Calculating the perfect nodal point
Dimitri
I was just thinking... since PTGui is calculating distortions and
figuring out how to move things around so that everything's aligned,
cound it possibly tell me in some way how much my camera's mounting
position is off and in what direction?
Yes, I know that there are plenty of ways to get "good enough" by
looking through the lens, but I would love to get a mathematically-
derived "move body/lens forward 1mm."
Thanks!
Dimitri Katsaros
Erik Krause
First of all the nodal point of a lens is not identical with the
no-parallax-point you need to adjust your pano head to for least parallax.
Second we can safely say it's impossible for a fisheye lens, since
fisheyes don't have a single no-parallax-point. And I doubt it's
possible for rectilinear lenses, since there are two directions in which
the NPP can be off (f.e. back or left) causing a misalignment in only
one direction in the image. Some of the information is lost because an
image is projected from the 3-dimensional world on a 2-dimensional surface.
Perhaps it's possible to do some calculations if the optical axis and
the rotational axis are already aligned and the camera is off front or
back only. In this case the calculations on
http://wiki.panotools.org/Parallax might be of some help.
best regards
--
Erik Krause
Dimitri
> First of all the nodal point of a lens is not identical with the
> no-parallax-point you need to adjust your pano head to for least parallax.
Okay, so I used the wrong terminology, but it seems you got the jist
of my question. :-)
> Second we can safely say it's impossible for a fisheye lens, since
> fisheyes don't have a single no-parallax-point. And I doubt it's
> possible for rectilinear lenses, since there are two directions in which
> the NPP can be off (f.e. back or left) causing a misalignment in only
> one direction in the image. Some of the information is lost because an
> image is projected from the 3-dimensional world on a 2-dimensional surface.
Ah, but in the context of all the images taken, why shouldn't it be
possible... changes in object placements within a row would indicate
adjustments in one axis, and a column in the other, no? If you don't
see how it can be done, that's okay, but I don't think I'm a
dunderhead for asking the question :-)
Thanks
Dimitri Katsaros
Erik Krause
>> Second we can safely say it's impossible for a fisheye lens, since
>> fisheyes don't have a single no-parallax-point. And I doubt it's
>> possible for rectilinear lenses, since there are two directions in
>> which the NPP can be off (f.e. back or left) causing a misalignment
>> in only one direction in the image. Some of the information is lost
>> because an image is projected from the 3-dimensional world on a
>> 2-dimensional surface.
> Ah, but in the context of all the images taken, why shouldn't it be
> possible... changes in object placements within a row would indicate
> adjustments in one axis, and a column in the other, no?
Unfortunately not. In case of images in a row you can determine how many
millimeters the rotation point is off horizontally. But this can be a
wrong front-back adjustment on the panohead tilt arm as well as a
right-left adjustment in the pan arm as well as both.
Same applies to images in a column: The difference can be back-front or
up-down or both.
In case of a single row or a single column panorama you have information
for one direction where you would need two, in case of a multi row
panorama you have information for two directions but you would need three.
> If you don't see how it can be done, that's okay, but I don't think
> I'm a dunderhead for asking the question:-)
This wasn't my intention and I'm sorry if you got this impression.
However, it is not that I can't see how it is done. I think that it
can't be done.
--
Erik Krause
Dimitri
What if a camera with live view was tethered to a laptop and the
parallax could be "seen" in real time... would that work? I realize
that this would involve another program/application entirely, but what
with automated object recognition and tracking software that already
exists (it's even a feature of VLC videolan player) it might be
achievable in that instance. If one also needed precise camera
position as well, I could see reading that info from the servos of
some of the pano-robots such as pixorb or rodeon.
Does this make sense or am I still dreaming?
Dimitri Katsaros
Kevin Wilton
All lenses have a front and rear nodal point, it's just
physics. And, furthermore, you can evidence the front one
with a pencil and two light stands.
If you want me to explain this off board, I would be happy
to, there are a lot of nodal point Luddites in here.
Kevin
Erik Krause
> All lenses have a front and rear nodal point, it's just
> physics. And, furthermore, you can evidence the front one
> with a pencil and two light stands.
If your technique helps to minimize parallax you certainly find the
center of the entrance pupil. "Nodal point" is a geometric optics term
which refers to a different point but which was used falsely for the
no-parallax-point for a long time. We should be clear in what we talk
about if we want to educate...
> If you want me to explain this off board, I would be happy
> to, there are a lot of nodal point Luddites in here
Not only in here :-)
See:
http://en.wikipedia.org/wiki/Cardinal_point_%28optics%29#Nodal_points
http://tinyurl.com/2qwm5
http://en.wikipedia.org/wiki/Entrance_pupil
http://tinyurl.com/d29lu
http://doug.kerr.home.att.net/pumpkin/Pivot_Point.pdf
http://toothwalker.org/optics/misconceptions.html#m6
http://www.outline.be/quicktime/tuto/
--
Erik Krause
Kevin Wilton
No need to contemplate parallax if you rotate around a
single point. Why is that so hard?
K
PS, Thanks for the links, and they are for...?
michael crane
Ken Warner
together, there are no parallax caused stitching errors --
what have I adjusted? The Nodal Point or the Entrance Pupil
or the No-Parallax-Point?
And should I really care what to call it if the pano stitches?
Kevin Wilton
kings clothes are anything other than what they are, nudity.
You work with what you work with, I shall work with what I
work with, but I don't seem to have have half as many issues
as you do.
Kevin Wilton
Erik Krause
> No need to contemplate parallax if you rotate around a
> single point. Why is that so hard?
Sorry, I don't understand.
--
Erik Krause
Erik Krause
> And should I really care what to call it if the pano stitches?
No, there is no need to care about if the pano stitches. But it is
necessary to use correct wording if you talk to others. You are always
free to call a chair "table" or a table "door" but you might be
misunderstood by listeners not used to your argot...
--
Erik Krause
Kevin Wilton
You need to read around the nodal point and get off my case
until you are sure, I am positive you could Google
something. Unless you would prefer me to explain?
K
Erik Krause
Please explain. I'm very curious...
--
Erik Krause
Kevin Wilton
Exactly, so why do you find it so hard to eat that parallax
whatever when it's the nodal point that concerns us, as
panoramists.
Don't trip over the big words you found on someone else's
blog. The science here is really, really simple, and if I
prefer to explain that to someone off-group so I/we don't
have to trawl through this dirge every time someone mentions
nodal point, then please allow me to do so.
If you prefer to call a table a chair, please also feel
free.
K
Kevin Wilton
I would be happy to, both off group and tomorrow as I am
also playing poker on line and have had a few too many.
K
John Houghton
> The nodal point.
Kevin, No it isn't. Have you taken the trouble to measure the
positions of the nodal points in one of your lenses using an
established technique such as the nodal slide? Please refer to this
document, for example:
I measured a Canon 70-200mm f/4 lens, using my NN5 as a nodal slide
this morning. At 200mm, the positions of the two nodal points were
both near to the front of the lens, about 1cm apart. (One of them is
easy as it has to be 200mm in front of the sensor plane). The no-
parallax point established by taking photos of near and far objects is
close to the lens mount - around 15cm away from the nodal points. The
entrance pupil position can be judged simply by looking into the lens
at the virtual image of the aperture. It is clearly located near the
lens mount.
Please do such a test yourself and see what your conclusion you come
to. A rectilinear zoom lens at the long setting is best for such a
test as the points are usually widely separated so that there's no
doubt about which points match up with the no-parallax point.
John
Keith Martin
>I am positive you could Google something.
Heh. You trust the Interwebs? ;-)
Seriously though, the term "No-Parallax-Point" is UNARGUABLY correct,
as it describes the point at which you get no parallax.
Additionally, I don't see how questions about the validity of "nodal
point" can be dismissed out of hand when there are two nodal points
to consider; front and rear. Surely only one point can possibly
produce no parallax in overlapping shots?
Could it be that the no-parallax-point and whichever nodal point you
use be coincident in your lens?
k
Erik Krause
> Exactly, so why do you find it so hard to eat that parallax
> whatever when it's the nodal point that concerns us, as
> panoramists.
You gave no evidence for your claim so far. I suggest you read the
supplied sources, especially the paper by Rik Littlefield. Otherwise I
must assume you are simply trolling.
--
Erik Krause
Kevin Wilton
That presumes I am after something?
Kevin Wilton
K
Heh. You trust the Interwebs? ;-)
I only trust what I know and trust, as would anyone, anyone
can upload anything to the web, I trust facts. I will read,
and have read most of the links Eric put up. While they make
sense, look florid and have complex equations, they are not
necessarily correct in this case.
Seriously though, the term "No-Parallax-Point" is UNARGUABLY
correct, as it describes the point at which you get no
parallax.
You only get parallax from TWO points, not one. Therefore,
the Nodal Point, being singular wins the day.
Additionally, I don't see how questions about the validity
of "nodal
point" can be dismissed out of hand when there are two nodal
points
to consider; front and rear. Surely only one point can
possibly
produce no parallax in overlapping shots?
Exactly right my friend!!!! The front one.
Could it be that the no-parallax-point and whichever nodal
point you
use be coincident in your lens?
Possibly, but equally a bus, train or car could get me to
work although coincidentally, only one is a car.
K
K
Keith Martin
>You only get parallax from TWO points, not one. Therefore,
>the Nodal Point, being singular wins the day.
In the English language, if there is a point that produces no
parallax then it IS, quite literally, the no parallax point. This is
a descriptive name that is either correct or shouldn't have that name
in the first place.
My, er, 'point' is this: whatEVER point in a lens produces no
parallax, whether that sits on a nodal point or not, that must be the
no-parallax-point. Ergo sum. Or better, quod erat demonstrandum.
BTW, I'm having trouble grokking "you only get parallax from two
points". I'm talking about NO parallax, not parallax. If I wanted to
describe positions of parallax in a lens I'd be discussing lines
rather than a point. Well, I guess I'm just a square. (Nods towards
Mr Abbott. ;-)
k
Kevin Wilton
Like I said earlier, I'm doing this and playing poker, as
well as being three sheets to the wind, let's continue
tomorrow?
K
Keith Martin
k
Dimitri
Oh My GOD
I am almost sorry I asked my question in the first place....
Look, perhaps I should have avoided terms like "nodal," I'm not
sure... what I am sure of is the core of my question:
Can anyone think of a way of semi-automating the placement of the
camera in such a way that is optimal for stitching? [e.g., something
tells the photographer to move the mount forward or back and/or (in
the case of multi-row) up or down]
Anything else is just off topic and a waste of time.
Respectfully
Dimitri Katsaros
Joergen Geerds
I think this PDF explains getting the no-parallax-point best:
http://www.outline.be/quicktime/tuto/TheGrid.pdf
Once you have found the point for your specific lens and focal length
(if you are using a zoom), make yourself a field table with the exact
values.
question for erik: does the no-parallax-point move when the focus
distance changes (for regular, non-fisheye lenses)?
joergen
Hans
Oh yes.
Unless you have an sight disability your eyes can do it in 60 seconds,
I doubt ant software can automate it faster.
To find the NPP alias the entrance point of any lens is really so
simple that thinking of making it automated is just laziness and lack
of basic knowledge.
There are loads of simple tutorials around.
Check NodalNinja site, John's site http://www.johnhpanos.com/epcalib.htm,
http://www.vrphotography.com/data/pages/techtutorials/technotes/nodalptalign-tn.html,
Rosauros site http://www.rosaurophotography.com/html/technical7.html
and many others.
Maybe I will even publish one soon as people seems not to understand
how simple it is.
Hans Nyberg
www.panoramas.dk
Kevin Wilton
question for erik: does the no-parallax-point move when the
focus
distance changes (for regular, non-fisheye lenses)?
No it does not because the same rules as Macro apply.
Therefore, if you are closer than 9 focal lengths, then you
have to take it into account.
However, minutest changes in the focal length of a zoom do
have effects that accumulate with the increase in the amount
of zoom.
Kev
Joergen Geerds
> No it does not because the same rules as Macro apply.
> Therefore, if you are closer than 9 focal lengths, then you
> have to take it into account.
affects in my work, but I just wanted to be sure.
> However, minutest changes in the focal length of a zoom do
> have effects that accumulate with the increase in the amount
> of zoom.
Yes, that's why I am such a fan of old manual lenses for pano work...
they are usually so stiff, that they don't have any zoom or focus
creep. on the other hand, i think i found the worst lens for pano
work: the canon 50mm f1.8: totally useless, the slightest wind will
push the front element, and bring the lens out of focus, and gravity
is even worse...
joergen
Kevin Wilton
of the
camera in such a way that is optimal for stitching? [e.g.,
something
tells the photographer to move the mount forward or back
and/or (in
Dimitri
I have spent the past hour reading a whole pile of links
that are extremely long winded and verbose, some don't even
seen to be correct. However, that doesn't matter, and
nomenclature doesn't matter either, just so long as you can
find the elusive point about which you wish to rotate your
lens. That, my friend, is simple, thank goodness.
No need to set up much more than a couple of light stands
and your tripod, and you only need to do this once, for each
focal length of your zoom and for each lens that you plan to
use, then scribe those marks on your pano head or make a
table of your own.
Simply set up your camera tripod and with your eye and a
piece of string, find the centre of the lens and place it
exactly over the centre of the tripod. You need to rotate
around the centre of the tripod in two planes, this is the
first.
Now set up a light stand so that one of the thick bottom
sections is in front of the camera.
Now place the second stand in between the camera and the
first stand with a thin top section at the same level as the
first thicker one. On top of the second stand you must
attach, with tape, a pen or pencil. If your stand is black,
use a white one.
The second stand must be close enough for you to reach when
you are looking through the camera, so you can make slight
adjustments and the far one must be close enough for you to
see both sides around the pencil. Usually the far one is
about 1.2M away and the second just in front.
When both are in perfect alignment and centred in the lens,
move the camera to the left or right and the two should not
move in relation to one another. If you have found this
magical point in space, they do not.
However, should there be any displacement; these are the two
simple rules to follow:
1) if you find are BEHIND this mysterious point then, when
you spin to the right, the amount of visible rear light
stand to the right of the pen grows while that on the left
shrinks; therefore you have to slide the camera forward to
correct this. Don't make timid movements or you'll be here
all day. Make bold adjustments and refine them on your 3rd
and 4th goes as you straddle the point in smaller steps.
2) if you are in FRONT of the point we do not mention, when
you spin to the right, the visible rear light stand to the
left of the pen grows while that on the right shrinks;
therefore you have to slide the camera backward.
Do both of these, forward and back until there is no
movement of the stands.
However, if you move to the left and then to the right and
the rear light stand appears to move by DIFFERING amounts,
you are not centred over the tripod and must address this
FIRST.
Total time to do this, 5 minutes.
I have pictures if you need them but you will have to email
me direct.
Kevin
John Houghton
I've know idea why people continue to tell you how to set up the
panohead manually rather than answer your interesting question. We
have no reason to suppose that you don't already know how to do this.
The only semi-automatic element I have tried in an otherwise manual
method is to measure the parallax shift (in pixels) in Photoshop for
two positions of the camera on the upper arm of the pano head that are
clearly on either side of the NPP. Then interpolate the NPP position
from the two scale readings using a straight line graph assumption.
This avoids the nudge it forwards a bit, then back a bit, then back a
bit more, then just a smidgin forwards, .... procedure.
John
Ken Warner
Kevin Wilton
space without tables would be useful for the days when you
don't have them, the dog ate them, your assistant left them
on the last job or they are in your other trousers, don't
you think?
John Houghton
> That sounds like it should work. Does it?
Yes, it works. I tried plotting a number of settings at 1cm intervals
to see how straight the graph would be and it looks reasonably
straight. (It's for a Tokina 10-17mm fisheye @10mm on a 40D). I
wouldn't actually do it this way, as it's vastly quicker and more
accurate to observe the movement of the entrance pupil directly in
this case.
John
John Houghton
> I tried plotting a number of settings at 1cm intervals
Sorry, I forgot the link to the graph (oh for a message edit option):
http://www.johnhpanos.com/npp-graf.gif
John
Erik Krause
> Yes, it works. I tried plotting a number of settings at 1cm intervals
> to see how straight the graph would be and it looks reasonably
> straight. (It's for a Tokina 10-17mm fisheye @10mm on a 40D). I
> wouldn't actually do it this way, as it's vastly quicker and more
> accurate to observe the movement of the entrance pupil directly in
> this case.
How do you determine whether the point is off in a back-front
direction or sideways? If the lens optical axis already hits the
rotation axis there is no problem...
best regards
Erik Krause
http://www.erik-krause.de
Erik Krause
> question for erik: does the no-parallax-point move when the focus
> distance changes (for regular, non-fisheye lenses)?
It might, depending on the lens. Usually the whole lens or the front
element is moved relative to the camera when focusing. Since the NPP
is a property of the lens it moves with it. There is less movement
necessary for a wide angle lens but very much for a telephoto one.
For a 50mm lens it is 8mm between infinity and 0.45m close limit, for
a 400mm lens it is 100mm at 2m close limit.
This might cause problems for focus stacking where there is very near
foreground.
However, some lenses move a rear or inner element to focus in which
case the NPP won't move (at least if moving element is behind the
aperture). But those lenses most likely change focal length during
focusing.
John Houghton
> How do you determine whether the point is off in a back-front
> direction or sideways? If the lens optical axis already hits the
> rotation axis there is no problem...
>
Lateral and vertical positioning is done separately.
John
Ken Warner
and cameras without an optical through the lens viewfinder.
The way I do it with my compact is the way you described earlier -- lots
of fiddling.
Dimitri
> Oh yes.
> Unless you have an sight disability your eyes can do it in 60 seconds,
> I doubt ant software can automate it faster.
True
I have been doing it myself for some time... I'm looking for an aid
that will get me to the EXACT point, not approximate. While pretty
much academic when shooting panos that are distant, the closer an
object is the more apparent even the slightest parallax is. (e.g.,
macro panoramas.)
Thank you
Dimitri Katsaros
michael crane
> Convincing graph. Seems like that would be very handy for compacts
> and cameras without an optical through the lens viewfinder.
good point
michael crane
where the light was coming from ?
mick
Dimitri
thing I can think of for macro applications would be to print a grid
to a thin glass plate, place it over a scene and use your photoshop
technique.
Does this sound plausible?
Thanks
Dimitri Katsaros
Uri Cogan
> I have been doing it myself for some time... I'm looking for an aid
> that will get me to the EXACT point, not approximate.
>
I don't have a 'calculated' method, but for what it's worth here is my
way of finding the NPP.
It helps to have the camera on a tripod, and connected to a computer
running some camera control software. I use Nikon Capture.
1. point the camera down (using the pano head) and shoot. Open the image
in Photoshop.
2. Adjust the camera's position on the lower rail until the centre of
the rotator appears EXACTLY at the centre of the frame as measured (by
the pixel) from the bottom of the frame, near the base of the camera.
3. Position a vertical needle about 30 cm in front of the lens, with a
detailed scene at least 10 meters away.
4. Shoot two images, one with the needle on the extreme left side of the
frame and then on the right side. Compare the relationship of the needle
to the background scene in both images.
5. Adjust the position of the camera on the upper rail until there is no
difference between the needle/background relationship.
6. Write the numbers down. Done.
--
Uri Cogan
http://www.uricogan.com
*** My Karma Ran Over My Dogma ***
Erik Krause
> just to muddy the water I understood that the npp varied according to
> where the light was coming from ?
That applies to fisheyes only. Their NPP moves depending on the angle
to the optical axis.
See http://tinyurl.com/y8wco5e for details.
Rectilinear lenses should have a single NPP which doesn't move.
Erik Krause
> 1. point the camera down (using the pano head) and shoot. Open the image in
> Photoshop.
>
> 2. Adjust the camera's position on the lower rail until the centre of
> the rotator appears EXACTLY at the centre of the frame as measured (by
> the pixel) from the bottom of the frame, near the base of the camera.
A mechanical and possibly faster way (only applicable if the lens
housing is round and symmetrical): Point the camera exactly down.
Lower the tripod until the front element of the lens is as high as
your table. Place the tripod near the table and place a book on the
table such that one of its corners touches the lens housing. Turn the
pano head 180° and spot the gap between the book corner and the lens.
Adjust the rail such that the book touches the lens slightly all the
180° round...
Keith Martin
>Their NPP moves depending on the angle
>to the optical axis.
Fortunately (if I may generalise somewhat) the overlaps we use tend
to be in the same areas of shots, which seems to reduce the problems
this could cause.
k
Serge Maandag
>> I tried plotting a number of settings at 1cm intervals
> Sorry, I forgot the link to the graph (oh for a message edit option):
> http://www.johnhpanos.com/npp-graf.gif
John, that is an oddly perfect graph :)
I performed sort of the same thing once. I set up 2 objects behind each
other some 20cm in front of my fisheye lens. I measured the parallax
difference in pixels between a shot where the objects were at -30 degrees
yaw and at +30 degrees yaw. I shifted the lens forward in steps, each time
measuring the parallax again. I plotted the results in a graph, parallax in
pixels as a result of the forward position of the lens.
My results where sort of linear, but the line was steeper around the NPP.
So a tiny offset from the NPP caused a relatively big parallax result.
Shifting further from the NPP had a less dramatic effect.
Regards,
Serge.
Uri Cogan
> A mechanical and possibly faster way (only applicable if the lens
> housing is round and symmetrical): Point the camera exactly down.
> Lower the tripod until the front element of the lens is as high as
> your table. Place the tripod near the table and place a book on the
> table such that one of its corners touches the lens housing. Turn the
> pano head 180� and spot the gap between the book corner and the lens.
> Adjust the rail such that the book touches the lens slightly all the
> 180� round...
>
Yes, this will mechanically align the body of the camera to the
panohead, but will not compensate for sensor displacement within the camera.
Serge Maandag
Keith,
It does, but only partially, I've written a page about it in the past:
http://360photo.org/57.html
Regards,
Serge.
Erik Krause
> >Their NPP moves depending on the angle
> >to the optical axis.
>
> Fortunately (if I may generalise somewhat) the overlaps we use tend
> to be in the same areas of shots, which seems to reduce the problems
> this could cause.
Hmmm, overlap is measured along the horizon only. If you look at the
usual seam lines above or below the horizon, it gets worse. See here:
http://360photo.org/57.html
Kevin Wilton
according to where the light was coming from ?
mick
Call it what you like, the point around which you rotate is
a product of the optics of the lens and not its
surroundings. It doesn't change.
Erik Krause
> > A mechanical and possibly faster way (only applicable if the lens
> > housing is round and symmetrical): Point the camera exactly down.
> > Lower the tripod until the front element of the lens is as high as
> > your table. Place the tripod near the table and place a book on the
> > table such that one of its corners touches the lens housing. Turn the
> > pano head 180° and spot the gap between the book corner and the lens.
> > Adjust the rail such that the book touches the lens slightly all the
> > 180° round...
> >
>
> Yes, this will mechanically align the body of the camera to the
> panohead, but will not compensate for sensor displacement within the camera..
It will align the lens - and hence the optical axis - to the panohead
(always given the lens housing is rotational symmetric to the optical
axis).
You won't want to align the sensor center to the panohead. If the
optical axis doesn't hit the sensor center you have a NPP
displacement again. You compensate for sensor shift with the shift
parameters d and e or with proper cropping (in case of a circular or
semi-circular fisheye).
Erik Krause
> Call it what you like, the point around which you rotate is
> a product of the optics of the lens and not its
> surroundings. It doesn't change.
You can measure it. Try it yourself or read michel thoby's page:
http://tinyurl.com/y8wco5e
John Houghton
shift in the images resulting from a given shift in the entrance pupil
would gradually increase as the camera was gradually moved closer to
the near object. In my case, I had the near object at a little over
1m away from the camera with the background a long, long way away. The
departure from linearity would be much less than for your placement at
only 20cm.
John
360 Precision
far more difficult then it needs to be. To arrive at a usable setting
should take roughly 60 seconds. If you then need to fine tune the
measurement it should take no longer than 5-10 minutes to get to with
0.5mm.
Most of the problems arise from people not understanding what errors
are actually caused by an incorrect entrance pupil. I've looking at
hundreds of PTGui projects over the past 5-6 years for customers and
maybe 1% have issues with incorrect EP settings. Most visible
stitching errors are caused by:
1. Bumping the tripod
2. Changing the lens zoom.
3. Bumping the tripod
4. Enblend issues or other blending options
5. Poor optimisation procedures
6. Bad control points
7. I'm sure there are others
Quite often people won't even try to stitch a set of images if they
think they've used the wrong settings. Most of time the pano will
stitch regardless. If it won't stitch automatically there's a few
simple procedures to follow:
1. Set control points equidistant from the lens in adjoining images.
2. Don't set control points where foreground objects meet background
objects. This is the only ares in which entrance pupil issues will be
visible.
These two simple steps will cure most problems.
Personally I don't care how people term the setting but there are
some reasons why nodal point is wrong.
1. All lenses have a front and rear nodal point. If you use either of
these settings you will most likely still have parallax issues
2. If the nodal point and entrance pupil are the same then it's
purely coincidental and can't imagine it's possible anyway
I'm guilty of calling it nodal point as well but just because it's
the most widely used term does not make it right.
Yes, the entrance pupil can change with focus and aperture but the
change is so minute these issues become irrelevant. Like they say,
"don't sweat the small stuff"
Matt
> --
> You received this message because you are subscribed to the Google
> Groups "PTGui" group.
> To post to this group, send email to pt...@googlegroups.com
> To unsubscribe from this group, send email to ptgui+un...@googlegroups.com
> Please do not add attachments to your posts; instead you may upload
> files at
> http://groups.google.com/group/ptgui/files
> For more options, visit this group at http://groups.google.com/group/ptgui
360 Precision
example will be different compared to images taken every 60º. This
change however is COMPLETELY IRRELEVANT unless for some bizarre reason
you shoot different images with a mixture of 60 and 90º increments.
No idea why anyone would ever do this though ????
Matt
Erik Krause
> The entrance pupil when shooting every 90º with a Sigma 8mm for
> example will be different compared to images taken every 60º. This
> change however is COMPLETELY IRRELEVANT unless for some bizarre reason
> you shoot different images with a mixture of 60 and 90º increments.
Sorry, your'e wrong. Read http://360photo.org/57.html
Matthew Rogers
Half of what Serge says on the link below again is irrelevant and will never show errors in any NORMAL panorama. You'd have to set-up a very specific scene for his assumptions to have any real effect on the output. Or unless of course you're trying to stitch together eight nicely coloured images created in photoshop which of course is pointless.
What this little experiment completely fails to take into consideration is where foreground versus background objects in a scene typical coincide. Choosing the 45º, well really 90º option for the coloured images based on the zenith and nadir makes no sense in reality. I once stitched over 1,000 panoramas for a customer with the wrong entrance pupil and errors between the zenith/nadir and the horizontal images were practically non-existant. The setting for the horizontal images have far more weighting in an error free output.
This to me looks like a someone trying to find a solution to a problem that simply doesn't exist. If it does for Serge then there's something wrong somewhere else with his set-up. Maybe because it's on the internet then it must be right :(
I guess my real world testing of almost 400 different camera and lens combinations bears no relevance to the discussion.
Matt
http://www.360precision.com
Jeffrey Martin
feel free to strike up a friendlier tone, otherwise your comments are
not welcome (with me anyway).
Jeffrey
On Jan 3, 1:39 pm, "Kevin Wilton" <wil...@eeek.tv> wrote:
> Trolling?
>
> That presumes I am after something?
John Houghton
> Trolling?
The Wiktionary offers this definition of troll:
"A person who posts to a newsgroup, bulletin board, etc., in a way
intended to anger other posters and to cause drama, or otherwise
disrupt the group's intended purpose".
John
Kevin Wilton
John
I see, I was just wondering what you were up to.
Serge Maandag
> Half of what Serge says on the link below again is irrelevant and will
> never show errors in any NORMAL panorama. You'd have to set-up a very
> specific scene for his assumptions to have any real effect on the output.
> Or unless of course you're trying to stitch together eight nicely
coloured
> images created in photoshop which of course is pointless.
Matthew,
I agree that there are far more important factors to consider, changing
light due to cloud movement and kicking the tripod legs have done much more
damage to my panoramas than the effect of entry angle dependent NPP
location. But the effect is not 0. It is obviously small enough for you to
dismiss it, but that was not my intention when I wrote the page, nor was it
probably Michels when he did his measurements.
It's a theoretical discussion. The art of trying to get it perfect, whether
that's relevant or not and whether that's feasible or not.
But it's clear that you (and Hans for that matter) are much rather
practical.
For me: I do see the effect. The zenith image never fits as nicely as my
other images. That means my nodal ninja is flaky or I am indeed suffering
from NPP displacement for the zenith image.
Serge.
Jeffrey Martin
conversation is "perfect".
I got 150 panoramas last summer in a couple days (timisoara panotools
meeting, some of you saw me, and some of you saw the nice results)
which were processed with NO human intervention, all of them handheld,
and most of them turned out "good enough"
example
http://www.360cities.net/image/girls-budapest-keleti-palyaudvar-main-train-station
again, not perfect, but "good enough" for this particular purpose.
(and yes, no human intervention :P
On Jan 3, 2:16 pm, "Kevin Wilton" <wil...@eeek.tv> wrote:
> K
>
> Like I said earlier, I'm doing this and playing poker, as
> well as being three sheets to the wind, let's continue
> tomorrow?
>
> K
Matthew Rogers
The example below is absolutely dreadful, every aspect of it is bad, the colour, CA, sharpness, stitching errors and bizarre HDR effect. I guess out definition of "Good Enough" and "perfect" are miles apart.
The only place this pano would go is the trash can I'm afraid.
Matt
PTGui Support
You're welcome here to discuss panoramic photography, both the practical
and theoretical aspects, but no personal attacks please.
Joost
Dimitri
the calculations done by PTGUI, could some mathematical confirmation
be arrived at. I'm not sure how I could have worded my inquiry
differently, but I apologize for its unintended effects. If you know
of a more suitable forum or people I could get in contact with to
discuss, please let me know.
Sincerely
Dimitri
Joergen Geerds
We're especially going to need it when robots take more panoramas, and
they have to self-calibrate themselfes... human intuition and manual
adjustments, as they are standard right now (aka eyeballing), doesn't
do well with machines, so in the long run we would need an algorithm
how to find the no-parallax point.
joergen
michael crane
> Thank you for chiming in, Joost.... All I was looking for was, given
> the calculations done by PTGUI, could some mathematical confirmation
> be arrived at. I'm not sure how I could have worded my inquiry
> differently, but I apologize for its unintended effects. If you know
> of a more suitable forum or people I could get in contact with to
> discuss, please let me know.
LOL
Erik addressed your question fully I think, probably, as you say, you
need to investigate some sort of spatial awareness robotic software or
something if you want to persue it.
regards
mick