stitching panoramas from a tilt/shift lens, with focus plane tilt but no shift
Wildernesscapes
images taken with a tilt/shift lens, such as the Canon TS-E 24mm, with
the camera rotated around the nodal point, with no lens shift, only
tilt. I am asking because I'd like to stitch together images taken
with my tilt-shift lens, rotated around the nodal point of my lens. I
don't use the shift feature to create a two image panorama. My goal is
to be able to create a stitched wide horizontal panorama in which i
tilt the plane of focus to match the receding landscape, rotating the
camera with each shot. A second goal would be create a 2-image
vertical panorama, with the lower image having a focus plane flater,
and the top image with a normal perpendicular focus plane (so I could
get both foreground flowers receding off into the distance, and the
tall mountains in the background, all completely in focus, as an
example). Are these things possible with PTGUI, which I use
exclusively for all my stitching?
I can generate/add accurate control points, in my prior trials, but a
large optimization error is incurred, the lens mm value get
automatically changed (from say 24mm to 58mm), and the final stitched
seam is very bad in the final photo. Also, if I am using a 1.6x
teleconverter, would I manually change the 24mm lens value to 38.5?
(but it still automatically changes to 58mm when optimizing.).
Thanks for your information.
Johnathan Esper
Erik Krause
> I am wondering if it is possible to stitch panoramas from source
> images taken with a tilt/shift lens, such as the Canon TS-E 24mm, with
> the camera rotated around the nodal point, with no lens shift, only
> tilt. I am asking because I'd like to stitch together images taken
> with my tilt-shift lens, rotated around the nodal point of my lens.
You probably mean the no-parallax-point, not the nodal point.
> I
> don't use the shift feature to create a two image panorama. My goal is
> to be able to create a stitched wide horizontal panorama in which i
> tilt the plane of focus to match the receding landscape, rotating the
> camera with each shot.
If you use the tilt feature of your lens you have a different effective
focus distance for the upper and lower part of your image which results
in different magnification and probably even in a different
no-parallax-point. It will be hard to stitch such images. You would need
to allow optimization for shift parameters (d and e), since the optical
axis doesn't hit the image center. Or use viewpoint correction to
compensate for different magnification.
> A second goal would be create a 2-image
> vertical panorama, with the lower image having a focus plane flater,
> and the top image with a normal perpendicular focus plane (so I could
> get both foreground flowers receding off into the distance, and the
> tall mountains in the background, all completely in focus, as an
> example).
There are easier ways to achieve a very wide depth of field. Use focus
bracketing in this case. Tufuse will combine the sharpest parts of a set
of images. You will need to align each stack of images in order to meet
field of view differences. This is easily done with PTGui if you allow
for individual lens parameters for each but the anchor image. Output
individual images and pass them to tufuse for focus stacking.
> Are these things possible with PTGUI, which I use
> exclusively for all my stitching?
Probably, but it will be hard. Could be morphing will be a solution. You
need panotools installed in order to use morphing.
> I can generate/add accurate control points, in my prior trials, but a
> large optimization error is incurred, the lens mm value get
> automatically changed (from say 24mm to 58mm), and the final stitched
> seam is very bad in the final photo. Also, if I am using a 1.6x
> teleconverter, would I manually change the 24mm lens value to 38.5?
> (but it still automatically changes to 58mm when optimizing.).
Why would you use a teleconverter with a tilt lens?
See also: http://www.ptgui.com/support.html#2_20
--
Erik Krause
Kevin Wilton
point.
No, if he is doing it correctly, he does mean the front
nodal point.
Erik Krause
Sorry, this has been discussed over and over again. The front nodal
point and the no-parallax-point have nothing in common:
http://toothwalker.org/optics/misconceptions.html#m6
http://doug.kerr.home.att.net/pumpkin/Pivot_Point.pdf
and of course the comprehensive "Theory Of The No Parallax Point":
http://tinyurl.com/d29lu
--
Erik Krause
Kevin Wilton
--
Erik Krause
Whatever you think, whatever you say, the point of rotation
is the front nodal point. Do you think the head is called a
Nodal Ninja by accident. Come on, get over it.
Kevin
John Houghton
> Do you think the head is called a Nodal Ninja by accident.
From the NN5 User Manual: Using a special head like the Nodal Ninja
and adjusting it properly so it rotates about the "entrance pupil" of
a lens will eliminate virtually any parallax.
In general, the entrance pupil of a lens is not coincident with the
front nodal point. See this list, for example, or make your own
measurements - it's not difficult: http://www.swissarmyfork.com/lens_table_1.htm
.
In the face of clear evidence to the contrary, it seems to me perverse
to seek to perpetuate the myth that you should rotate about the front
nodal point.
John
Wildernesscapes
entrance pupil argument. Actually, I didn't know there was any
difference. I wish there were better diagrams of lenses in the article
on the entrance pupil though. I'm a bit confused still where the light
rays come together in a single point - is that the nodal point or
entrance pupil? I intuitively did in the past rotate around the
entrance pupil (thinking that was the nodal point), or used values off
the Nodal Ninja's database. Checking the shift boxes in the advanced
optimization tab did improve my results quite a lot. so thank you. I
installed Panorama Tools, but does anyone know how to morph using
Panorama Tools/PTGUI, as you mention? I just installed the self-
installing package linked to off of Ptgui's website - is that the
version of Panoorama Tools I need? Also, should I be checking off
the shear tick boxes too? I use a 1.6x teleconverter with my 24 mm
tilt/shift lens, because Canon doesn't make a 35mm tilt/shift, and I
like that focal perspective. Doesn't anyone else do that too? I
don't use focus bracketing, because there's often some movement of the
subject, like blowing flowers, so that's why I use tilt-shift lenses
so much. Thank you for your help,
Johnathan Esper
www.wildernessphotographs.com
John Houghton
> I'm a bit confused still where the light
> rays come together in a single point - is that the nodal point or
> entrance pupil?
Only those light rays directed at the entrance pupil can make it
through the lens system. So that is where they necessarily cross over
or come together.
The nodal points are defined thus: In any system of lenses, there are
two points called nodal points such that a ray incident towards one
nodal point will emerge as from the other nodal point in a direction
parallel to the original direction. See diagram:
http://www.johnhpanos.com/nodal_pt.gif
Note that the rays need not physically pass through the nodal points.
Incident light rays parallel to the lens axis pass through the focal
point and don't pass through either nodal point neither do they appear
to do so. So to speak of the nodal point being where the light rays
come together in a single point is just waffle.
> does anyone know how to morph using
> Panorama Tools/PTGUI, as you mention?
Morphing: It's generally used to correct stitching errors caused by
parallax or lens distortions. A network of triangles based on the
control points is transformed to bring all the morphed points into
exact alignment.
You must first select PTStitcher on the Create Panorama tab. On the
Panorama Settings tab, you'll then have the Morph-to-fit option list.
Choose either to morph all points or just a selection of them. If you
choose the option to morph only a selection, you pick the ones to
morph by right clicking the points on the Control Points tab and
checking the Morph option that is now available.
You would normally want to exclude the points to be morphed from the
optimization process since including points affected by parallax is
liable to distort the general alignment of images.
It's best to do a few tests using one seam at which you have a
parallax problem. You need to assign several points in the area
affected, and also around it to avoid unwelcome distortion in areas
remote from the parallax error.
While you can get good results using morphing, it may be easier and
more controllable to use the morphing features now available in
Photoshop that weren't available when morphing was added to Panorama
Tools.
> I just installed the self-installing package linked to off of Ptgui's website - is
> that the version of Panoorama Tools I need?
Possibly - a quick search failed to find the link. I use the package
at http://photocreations.ca/panotools.
> Also, should I be checking off the shear tick boxes too?
Only if you're scanning film images, which might give rise to shear
distortion.
text -
Kevin Wilton
make it
through the lens system. So that is where they necessarily
cross over or come together.
K: This is true, no question.
J: The nodal points are defined thus: In any system of
lenses, there are two points called nodal points such that a
ray incident towards one nodal point will emerge as from the
other nodal point in a direction parallel to the original
direction. See diagram:
http://www.johnhpanos.com/nodal_pt.gif
K: If only there were two rays of light to make up an image,
life would be so simple and this would be exactly the case.
However, light from all parts of the object pass through the
all parts of the lens to make up all of the final image.
Therefore, light from all parts of the object DO pass
through the nodal point, even though they pass through all
of the other parts of the lens as well. I know this is
pedantry but if we are going to the Nth degree to be totally
correct, then we should be totally correct.
J: Note that the rays need not physically pass through the
nodal points.
Incident light rays parallel to the lens axis pass through
the focal
point and don't pass through either nodal point neither do
they appear to do so. So to speak of the nodal point being
where the light rays come together in a single point is just
waffle.
K: Hmmm... Waffle.
John Houghton
> However, light from all parts of the object pass through the
> all parts of the lens to make up all of the final image.
>
obstructions caused by the aperture and other parts of the lens
housing.
> Therefore, light from all parts of the object DO pass
> through the nodal point, even though they pass through all
> of the other parts of the lens as well.
A few of the light rays undoubtedly do pass through the nodal points,
but by no means all. The point is that ALL light rays must pass
through the entrance pupil if they are to contribute to the final
image. That is why it becomes the viewpoint from which the scene is
viewed by the camera - not the nodal point, front or rear.
John
Kevin Wilton
Sigh, when you get a bit between your teeth you just don't
know when to let up.
Light from the object reaches all parts of the lends,
therefore, at wide open, all of the lens is responsible for
all of the image. And, as we have shutters, we selectively
block SOME of the light passing through the glass to the
sensor. HOWEVER, regardless of the amount of glass we use,
all of the REMAINING lens is still responsible for all of
the image. Therefore, light from all of the image must pass
through the nodal point.
You're picking on the wrong one here J...
Kev
John Houghton
> Therefore, light from all of the image must pass
> through the nodal point.
Kevin, This may be true, but as I have already pointed out, not all
light rays (from the subject) pass through the nodal point. A light
ray that enters the lens parallel to the lens axis does NOT pass
through the nodal point. OTOH, all rays that pass through the lens do
pass through the aperture, and for the rays to pass through the
aperture they must be directed at the entrance pupil. So all those
rays that will contribute to the final image will converge onto the
entrance pupil, which obviously becomes the viewpoint of the scene.
Have you performed the tests that I suggested you perform to measure
the positions of the nodal points, entrance pupil and NPP of a zoom
lens?
John
Kevin Wilton
You really are becoming confused here and just because you
point something out doesn't make it correct, especially in
this case.
Light radiates from the object in ALL directions. If it did
not, then there would be points where you could stand and
you would see nothing, just blackness, no light coming to
you from the parts of the object. Fundamental.
Therefore, if light does radiate in every possible direction
from every possible part of the object, yes even shiny
chrome, then it follows that every part of the object must
illuminate every part of the lens. Just as it illuminates
you, the camera, your tripod and the wall behind.
If that is the case, then light from every single, tiny,
minute part of the object MUST also pass through the Nodal
Point, just as light from all parts of the object visible to
the lens also pass through points just next to the Nodal
point. And a little further away in any direction until you
have covered the entire lens with light.
So, it follows that if every point of the whole lens sees
light from the whole of the visible object, then light from
every part of the object must pass through, when allowed by
the shutter, every part of the lens. Therefore, that MUST
include the Nodal point. It's not hard.
If light was only coming at the lens in parallel lines, like
your diagram then, as you stopped down you would vignette
the corners until there was nothing left but you don't. Even
stopped down to f/64 or f/90, a lens is still capable of
resolving the whole image. If this is the case, then light
from the extremes of the object, whatever it is and wherever
it is, must be passing through the centre of the lens.
Take this to the Nth extreme and look at a pinhole camera
image, it's not a dot of light for an image now is it.
Granted there is no lens involved but the pinhole is acting
as the shutter would, a tiny hole at the centre of where the
lens would be, allows light through that still manages to
paint the whole image.
What you are arguing is that only light along the centre
axis passes through the Nodal point, the rest of the light
from an object comes in parallel to that, touching all over
the surface of the lens but not passing through the nodal
point. This is plainly not correct.
K
John Houghton
in going over it again here. Do take the trouble to read it:
http://www.janrik.net/PanoPostings/NoParallaxPoint/TheoryOfTheNoParallaxPoint.pdf
John