Nikos Giannakopoulos
Posts: 137
Location: Athens, Greece
Registered: 10 Jan 2008
|
More parallax error towards to the edges of Sigma 8mm but not to the center
Posted: 13 Aug 2008 at 7:41 GMT
|
|
Hello everyone...
Yesterday I just sat down and tried to recalibrate the nodal point on my Sigma 8mm on 350D, working upon John's tutorial (www.johnhpanos.com/epcalib.htm) and I came up to the conclusion of the thread's title.
Or maybe I'm doing something wrong... anyway I'm just posting the images that shows what I mean, and maybe someone can clear out this phenomenon.
I shot two pictures with the same pitch (0°) but with ongoing yaw +90° as I always shoot 4 images. In front of the camera there is another tripod with a cardboard on top just behind a glass door, where there is also a white tape glued on the glass. As someone can notice between the two frames there is (almost) no parallax (or maybe there is just about 1 pixel or so).
But looking down on the bottom of the tripod being photographed and particularly at the clamp, there is noticeable parallax of more than 1 pixel.
yaw 0°
yaw 90°
Furthermore, I shot two more frames with the same yaw (0°) but with ongoing pitch -90°
pitch 0°
pitch -90°
As someone can see in this example the line made with a marker on the tape which is glued on the glass, doesn't align with the joints on the floor of the balcony. I guess the same will apply for a nadir shot.
Have I not set up the nodal point correctly and the camera must be moved on the horizontal or vertical arm of the head rotator, or this is just "normal" for fisheye lenses? I also think that this phenomenon tends to produce more stitching errors with PTGUI near nadir and / or zenith, and some times it needs a bit more effort with manual control points.
There are also times when auto align and optimizing produces a "too good to be true" result without any other effort while with the same setup/ head calibration on another location it produces a "good" result, with visible stitching errors.
I guess Fish eye lens are a bit of mystery as the entrance pupil is always changing and maybe it is the same on a circular direction towards the center of the lens.
Thanx in advance for any info you may have
Friendly Nikos
|
|
alert moderator
|
|
DorinDXN
Posts: 2853
Location: Timisoara, Romania
Registered: 14 Nov 2006
|
|
Hans Nyberg
Posts: 2791
Location: Denmark
Registered: 28 Aug 2005
|
Re: More parallax error towards to the edges of Sigma 8mm but not to the center
Posted: 13 Aug 2008 at 8:36 GMT
|
|
Yes this is a well known issue for fisheye lenses.
It is why we say the nodalpoint is floating.
Here is my page about it.
www.panoramas.dk/panorama/fisheye-NPP/parallax.ht...
The difference for the mounting can be as much as 8mm.
Meaning you will have no parallax at centre with a setting of 89mm on the upper arm but to have no parallax at the zenith and nadir you have to use 97
Remember that the more images around you take the larger this error becomes. 6 around with a Sigma 8mm is a bad idea as the centre blending is now closer to the centre of the image with a larger change in parallax in relation to the zenith and nadir.
You have to use a medium. The idea that the Nodalpoint is a matter of an exact setting within 1 mm is just not correct.
I have done tests shooting the same panorama while changing the mount from 92mm to 98mm, All of them stitched without problems.
If you also take zenith there is another much more important thing to look for.
Your camera mount on the arm has to be perfect. Not tilted 1 degree up or down on the arm. Small errors for that will give you problems not with the images around but with the zenith.
Hans
|
|
alert moderator
|
|
Roger Hein
Posts: 40
Location: North York, Canada
Registered: 27 Sep 2006
|
Re: More parallax error towards to the edges of Sigma 8mm but not to the center
Posted: 13 Aug 2008 at 19:35 GMT
|
|
Hans Nyberg said: If you also take zenith there is another much more important thing to look for.
Your camera mount on the arm has to be perfect. Not tilted 1 degree up or down on the arm. Small errors for that will give you problems not with the images around but with the zenith.
Hans
Hans - interesting info.
For someone like myself I would normally tilt the camera down -5 or -10 deg for a smaller nadir hole. I also take a zenith shot. Generally this has worked well but once in a while a get a situation where my stitch has a small error or two. It's been easy enough to 'fix' but I'm always looking for ways to improve.
I initially 'found' my 'nodalpoint' with the camera set at 0 deg tilt but from my understanding of reading your info this may not be the best route?
Roger...
|
|
alert moderator
|
|
Hans Nyberg
Posts: 2791
Location: Denmark
Registered: 28 Aug 2005
|
Re: More parallax error towards to the edges of Sigma 8mm but not to the center
Posted: 13 Aug 2008 at 20:12 GMT
|
|
Roger
I think you misunderstand me.
I say "tilted up or down on the arm"
Meaning that it is no longer in the correct NPP.
As far as I know only 360precision has a mounting mechanism for the camera so that you can mount it precisely on the arm each time.
To do this on other panoheads you have to use a spirit level on the arm and on the back of the camera to assure that it is perfectly mounted.
Hans
|
|
alert moderator
|
|
Roger Hein
Posts: 40
Location: North York, Canada
Registered: 27 Sep 2006
|
Re: More parallax error towards to the edges of Sigma 8mm but not to the center
Posted: 13 Aug 2008 at 20:17 GMT
|
|
Hans,
Yes I did! Thanks for the clarification.
Roger...
|
|
alert moderator
|
|
Rosauro
Posts: 300
Location: Toronto, Canada
Registered: 15 Dec 2006
|
Re: More parallax error towards to the edges of Sigma 8mm but not to the center
Posted: 14 Aug 2008 at 6:55 GMT
|
|
Hans Nyberg said: To do this on other panoheads you have to use a spirit level on the arm and on the back of the camera to assure that it is perfectly mounted.
I agree. This is a step most miss or disagree to using and solely rely on the bubble on the base. Most rely on software in post-prod fixes.
|
|
alert moderator
|
|
Nikos Giannakopoulos
Posts: 137
Location: Athens, Greece
Registered: 10 Jan 2008
|
Re: More parallax error towards to the edges of Sigma 8mm but not to the center
Posted: 14 Aug 2008 at 7:32 GMT
updated: 14 Aug 2008 at 8:00 GMT
|
|
Thanx Dorin and Hans for the info you posted.
Generally I think that stitching a spherical panorama with fisheye lens is different each time... and applying a template isn't always working (at least for me so far)... maybe because I'm shooting 4 images with -15° tilt (+ 1 zenith) and maybe I'm not having exactly the same degrees each time, or maybe a minor tripod movement occurred while shooting, etc (the camera is mounted perfectly on the upper arm)
I've tried a template which was "too good to be true" on another new stitch and it had errors... so I had to create CPs from scratch (Auto align and then manual edit).
Every new panorama is a new challenge, and it is very rare that with one click of optimization the result will be "too good 2b true"...
For example I was astonished with the result that PTGUI gave me with one auto align - optimize.
On the following panorama a too good to be true was only with 2 or 3 clicks!!
www.360visits.gr/en/panoinfo.php?cat=85
If you look on the zenith... the electricity wires were an absolute line!! No Stitching errors at all, even though they are so thin!
On the other hand on the next panorama
www.360visits.gr/en/panoinfo.php?cat=42 (Marker 3 - on the bottom of the page)
the electricity wires had errors and it was difficult to add CPs there, so I decided to erase them in Photoshop.
I guess there are always ways to defeat errors, (in PTGUI or Photoshop) and that is what makes each project interesting...
Lately I tried/bought Tufuse Pro and I was very impressed by its results and the more coherent adjustments prior to Fusing the images, so I guess one more thing was added to the interesting procedure.
Nikos
|
|
alert moderator
|
|
back to threads list |
this thread is closed
