorama.

Although that would introduce a certain amount of vertical parallax, with a good blending routine you can probably get away with it, especially if the upper row contains few foreground features. With a liberal amount of overlap, there is likely to be an invisble solution possible, nature being easier than architecture.

I actually am much more interested in using tilts to make the depth of field adhere more closely to my needs. As such, I may still use a spherical pano head to rotate the camera.

Unfortunately the tilt also causes a (very) small offset of the entrance pupil, but it's usually not a problem unless there is some extremely close foreground detail in both/multiple rows.

Cheers,
Bart

[fike]

Although that would introduce a certain amount of vertical parallax, with a good blending routine you can probably get away with it, especially if the upper row contains few foreground features. With a liberal amount of overlap, there is likely to be an invisble solution possible, nature being easier than architecture.

I use an arca swiss bracket on my spherical head that I think would allow me the latitude needed to lower the camera enough that this effect would be minimal or non-existent.  

Unfortunately the tilt also causes a (very) small offset of the entrance pupil, but it's usually not a problem unless there is some extremely close foreground detail in both/multiple rows.

I think that with some experimentation that I should be able to calibrate for this.  Also, as you mentioned, in landscape work you can generally blend the seems well enough to hide the misalignments that are likely to be minor.

[JeffKohn]

As for me, I am thinking that I will use the shift only to move between first row and second rows of the panorama.

IMHO there are plenty of reasons to get a T/S lens (I have 3 of them), but this is not one of them. All this really accomplishes is to create extra work in stitching, with no real gain in image quality. A pano-head that can handle multi-row stitching makes more sense, and also eliminates any worries about parallax error on the vertical axis (today's stitchers/blenders can often do a good job of hiding parallax, but not always).

I don't agree with some of the other comments in that Tawbaware thread, though. Rotational stitching with a pano head is not faster than using the shift-only approach. Part of the reason I like the shift-only approach is that it's very fast and easy to get a 3-image stitch, either for a modest bump in resolution or to get a quick wide-aspect composition without having to get the pano head out. The other reason I like shift-only stitching is that I can easily stitch from the RAW using Photomerge (the 'reposition only' option is what you want for flat-stitching). It's quick and doesn't require generating intermediate tiff's.

Add DOF control with tilts, and in-camera perspective control where you can actually see the final composition, and T/S lenses are immensely useful even for landscape shooting IMHO.

[Patricia Sheley]

[attachment=21766:C_Sheley...tHarmony.jpg]

There's an article in L-L that I found quite helpful.  The picture near at the beginnig of the article was the best I have seen to date putting the whole Scheimpflug principal into focus for me.  Later the author provides tables indicating angles of tilt against the height of the camera - essentially the closer the camera is to the ground (assuming ground is the plane you are trying to keep in focus), the greater the angle you need to tilt.

Here is the link:

http://www.luminous-landscape.com/tutorial...cusing-ts.shtml

Thank you Andrew for this referewnce to David Summerhayes' excellent article...for whatever reason, the way he discussed it just "clicked"   Much appreciated... Patricia

[Kirk Gittings]

FWIW, I agree with Jekk Kohn's perspective above. Shift stitching works so effortlessly. I use Canon T/S lenses and CS4 for architecture stitches many times weekly. I use the new 24 (also with the 1.4 Teleconverter to render it a approx. as a 35T/S), the 45 and 90. I dare say that these are my bread and butter. Working on a tripod and leveling the camera renders exposures that are easily stitched in Photomerge on automatic setting.

[Patricia Sheley]

FWIW, I agree with Jekk Kohn's perspective above. Shift stitching works so effortlessly. I use Canon T/S lenses and CS4 for architecture stitches many times weekly. I use the new 24 (also with the 1.4 Teleconverter to render it a approx. as a 35T/S), the 45 and 90. I dare say that these are my bread and butter. Working on a tripod and leveling the camera renders exposures that are easily stitched in Photomerge on automatic setting.

This is great encouragement for me...I have the old TS-E 24mm and use it with the  1.4 x all the time... an easy take along...I've been shooting infrareds this week, and took the time to do a few shift panos...I'm terrified of trying Photomerge the first time...also using CS4    maybe it's time for this ancient to bite the bullet...

Do you find the new 24 better incrementally? Thanks for your input...time to try a stitch...hoping the first time only is the hard part... Pat

[Kirk Gittings]

This is great encouragement for me...I have the old TS-E 24mm and use it with the  1.4 x all the time... an easy take along...I've been shooting infrareds this week, and took the time to do a few shift panos...I'm terrified of trying Photomerge the first time...also using CS4    maybe it's time for this ancient to bite the bullet...

Do you find the new 24 better incrementally? Thanks for your input...time to try a stitch...hoping the first time only is the hard part... Pat

Pat, actually I find the new 24 dramatically improved over the old one.

[Bart_van_der_Wolf]

FWIW, I agree with Jekk Kohn's perspective above. Shift stitching works so effortlessly. I use Canon T/S lenses and CS4 for architecture stitches many times weekly.

Hi Kirk,

How do you offset the shift of the entrance pupil? I occasionally "flat-stitch" and use stopbars on my RRS rail so that I can quickly shift the lens and countershift the camera. The speed at which the tiles can be captured may help in avoiding shifting cloud/shadow artifacts and such. The approach with stopbars only works for a fixed offset distance along the shift dimension. What do you do when you need to shift diagonally, or do you only shift vertically for height, or horizontally for FOV?

I find that acquiring tiles for rotated stitching can be performed quite fast because only one repositioning is required and not two for each additional tile.

Of course the ease of postprocessing for a flat stich is very nice.

Cheers,
Bart

[Kirk Gittings]

Hi Kirk,

How do you offset the shift of the entrance pupil? I occasionally "flat-stitch" and use stopbars on my RRS rail so that I can quickly shift the lens and countershift the camera. The speed at which the tiles can be captured may help in avoiding shifting cloud/shadow artifacts and such. The approach with stopbars only works for a fixed offset distance along the shift dimension. What do you do when you need to shift diagonally, or do you only shift vertically for height, or horizontally for FOV?

I find that acquiring tiles for rotated stitching can be performed quite fast because only one repositioning is required and not two for each additional tile.

Of course the ease of postprocessing for a flat stich is very nice.

Cheers,
Bart

I don't worry about it. Its not an issue. Yes I also routinely do diagonal shifts when I want say a rise with the flat stitch.

[Bart_van_der_Wolf]

I don't worry about it. Its not an issue.

Unfortunately, for detailed foreground features, I do find a 23 or more pixel shift (T/S-E 24 II on 1Ds3) to be an issue, a problem even. For simple foregrounds things are simple, even handheld shooting would work.

Cheers,
Bart