PTGui Apex/Nadir Stitching Problem with Nikon 10.5mm DX Lens

Discussion forum for Tawbaware's PTAssembler software, Helmut Dersch's Panorama Tools software and any other photography related software

Moderator: spamkiller

Post Reply
kappel
Posts: 4
Joined: Thu Jul 02, 2015 12:39 pm
Location: Tampa, Florida, USA

PTGui Apex/Nadir Stitching Problem with Nikon 10.5mm DX Lens

Post by kappel » Fri Jul 03, 2015 3:53 pm

Recently acquired Nikon 10.5mm f/2.8 DX full-frame fisheye on Nikon D7000 for creating VR panoramas. Did test photos to learn how to use and stitch images for this new lens with PTGui version 9.1.9. Can stitch the single horizontal row without problems. Can not stitch the Nadir and Apex images in with the horizontal row even though created what looks like good set of control points. The distance on the Apex/Nadir control points is very large. Process followed is 6 portrait images taken 60 degree apart for horizontal row, followed by Apex shot straight up and Nadir shot taken at an angle. Viewpoint correction on Nadir shot does not seem to help. Have followed similar procedure with my Nikon 14-24mm f/2.8 on Nikon D7000 with no problems. I am thinking that the problem has to do with characteristics of new lens. Any ideas would be appreciated.

I have uploaded the 6 horizontal images, 1 apex image (taken at same position as first shot of horizontal row), and 1 nadir image as well as the PTGui project file at: http://ge.tt/210J1fJ2?c

I also uploaded one image for the flat pano output for the single horizontal row and also one image for the complete horizontal plus apex and nadir.

I have additional Apex and Nadir shots that I did not upload if they are needed. However, in the past, all I ever needed was usually just one Apex and one Nadir shot.

johnh
Posts: 2118
Joined: Sun Jul 20, 2003 2:43 pm
Location: UK
Contact:

Re: PTGui Apex/Nadir Stitching Problem with Nikon 10.5mm DX

Post by johnh » Fri Jul 03, 2015 7:02 pm

The problem with your stitch is that the zenith image is in a different orientation to the other images. This is because the auto rotate sensor doesn't work when the camera is pointing directly up or down. It's best to disable this feature in the camera to avoid this problem, but you can rotate the zenith into portrait orientation using Photoshop before input to PTGui. NB. It does matter whether you rotate the image clockwise or counterclockwise by 90 degrees. The application of lens shift parameters needs to be consistent. Probably you want to rotate clockwise. The images then align not too badly.

The viewpoint nadir aligns ok but you must assign control points ONLY on the flat floor.

My project file is at http://ge.tt/6Nb9YfJ2/v/0?c .

John

kappel
Posts: 4
Joined: Thu Jul 02, 2015 12:39 pm
Location: Tampa, Florida, USA

Re: PTGui Apex/Nadir Stitching Problem with Nikon 10.5mm DX

Post by kappel » Sat Jul 04, 2015 4:50 pm

Hi John,

Thank you for being so helpful and especially for providing your PTGui project file which was most useful.
For the nadir shot, I created all control points on the flat surface of the floor, masked the tripod as you did, and set the viewpoint correction for that image.
For the apex shot, I rotated the image clockwise to "portrait" and created all control points on the flat surface of the ceiling.
With these changes PTGui reported "very good" on the alignment with max distance of about 16.
The resulting flat panorama and VR movie showed no noticeable misalignments or out of place artifacts.
I can declare the 10.5mm DX lens test a success and proceed now to a real panorama.
I have posted the updated images and project file again at:
http://ge.tt/3cVMOiJ2?c
Please take a look at these if you can and let me know is there are any other suggestions that you may have.

A couple of other questions for my better understanding:

(1) In the past, I have successfully used PTGui to stitch apex shots with a different orientation (landscape versus portrait) than the images in the pano rows which were taken as portrait. Those panos were all taken with Nikon 14-24mm f/2.8 on a Nikon D7000 (DX). Any thought as to why this same process did not work for this pano using the Nikon 10.5mm f/2.8 DX on same Nikon D7000?

(2) My first trial attempt with the 10.5mm was a two row (first downward angle, next upward, apex, and nadir). There was plenty of overlap between rows and I had a full 360x180 coverage. This did not stitch well, but I was not yet following the other suggestions you had made on the newer pano trial which was single row, apex, and nadir. My reason for the two rows was possibly better overlap and easier stitching with the 10.5mm. Any thoughts on using the 10.5mm for two rows with a downward row and an upward row but no horizontal row?

(3) I have read frequently that people start a multi-row pano with the horizontal row. For a three row pano, I usually start with the downward angle first, then horizontal, then upward angle, then apex, and finally nadir. I have not had trouble stitching in PTGui, but do you have any comments on the process?

(4) Any comments on use of Nikon 14-24mm (at 14mm) f/2.8 versus Nikon 10.5mm f/2.8 DX for high quality panos and VR panos. I do mostly large buildings (e.g., churches). I know the resolution of the 10.5mm will be less, but the 14-24mm takes me about 114 shots (3 rows, plus 4 apex, 4 nadir) which includes 3-exposure bracket at each position. Using natural night and interior lights, this takes about 5 hours which includes 1 hour setup, 4 hours of photography, and 0.5 hour of take down. I am planning on still using the 14-24mm but would like to use the 10.5mm for less important panos. Any comments?

johnh
Posts: 2118
Joined: Sun Jul 20, 2003 2:43 pm
Location: UK
Contact:

Re: PTGui Apex/Nadir Stitching Problem with Nikon 10.5mm DX

Post by johnh » Sat Jul 04, 2015 6:30 pm

Your stitch looks fine though there's a glitch in the cord hanging from the fan. The horizontal row together with the nadir optimize very well, giving an average cp distance of 0.88 and a maximum of 2.33 (after correcting a couple of points). However, the zenith doesn't align very well and there are few clues as to the cause. Maybe the head setup isn't quite as good as it might be. You should consider shooting the zenith at a pitch of 70, say, instead of 90. You still fill the zenith hole perfectly adequately, but there is better overlap down towards the horizon where there are many more features for control points. You could in this case also capture the fan and hanging cord in its entirety in the zenith, thus making it possible to avoid stitching errors there.

Q1. If you mix portrait and landscape shots and use a global horizontal fov parameter, it cannot be evaluated correctly for both orientations, as they are clearly different. Remapping the images onto the spherical stitching surface is therefore going to be badly compromised. (There's an option to assign individual lens parameters for the odd man out, which is the proper way of dealing with the mismatch if you don't want to rotate the image before stitching). I have no idea why you have not experienced problems in the past with mixed orientations. It may be that the image content was not sufficiently detailed to reveal obvious alignment errors - e.g. blue sky or plain ceilings.

Q2. A single row should work quite well enough. Two rows give complicated joins at the horizontal level where usually most of the interest lies. Moreover, you may be using the lower image quality at the top/bottom of the images for the central region of the panorama, where best quality would be preferred.

Q3. It doesn't really matter in what order you shoot the rows, though with subjects containing moving objects like clouds, you want minimal delay between overlapping rows. When running PTGui, if you initialize the yaw, pitch and row parameters with their known nominal values, you should minimize the risk of stitching going awry.

Q4. Choosing between the 10.5mm and 14mm options is something that you must decide for yourself. You have the equipment to do your own tests and you know what quality you are looking to achieve.

John

kappel
Posts: 4
Joined: Thu Jul 02, 2015 12:39 pm
Location: Tampa, Florida, USA

Re: PTGui Apex/Nadir Stitching Problem with Nikon 10.5mm DX

Post by kappel » Sun Jul 05, 2015 3:13 pm

Hello John,
Your general comments on the stitching were very good. I will indeed try some tests with zenith at 70 degrees. Do you think that I would need to take the 70 deg. up shots in the same manner as the horizontal row (6 shots at 60 degrees apart) or might a single 70 degree (or higher angle) up shot cover the entire apex hole? That way perhaps I could just take 1-4 shots at that angle saving some time (usually take 3 exposures per position at about 8s, 15s, 30s).

This pano was definitely trickier to align since it was a test at home and the ceiling was almost featureless. I am thinking that the pano head is probably not the problem. I was using the Really Right Stuff tripod, BH40 ballhead, PC-Pro panning clamps, and multi-row pano gear (Lightweight Multi-Row Pano Elements Package Pro). It is quite accurate and stable. I will also have a chance to test the 10.5mm in several large churches this summer.

In response to this post, I also received an email from John Law (UK). He said, "I use a 10.5 on a D810 all the time with a monopod. With the lens hood removed (usually called ‘shaving’) only 4 images in portrait mode are needed to capture the whole sphere.Some people only use 3 - zenith and nadir shots are not needed!". Here are a number of recent examples - some converted to videos. john-law.net/projects." Any comments?

(Q1) I also don't know why I was successful in the past with the Nadir in landscape orientation using the 14mm versus all other images being portrait. I will try to take some time to post a PTGui project and images so you can see an example of this.

(Q2), (Q3), (Q4) - Great comments

Thanks so much!
Lee

johnh
Posts: 2118
Joined: Sun Jul 20, 2003 2:43 pm
Location: UK
Contact:

Re: PTGui Apex/Nadir Stitching Problem with Nikon 10.5mm DX

Post by johnh » Sun Jul 05, 2015 4:03 pm

Only one zenith shot angled at 70 degrees is needed, as the hole is fully covered. You can verify this simply by dragging the zenith image in your project on the Panorama Editor window (or directly setting its pitch value to 70 on the Image Parameter tab).

A test that you could usefully try with your panorama head is to take two shots with the camera pointing vertically down, rotating the panorama head by 180 degrees between them. Check for parallax effects when the images are aligned in PTGui.

John's comments about only taking 4 shots around relate to his full frame camera. You need 6 for yours + zenith, but of course all sorts of options are available if you choose to change all your equipment.

John

kappel
Posts: 4
Joined: Thu Jul 02, 2015 12:39 pm
Location: Tampa, Florida, USA

Re: PTGui Apex/Nadir Stitching Problem with Nikon 10.5mm DX

Post by kappel » Sun Jul 05, 2015 4:26 pm

Hi John,

I will try all of your good suggestions.

On the RRS multi-row pano equipment, I think the NPP alignment on one axis is good since I have checked the parallax on it several times. However, the alignment on the other axis (the centerline of lens with the rotation point) was only eye-balled and could be off a bit.

I did suspect John L. was using his D810 in full-sensor mode to be able to get away with so few shots. I am using the D7000 cropped sensor with the 10.5mm DX. I have sent him a follow-on email: "Are you using the D810 in full- or cropped-sensor mode with the 10.5mm DX lens? I did view your photos, but I could not tell which were done with the 10.5mm. Could you point me to some specific photos taken with the 10.5mm? How do you achieve sufficient stability for stitching with a monopod? I normally take long exposures 8-30 seconds and a monopod would seem to be a problem?"

Thanks again,
Lee

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest