Panoramic head vs. shift - question

[AreBee]

Question: All else being equal, if rectilinear projection is adopted when rendering a set of images shot using a panoramic head, will the stitched image overlay perfectly with the same field of view obtained from a stitched set of images shot using shift on a camera with movements, in which the lens is assumed stationary?

I am interested in ascertaining if the 'faceted' nature of an image obtained from a panoramic head is taken into account by the projection method adopted.

[Ellis Vener]

Question: All else being equal, if rectilinear projection is adopted when rendering a set of images shot using a panoramic head, will the stitched image overlay perfectly with the same field of view obtained from a stitched set of images shot using shift on a camera with movements, in which the lens is assumed stationary?

A: IF the horizontal and vertical angles of view are equal and that  in both cases  and IF any near/far parallax errors are minimal to non-existant, you more likely than not will see no differences.  But those are two big "ifs".

[Paul2660]

In my experience, it's very dependent on the lens, and if your setup is nodal and level.  If you use a ultra wide, say 14mm to 20mm, even nodal you will find distortion as you pan (you can see objects distort as you pan across).  This is worse if the lens is held in the landscape orientation.  So for ultra wides, I usually go vertical and take more shots.  With wides, the cylindrical solution often works better albeit with some loss of image on the right and left side.  

In regards to movements, if your setup allows shifting the body, not the lens, (say with a tech camera) or a zork adapter on a 35mm camera, then you can stitch all day long, not worry about nodal points etc. and easily get a great solution rectilinear.  You don't have to worry about being level either as you are only move the camera body or MF back.  I use this all the time and get great results with stitched images.  The main limitation is that 15mm tends to be the max movement with MF and 18mm with a zork and that won't really a true 3:1 pano image.  But it's still a great image.

As much as I have stitched and panned over the years, with 35mm, I tend now to do everything hand held if the subject allows it then throw the images into PtGui and see what I get.  Most times I will get a solution that works for me may not be rectilinear, but PtGui has a lot of different projections to try.  You also can warp the final product a bit in CC to help square it out, but you have to watch this as you can start to lose detail pretty quick.  You can do this in a landscape most times but not with a static known object like a building.  Trees, rocks, etc. will easily blend together, they may be extended or warped a bit, but it all looks fine when put together with a good stitching product.  

There are times, few now, that I still setup for a nodal stitch, but I find in my work I most times just don't need it.

Paul

[alan_b]

Generally/theoretically, yes.  A couple more practical wrinkles in addition to what's been mentioned already:
1. The shift lens may not give a "perfect" rectilinear projection due to its own distortion.
2. There will be a difference at the microdetail/texture scale, depending on angle between frames of the stitch. Too much angle between frames and there will be noticeable change in direction of the stretching of grain/detail.
3. The shift lens will probably show more optical defects near the edge of its image circle.

Finding the sweet spot between all these degradations takes some experimentation.  (I'm sure someone smarter than me could come up with a graph to identify sweet spots.)

[AreBee]

Ellis, Paul and Alan,

Thank you very much.

I assumed that there would be no difference, my reasoning being that for that to not be true a change in direction would be visible in straight line features where they crossed 'edges' of individual images shot with a panoramic head (ignoring how blending actually takes place), which would have been contrary to the property of rectilinear projection - straight lines in the scene rendered as straight lines in the stitched image.

However...

There will be a difference at the microdetail/texture scale, depending on angle between frames of the stitch. Too much angle between frames and there will be noticeable change in direction of the stretching of grain/detail.

The above seems to suggest exactly that.

Please can you clarify?

[kers]

Generally/theoretically, yes.  A couple more practical wrinkles in addition to what's been mentioned already:
1. The shift lens may not give a "perfect" rectilinear projection due to its own distortion.
2. There will be a difference at the microdetail/texture scale, depending on angle between frames of the stitch. Too much angle between frames and there will be noticeable change in direction of the stretching of grain/detail.
3. The shift lens will probably show more optical defects near the edge of its image circle.
….

I agree 100% with Alan…
If there is no distortion in the lens used for shifting a 100% match is possible in position but not in image quality. Image quality depends on the lensquality in both cases.
There always will be a degrading of the image quality from center to the boundaries .

In the case of the TS lens; because of the lens quality that is less in the corners.
In the case of stitching; because the images in the corners have to be stretched.
The better image could be both, but in reality it is more likely to get the better image through stitching.

[BernardLanguillier]

It depends on the focal length used to do the spherical stitch and of the targeted angular coverage.

Besides, one has to keep in mind that the image quality of wide T/S lenses, even in their central non shifted section, is pretty average (even for the best in class Canon lenses). They are in essence MF designs that attempt to have a reasonnable uniformity over a large image circle but fail to reach high standards of image quality compared to longer non T/S designs. This is plenty clear if you review the www.lenscore.org results.

In the end, the image spherical stitched with a top normal prime such as the Sigma 50mm Art or Otus, is vastly superior to the T/S image in all accounts. It's not even close. The gap will be larger if you use longer focal lengths to stitch, at the cost of longer stitching times are risk if something moves in the image. This is btw already the case when comparing a top prime such as the Nikon 24mm f1.4 at its best aperture (say f6.3) and the equivalent stitch.
 
Cheers,
Bernard

[Hulyss]

I use a rail for stitching and a very sturdy tripod. I make a first batch of tree photos (or more, depending of the rail) and then I elevate the column and restart. Need some practising to make regular overlap but overall, my flawless method.

[alan_b]

2. There will be a difference at the microdetail/texture scale, depending on angle between frames of the stitch. Too much angle between frames and there will be noticeable change in direction of the stretching of grain/detail.

Please can you clarify?

Look at the seams of a stitch in a noisy area and you'll see the change in stretching direction of noise & fine detail. 

[Jim Kasson]

Here's a little case study. I came down on the side of spinning, rather than sliding:

http://blog.kasson.com/?p=2707

Jim

[shadowblade]

Here's a little case study. I came down on the side of spinning, rather than sliding:

http://blog.kasson.com/?p=2707

Jim

You also did the test with a shift lens that has notoriously poor image quality when shifted even just a little bit.

[robdickinson]

I often use both. In combination.

With just a rotator and the camera level you can easily use shift to set the horizon and then shoot a panoramic in the usual rotational way.

[Cayman]

Besides, one has to keep in mind that the image quality of wide T/S lenses, even in their central non shifted section, is pretty average (even for the best in class Canon lenses). They are in essence MF designs that attempt to have a reasonnable uniformity over a large image circle but fail to reach high standards of image quality compared to longer non T/S designs. This is plenty clear if you review the www.lenscore.org results.

Bernard, I have looked at the lenscore results but have trouble reconciling them with the results at http://www.the-digital-picture.com.   If you look at the example of the 24 TS-E II there it does better than just about any lens it is put up against.   Any thoughts onto why that is?

[BernardLanguillier]

Bernard, I have looked at the lenscore results but have trouble reconciling them with the results at http://www.the-digital-picture.com.   If you look at the example of the 24 TS-E II there it does better than just about any lens it is put up against.   Any thoughts onto why that is?

No idea.

The way lenscore works is pretty unique as you can read if you click on "about" in the left navigation bar.

This being said, Lenscope confirms that the 24mm TS-E is an excellent wide angle lens, probably the best indeed, but it still falls short by a large margin compared to longer focal lengths.

Cheers,
Bernard

[Bart_van_der_Wolf]

The way lenscore works is pretty unique as you can read if you click on "about" in the left navigation bar.

This being said, Lenscope confirms that the 24mm TS-E is an excellent wide angle lens, probably the best indeed, but it still falls short by a large margin compared to longer focal lengths.

The LenScore metrics are relative numbers, relative to one lens set as a base. As usual, the easier optical designs like longer focal lengths score higher (and are used as a base), but amongst the 24mm lenses the TS-E 24mm II scores high. What their scores also do not reflect is how the lens will combine with a camera of a certain sensel pitch. It's the combined lens+camera MTF that makes a system MTF, which is what counts in the end.

Also, one of the benefits of stitching is that one can (should?) use a longer (better corrected) focal length, because the required FOV can still be achieved by just adding more tiles for stitching. Even for a single row stitch, by rotating the camera to portrait orientation allows to use a (e.g. 40%) longer focal length (35mm vs 24mm) with the same vertical FOV as a single landscape oriented image.

In general, also because the rotated stitching tiles only use the center of the image circle, the stitched result should be equal or (in the case of a better longer focal length lens) superior to a shifted result, although less convenient. Using something like a 20-30% overlap will make sure that the blending engine can pick the highest resolution parts of the image circle. In addition, geometrical lens distortions and vignetting will/can be automatically corrected by the stitcher application, which is sometimes more accurate than separate lens corrections that are based on a generic model.

Cheers,
Bart

[shadowblade]

The LenScore metrics are relative numbers, relative to one lens set as a base. As usual, the easier optical designs like longer focal lengths score higher (and are used as a base), but amongst the 24mm lenses the TS-E 24mm II scores high. What their scores also do not reflect is how the lens will combine with a camera of a certain sensel pitch. It's the combined lens+camera MTF that makes a system MTF, which is what counts in the end.

Also, one of the benefits of stitching is that one can (should?) use a longer (better corrected) focal length, because the required FOV can still be achieved by just adding more tiles for stitching. Even for a single row stitch, by rotating the camera to portrait orientation allows to use a (e.g. 40%) longer focal length (35mm vs 24mm) with the same vertical FOV as a single landscape oriented image.

In general, also because the rotated stitching tiles only use the center of the image circle, the stitched result should be equal or (in the case of a better longer focal length lens) superior to a shifted result, although less convenient. Using something like a 20-30% overlap will make sure that the blending engine can pick the highest resolution parts of the image circle. In addition, geometrical lens distortions and vignetting will/can be automatically corrected by the stitcher application, which is sometimes more accurate than separate lens corrections that are based on a generic model.

Cheers,
Bart

Individual frames from a rotational pano need to be stretched and distorted quite a lot to produce the equivalent of a rectilinear wide angle shot. The wider the virtual lens, the greater the distortion at the edges. In order to resolve the same detail as a shifted tilt-shift lens, you need to start with more pixels and a longer focal length with a rotational pano, in order to retain enough detail after stretching and distorting them to keep the image rectilinear. Fortunately, this is very possible to do.

[Bart_van_der_Wolf]

Individual frames from a rotational pano need to be stretched and distorted quite a lot to produce the equivalent of a rectilinear wide angle shot.

The same stretching occurs in the corners of a lens projected image on a flat sensor, and the optical quality at the edge of the image circle is much lower. The only difference is that the sensor will capture the stretched (and lower lens resolution!) pixels accurate, and the rotated pano will have to interpolate its (much higher quality) pixels very accurately. The interpolation is rather accurate, because it's only a small stretch if you do not exceed what would be in the image circle of the shifted lens. Using a longer focal length for stitching will close the 'gap' even more.

Here is a demo/example of how the image quality deteriorates when one is approaching the image circle of a very good TS-E lens (click on the image for a full resolution version, full size=9,846,388 bytes!):

And here is another (click on the image for a full resolution version, full size=11,560,629 bytes!):

Note that in both images also a small amount of tilt was applied (which after all is the forte of a T/S lens, so why not use it), which would require stitching with a regular rotated pano capture. Also, they are straight Raw conversions, without CA correction or special sharpening (other than a uniform sharpening to recover from diffraction at f/11).

The wider the virtual lens, the greater the distortion at the edges. In order to resolve the same detail as a shifted tilt-shift lens, you need to start with more pixels and a longer focal length with a rotational pano, in order to retain enough detail after stretching and distorting them to keep the image rectilinear. Fortunately, this is very possible to do.

Exactly what I was referring to, and it would even accommodate wider FOVs than possible with many T/S lenses at full +/- shift. The TS/E 24mm would offer something like a 102 degrees HFOV in landscape orientation, with its virtual 60x24mm sensor.

However, with such extreme FOVs, that will look ugly anyway (when exceeding something like 110-120 degrees in a rectilinear projection), one does need to consider DOF and focus distance issues. The rotated pano will have a circular focus distance/DOF-zone, where a shifted T/S lens will have a (hopefully somewhat) flat focus distance/DOF-zone. In practice these T/S lenses do not always have zero field curvature, so their focus is not necessarily in a flat plane. Also in practice, a rotated pano will have lots of DOF to keep tiles well focused, and if not then one needs to adjust focus for each tile (which is a nuisance but doable). More commonly, if the FOV is so extreme, one will use a different projection method if the subject allows.

Cheers,
Bart

[Jim Kasson]

You also did the test with a shift lens that has notoriously poor image quality when shifted even just a little bit.

If you don't like mine, please post a comparison using a lens of which you approve.

Jim

[AreBee]

Thanks folks.

I've concluded that in principle both images will overlay exactly but that the result returned via shift will, as yaw angle interval increases in the case of the result obtained from a panoramic head, be visually more 'correct' (no visible change in stretching direction because all images are shot in one plane), although in terms of image quality it may be poorer.

[kers]

Thanks folks.

I've concluded that in principle both images will overlay exactly but that the result returned via shift will, as yaw angle interval increases in the case of the result obtained from a panoramic head, be visually more 'correct' (no visible change in stretching direction because all images are shot in one plane), although in terms of image quality it may be poorer.

+ in the case you use a shift lens you have more control - at the moment of shooting - over the end result/ composition.
I think that is also an important factor.