PDA

View Full Version : Fisheye NPP changing based on pano spacing?



shidarin
09-27-2012, 05:23 PM
I was reading a review of the popular Samyang Fisheye lens (http://michel.thoby.free.fr/SAMYANG/Early%20test%20report.html), and it started talking about the NPP locations, and provided this image showing different NPP (http://michel.thoby.free.fr/SAMYANG/Samyang_shaved.jpg)s for different 360 pano intervals- a different NPP for 60, 45, and 90.

Everything else I've read talks about each lens having a single NPP- and by the very nature, why would the NPP change if the interval angle changes?

Just looking for some enlightenment, thanks.

Hugh
09-27-2012, 06:22 PM
The NPP (No Parallax Point) can be thought of as the point for rotating the lens round to achieve the "minimum" parallax, so can be thought of as a single point for a lens, but has to be chosen with respect to the Entrance Pupil.
The Entrance Pupil can be a single point for a particular lens, but is more likely to change as the angle of the incoming ray moves away from the principal ray (axis of the lens).
Some examples of the way the Entrance Pupil behaves can be found at:
http://www.hugha.co.uk/NodalPoint/Index.htm
which shows that for a fisheye lense the change can be quite significant.
Also page 8 of:
http://www.hugha.co.uk/Acrobat%20PDF%20Files/Spherical-Panoramas-for-HDS-Point-Clouds.pdf

When using a lens for creating panoramas there is an argument for selecting the Entrance Pupil where the angle of the ray relative to the principal ray is the same as the angle at which successive images are "joined".
e.g. If you take 6 shots round at 60 degree intervals with a Samyang 8mm then you would chose the NPP to be at just under 3mm in front of the gold ring where the rays at 30 degrees each side of the principal ray meet it, but if you take 8 shots round at 45 degree intervals then you might chose an NPP point at about 1.5mm in front of the gold ring, where the rays at 22.5 degrees each side of the principle ray meet it.
The Samyang 8mm has less "spread" of the Entrance Pupil than more "conventional" fisheye lenses such as the Sigma 8mm and Nikon 10.5mm so the change is less noticeable with the Samyang 8mm.

I hope this goes some way in answering your question.

nick fan
09-27-2012, 08:23 PM
I was reading a review of the popular Samyang Fisheye lens (http://michel.thoby.free.fr/SAMYANG/Early%20test%20report.html), and it started talking about the NPP locations, and provided this image showing different NPP (http://michel.thoby.free.fr/SAMYANG/Samyang_shaved.jpg)s for different 360 pano intervals- a different NPP for 60, 45, and 90.

Everything else I've read talks about each lens having a single NPP- and by the very nature, why would the NPP change if the interval angle changes?

Just looking for some enlightenment, thanks.

Michel Thoby is the authoritative source of studies of floating NPP / LPP. Read more on his studies of other lenses.

NIck

shidarin
09-27-2012, 10:07 PM
Thanks guys, that was a really informative info and fits with what I was thinking, but I didn't want to assume or start to piece together information incorrectly.

Armed with this knowledge, I'm off to read more of Michel Thoby's studies.

DennisS
09-28-2012, 06:42 AM
What is really interesting is that a fisheye lens has a different setting depending on where you want to optimize your image. If you remove parallax from the center portion of the panorama, you induce parallax along the top and bottom. I lost the links to pictures that clearly show this. While calibrating for 4 shots around, I can get the center to be perfect but the top and bottom are out. I optimize for the bottom and get acceptable results along the center. Longer focal length lenses do not suffer from this issue.

hindenhaag
09-28-2012, 08:25 AM
This is Michel's article about LPP.

http://michel.thoby.free.fr/Fisheye_history_short/Beyond-the-pupil.html

And this is the link to set up some nice music, some antipasti, and then take your time to read and try to understand the basics of Michel's articles:001_smile:

http://michel.thoby.free.fr/

All of us learned a lot about fisheye lenses use as well as from Hugh's articles. Thx a lot to both of them.

You have to decide how many shots you want to take around with your camera lens combination. Sometimes you can go for 6 or 4 shots around, sometimes for 4 or 3 shots around. With less shots around you will end up with less overlap. But you will be quicker to get your set around. May be important in busy places with a lot of people moving around, in concerts with people moving their arms to the music, journalism etc. The "normal minimum overlap without risk" is 25%. I prefer 30%. May be you have to shoot some more pics, just some seconds to take them, but this might save your time to go back to New York to shoot again and you lost the special moment and atmosphere anyway.

Use Frank's Calculator to check your overlap: http://www.frankvanderpol.nl/fov_pan_calc.htm

Just change the number of pics around in "Calculate Overlap".

Success,
Heinz

kevin6270
10-11-2012, 10:30 AM
Got my shiny new Samyang 8mm and NN4 yesterday. Shaved the lens hood and all ready to go:001_smile:

I'm planning to test a 3 shot pano at 120 deg on a D600, and wondered if anyone knew what the LPP is for the lens at this rotation or better still the NN4 rail settings. I can't quite determine the LPP from Michel's excellent article but it looks like 20mm behind the lens front (or 4mm ahead of the gold ring if I'm reading it correctly ?)

Any help appreciated.

Kevin

John Houghton
10-11-2012, 11:34 AM
Kevin, It's best if you do some tests yourself to determine the optimum settings. It doesn't take long. See:
http://www.panoramas.dk/panorama/Fisheye-NPP/parallax.html
http://www.johnhpanos.com/epcalib.htm
http://www.easypano.com/forum/display_topic_threads.asp?ForumID=1&TopicID=4162&PagePosition=1
.
John

Hugh
10-13-2012, 05:55 AM
Thanks guys, that was a really informative info and fits with what I was thinking, but I didn't want to assume or start to piece together informationincorrectly...
You are welcome shidarin.
I find that knowing about the behavior of the lens allows me to make"informed" decisions about setting up a lens and how many shots to take, but the reality is that we can only have one NNP (Nodal Ninja Point) and the methods suggested by John achieve the same result.
I have published this information on the Internet, because I have not been able to find this representation in photogrammetric journals or the Internet.
The only oasis I have found in this dearth is Michel Thoby (http://michel.thoby.free.fr/SAMYANG/Early%20test%20report.html)'s web site and I consider Michel to be the “guru” on the subject.
My reason for being interested in lens behavior goes back to the 1070s when we used hugely expensive photogrammetric cameras for Close Range (Terrestrial) Photogrammetry, such as the Wild P31, Wild P32 and Zeiss UMK..
These cameras cost something like ten times a Leica M or Hasselblad, so it was desirable to use “off the shelf” cameras, especially in hostile and nasty environments.
The cameras were simple in operation, but the price reflected the “symmetrical” distortion free lenses, a couple of microns in the corners of a 70mm x 70mm format in the Wild P32!
To use “off the shelf” cameras we needed to be able to correct for distortion at a measured point, so we needed to know how the lens behaved.
Although I could not find any information in journals, magazines, etc. I was able to talk to lens designers in Wild Heerbrugg, now Leica Geosystems and learned much from this.

(or 4mm ahead of the gold ring if I'm reading it correctly ?) Any help appreciated

Kevin
From my measurement of a Samyang 8mm Canon fit lens, 4mm in front of the gold ring is about right.
Note that from Michel’s page:
http://michel.thoby.free.fr/SAMYANG/Samyang_shaved.jpg (http://michel.thoby.free.fr/SAMYANG/Samyang_shaved.jpg)
that the position depends on whether you take 6 shots round at 60° or 8 shots round at 45°, which agrees with my measurements on:
http://www.hugha.co.uk/NodalPoint/Index.htm#Samyang8mm

All of us learned a lot about fisheye lenses use as well as from Hugh's articles. Thx a lot to both of them.

You have to decide how many shots you want to take around with your camera lens combination. Sometimes you can go for 6 or 4 shots around, sometimes for 4 or 3 shots around. With less shots around you will end up with less overlap. But you will be quicker to get your set around. May be important in busy places with alot of people moving around, in concerts with people moving their arms to the music, journalism etc. The "normal minimum overlap without risk" is25%. I prefer 30%. May be you have to shoot some more pics, just some seconds to take them, but this might save your time to go back to New York to shoot again and you lost the special moment and atmosphere anyway.
Success,
Heinz
Thanks for your comment Heinz, much appreciated.
I too have learned a lot from Michel’s site.
The choice of the number of shots around depends on many factors, and the position chosen for the NNP is affected by the number of shots around except for lenses such as the Sigma 10 to 20mm f4-5.6 EX DC HSM, where the Entrance Pupil is a single point.
I have tried to show this in the attached diagrams and it is the variation in the location of the Entrance Pupil relative to the angle of the ray that causes the phenomena mentioned by Denis.
I don’t know if it is the influence of my photogrammetric background, but I tend to go for a larger overlap (50%) than most resulting in more images to create the panorama.
I have a number of reasons for doing this:
By having more images I am making more use of the central part of the lens where the displacement of the Entrance Pupils is much tighter to try and minimise the effects of the lens distortion from the fisheye lens;
The larger overlap gives a greater area for Control Point generation;
The larger overlap gives me much more scope for using PTGui’s wonderful Mask tool,especially in very crowded environments such as hospitality suits where a lot of people are moving and moving around in closed proximity to the camera.
Of course, using the Mask option in PTGui does mean that I am sometimes choosing parts of the image away from the centre, but PTGui does a great job and I really appreciate the genius that is Joost that makes it all work!
When shooting hospitality suits with people milling around, some very close to the camera, I use a Sigma 4.5mm lens to get a full circle image to minimise the number of shots.
This does have the disadvantage that I am using only one third of the sensor area,but with 24Mp I am getting a decent resolution in the final panorama.
I could follow Heinz’s advice of 3 shots round to get the shots off as quickly as possible, but I have found that people can move very fast, especially if they are picking up a glass of wine! and sometimes it is necessary to wait a bit for people to move out of shot – I don’t want the lady with the large pink hat to appear twice in the panorama!
Consequently I take four shots round, which gives me the overlap advantages already mentioned, but I actually go round three times (12 shots) so I have plenty of material for eliminating unwanted movements, postures, gestures, etc.
I often wonder what the guests make of this gutywalking round and round a camera on a tripod without apparently taking any photographs because I am not looking at the screen or through the view finder and I am very conscious that the Guests, of their Hosts, have paid a lot of money to enjoy good food and drink and their favourite sport, but I find that I am quite inconspicuous and only occasionally is a Guest looking at the camera.
I guess it is because the fisheye lens is unobtrusive and I am not aiming the camera at individuals.
I have now made over 500 of these panoramas, some better then others, and can only count seven occasions when people have spoken to me, so I am happy that they are enjoying themselves and my camera and I are not a distraction.
Most of the seven were interested to know what I was doing, one knew about panoramas and we had a good discussion and only one asked me not photograph him with a polite “No photographs please”, so I reckon the system works.
The thing is really to find what suits you best and gives you the results you are looking for and for me I find know what the lens is doing helps.
When I set up the Sigma 4.5mm on a R1 I spent a few minutes with the laser determining the positions of the entrance Pupil for different angles, looked at the results and set the NNP 4mm in front of the gold ring and did not need any further adjustment, but as mentioned at he beginning, the reality is that we can only have one NNP (Nodal Ninja Point) and the methods suggested by John achieve the same result.

kevin6270
10-16-2012, 08:45 AM
Kevin, It's best if you do some tests yourself to determine the optimum settings. It doesn't take long. See:
http://www.panoramas.dk/panorama/Fisheye-NPP/parallax.html
http://www.johnhpanos.com/epcalib.htm
http://www.easypano.com/forum/display_topic_threads.asp?ForumID=1&TopicID=4162&PagePosition=1
.
John

Thanks John. Used these to determine the optimum NPP, but was getting some obvious stitching errors and poor control point numbers (averaging 8-9). I was pretty sure it wasn't my NN set up, so tried various settings in Ptgui. Finally amended the Minimise Lens Distortion setting from Medium to "Heavy + Lens shift" and hey presto - perfect stitching and ave CP of 1.08 and high of 2.4.

Really pleased with the end result, but don't understand why this simple change made all the difference.

hindenhaag
10-16-2012, 09:09 AM
Hi Kevin,

Using a fisheye lens with exif info PTGui automatically is set to "Heavy + Lens Shift" in the last version to correct sensor displacements during production and to minimize lens distortion. Joost has optimized PtGui to find CP with this special set up to spread CP in vertical and horizontal direction. Because Samyang is a manual lens, PTGui misses this info. Set lens to Full Frame and 9mm, and use Heavy + Lens shift optimizing. To create a template, which will optimize stitching in future with the same camera lens combination, you can try to go down to max distance CP in between 1.0 and 2. We can try to help you if you send your pics and PTGui PTS. file via www.ge.tt .

Then we can send a template for your set up. Next time you do a new project, you can use "apply template" to optimize CP or your special set up.

Heinz

kevin6270
10-17-2012, 10:03 AM
This is the most helpful forum, thanks !!

I've posted the files to http://ge.tt/4Z2vrgP?c so hopefully you can pick them up. Pics were shot at -15 deg (x4) and +60 deg.


Kevin

John Houghton
10-17-2012, 11:19 AM
Kevin, The images stitch quite well, with only a tiny glitch in the beam across the ceiling. Taking just two of the images separated by a yaw increment of 180 degrees reveals evidence of significant misalignment of the panorama head:

http://www.johnhpanos.com/test1Panorama.jpg

The bottom rail setting seems fine. The camera needs a small rotational adjustment on its mount to the upper rail. This may be the cause of the glitch in the beam. It would be a good idea to stick a few temporary markers on the beam (e.g. post-it notes) to act as features for control points to help the optimizer.

John

kevin6270
10-17-2012, 12:05 PM
Thanks John, I'll try that. I found centering above the NN logo tricky with the fisheye - even using full zoom in live view - as the tolerances are so tight. I swapped out the Samyang for an 85mm just to get the lens centred, but it seems that's not quite worked. Need to become better at that !

Also, looking forward to seeing Heinz feedback on optimal settings, and template.

hindenhaag
10-17-2012, 11:03 PM
Hi Kevin,

the download link does not work for me. It says: download is not finished.

Heinz

Ok, just managed pic by pic. Download all files didi not work. I just had a quick look around. Because of results I changed to Circular Fisheye lens parameter and cropped. In case of problem I always start with the shots around only. In optimizer change to heavy and lens shift and optimize. Del worst CP. When nothing is deleted, I changed to control points. I just marked the biggest ones with shift and del all above 20 etc. I just went down in max CP and ended up around 2 without placing anything manually.

Now change to control points and compare pic to pic and change with red arrows on top. Now you see in most cases CP sit in a bulk position specially in picture 2. Cp on the galls only.

What we always say: CP needs a good spread in vertical and horizontal. Have a look to the ceiling, where you want to add your zenith shot: nearly no CP.

I will explain this to you later because I have to leave now to go to work. Just go through and try to analyze the placement of CP. And try to find out which are in "a place of nowhere, this means in the middle of a brown spot instead of its sharp corner. When this is in a place of bulky CP, you can just delete this one without trying to correct the distance.

http://ge.tt/2wlYOjP/v/0

So long,
Heinz

John Houghton
10-18-2012, 07:54 AM
In case of problem I always start with the shots around only....
http://ge.tt/2wlYOjP/v/0
Heinz
Heinz, In this case, you appear to have started with only 3 of the 4 shots around, plus the zenith. An oversight, I'm sure.
.
John

Tiziano
11-10-2012, 05:03 PM
I'm quite new to all this but am researching this lens. I'm using a Nikon D90, would the settings discussed here be applicable to my cropped camera as well?
I'm going with an NN4 BTW.