NN5 + 7d + Tokina10-17 settings
Page 1 of 2 12 LastLast
Results 1 to 15 of 18

Thread: NN5 + 7d + Tokina10-17 settings

Share/Bookmark
  1. NN5 + 7d + Tokina10-17 settings

    #1

    Hello all. I have resisted requesting help here as there is so much fantastic information i felt it better to learn myself. However, i have come as far as i can along the learning curve and i would appreciate some advice. Ok, where to begin...

    7d + NN5 + nadir adaptor + Tokina10-17.
    Spherical panos.

    Aim- successful indoor 360 panos to make vr tours and master the settings in a tricky environment for ease of use outdoors.

    Work flow...
    -6 shots @ 30deg + zenith + nadir + nadir patch using nadir adaptor swung to 180deg, tripod moved more than DPD in direction of front leg and re-leveled.
    -ptgui pro
    -align
    -mask tripod
    -viewpoint correction on nadir patch
    -re-align.
    -extra targeted masking when necessary to reduce stitching errors in wide angle nadir shot & patch.

    I have read and followed these links to find the optimal stitching position & nnp...
    Tutorials: http://www.johnhpanos.com/tuts.htm
    No parallax point: http://www.johnhpanos.com/epcalib.htm
    Lower rail setting for optimal stitching position:
    http://www.easypano.com/forum/displa...PagePosition=1
    Entrance pupil database: http://wiki.panotools.org/Entrance_Pupil_Database

    I found the LRS using the optimal stitching point tutorial above and confirmed this was in fact in line with the centre of my lens & nn5 rotator (about 57mm).

    I have found the URS using 2 techniques (tape on the window with buildings in the background) and by setting up a vertical pole in my living room in line with a corner of the wall. In both vases i found to the LRS to be approximately 88mm which was lower than i expected. I tested settings lower and higher and using these techniques it kept coming back to about 88mm (cropped images attached). The mis alignments were more notable at +- 88.

    My main problem is stitching errors in the bedroom where there are a lot of rectilinear patterns (book shelves, matt, doors etc). Primarily the stitching errors are most evident on the matt below the tripod, but only when the nadir patch is included. Annoyingly they are not consistent. I have tried to make panos with LRS=57 & URS with values between 86 & 100 and sometimes, confusingly, these did not effect the amount of stitching error, but sometimes its positioning along the edges of the matt. I cannot find consistency in the factors which may effect the stitching errors despite many hours trying and re-trying +- the optimal settings i found. In fact sometimes i got better results with non-optimal settings. Sometimes i got better results with settings i had previously used unsuccessfully. Sometimes the entire image would distort into a curve, but only with the nadir patch included. Of course my HDR attempts sometimes compounded the errors.

    I have tried manually adding control points across the edges of the matt in the nadir & nadir patch, but confusingly, the stitching errors seemed to directly contradict these control points. Targeted masking has certainly helped reduce stitching errors, but it does not do so consistently, especially on either side of the tripod along the edge of the matt.

    I read that the nnp in a fisheye lens is not stationary and is most affected at 90deg, i.e. the nadir. I wonder if this may be the problem but then everything i've read suggests optimum settings can be found.

    I am ware that the room in which i am testing this is a challenging environment for panos but i figure if i can nail it here then i'm all set really. However, it is this complete lack of failure/success consistency across various settings which is head wrecking!

    So, if you're still reading!...
    Hints on work flow?
    Suggestions on why i may be having trouble?
    It would be great if anyone can suggest tried and tested LRS & URS for my camera / lens combo.
    Are there circumstances when a decent stitch is simply not possible despite optimal calibration or can they normally be forced with clever masking & control points?
    Is proximity to walls etc a factor?

    Any advice or suggestions gratefully received coz i'm pretty much at a loss!

    Thank you
    J
  2. #2
    Users Country Flag
    Join Date: Aug 2008
    Location: Netherlands
    Posts: 1,741

    Hi Johnny,

    Cloud you please send a set of your pics with Ge.tt to test your problems?

    Thx,
    Heinz
  3. #3
    Users Country Flag
    Join Date: Sep 2009
    Location: Los Angeles, CA
    Posts: 359

    Quote Originally Posted by JohnnyLawlor View Post
    .
    I found the LRS using the optimal stitching point tutorial above and confirmed this was in fact in line with the centre of my lens & nn5 rotator (about 57mm).
    .
    Of course my HDR attempts sometimes compounded the errors.
    .
    .
    Johnny,
    .
    Centering the rotator in the viewfinder is not the way to set or verify the LRS.
    .
    This is the method I use to calibrate all my rigs. Works great. http://www.easypano.com/forum/displa...PagePosition=1
    .
    Do not attempt any HDR work until you have mastered the basics. There are too many places for things to go wrong as it is. Trying to do everything at once when you do not know if your rig is properly calibrated yet is asking for trouble.
    .
    Post a set of downscaled images for us to have a look at.
    .
    Dennis
  4. #4

    Quote Originally Posted by JohnnyLawlor View Post
    Work flow...
    -6 shots @ 30deg + zenith + nadir + nadir patch using nadir adaptor swung to 180deg,
    I'm not sure what "@30deg" means nor do you say what focal length zoom setting you are using. I've had good results at 11mm and 14mm. As often as not, stitching errors turn out to be the result of unskilled use of PTGui so it would great if, as others have already suggested, you make a set of images available (single exposure set, good quality jpegs at half size would be fine). The free service at http://ge.tt/ can be used without registering if you don't have your own web space to upload your images. Post a download link here.

    John
  5. #5

    Sorry for the delay replying. I had some trouble posting my original question and for a few days I nearly gave up. Anyway, thanks for taking the time and interest to respond.

    Dennis, i did follow that tutorial. I just meant that i noted the results were very much equivalent to centering the viewfinder. Sorry about the confusion, I was trying to express many hours of trouble shooting in one post!

    John, nicely spotted type error! I meant 6shots @ 60deg around.

    Here is a sample.
    www.johnnylawlor.com/JL pano.zip (LRS 57, URS 89)
    You'll notice the 2 culprits of broken seams primarily around the mat, along with the curved distortion of the overall image.

    I must emphasise that i have tried many times using upper rail settings methodically between 85 & 100, with increments of 2-3mm so as to observe any differences. Some have been considerably better than this one i have included, whereas others have been worse. The point is i have been unable to find a common set of factors and the successes/failures have been inconsistent, making it difficult to confirm my optimal settings.

    I agree this probably indicates my lack of experience with ptgui. I have been mostly depending on the automatic control points, though i have tried manually adding more along the edges of the mat. Confusingly, i had the impression that these were ignored, or relegated by ptgui. But then if i have to remove some of the automatic control points it's difficult to know which ones i should remove.

    Any comments about my optimal settings? As noted earlier using the saw tooth method (and checking in the viewfinder) and got LRS=57. Using tape on the window and a building in the background along with lining up vertical lines of furniture i got URS=88 (which was less than i expected, hence trouble shooting at higher settings also).

    J
  6. #6

    oops, i put a space in that link. please copy and paste the whole link as far as '.zip' for it to work.
  7. #7
    Users Country Flag
    Join Date: Aug 2008
    Location: Netherlands
    Posts: 1,741

    Hi Johnny,

    I got the same result with faults only in the carpet.

    I got a very good result stitching the pics without the nadir shots. Max distance is around 1.5. Import pics, align, > advanced, >optimizer, set "minimize lens distortion" to "Lens shift + heavy". Go to "control points" in the top tab list and choose "delete worst control points" till no CP is deleted.

    Adding the Nadirs shots go to "advanced" in optimizer and set the "View point correction" tab normally for Nadir2 because this is moved aside.

    The problem with the nadir pics is orientation: the first one is 180° visa versa and both are a little bit turned to the anchor pic you shot in the beginning. The goal is to present the pics in one "Row". This delivers CP close to each other and not on the opposite site. Shooting the Nadirs you should not move and try to stay in same orientation. In your case its difficult to get CP's on the carpet margins in the area of the tripod. You can hardly find a relation between your normal shots and the Nadirs to find CP's.

    To proof your settings I need a set of "Smooth Set Up" pics to check your LRS. Use F-max, pitch down a little bit till you catch the corner of the rotator, focus to the corner and take your set around. Then import this set to PTGui and check for teeth. Having done this you can not compare with the viewfinder pic. Going through PTGui will correct the sensor and lens misalignments from production. I always use this method with good results.

    I feel you setting are quite good.

    Here is the template i made for you: http://ge.tt/2lqoYN81/v/0

    Import this to your image folder or your preferred folder for your templates. You can set the folder under PTGui "Preferences" in top row. Go to "Folders&Files" to choose the place to store it.

    Next time you do a new project, Import images and go to Top Row "File" open it and "apply template". Then you import the parameters of your last project without control points to get lens corrections etc.

    Heinz
    Last edited by hindenhaag; 12-08-2013 at 01:48 AM.
  8. #8

    I found the same as Heinz. The two down shots are upside down. Turn off auto rotation in the camera to avoid this. (The auto rotate sensor does not work when the camera points directly up or down when gravity acts at right angles to the plane of the sensor). My stitch is at http://ge.tt/6Xx2IN81/v/0?c (with the two down shots rotated by 180 degrees. Viewpoint correction is needed for both shots as the head setup is not adequate.

    John
  9. #9

    After a 35 mile bicycle ride, I had some more ideas and further investigation revealed the major cause of the alignment difficulties: the two nadir shots were taken at a focal length setting of 11mm, whereas all the others were taken at 10mm. Evidently the zoom ring was nudged round when the camera position was being adjusted. After assigning individual lens parameters to the last two shots, the first nadir aligned very well without recourse to viewpoint correction, so the panohead setup seems perfectly adequate. The second nadir aligned ok with viewpoint correction as before.

    John
  10. #10
    Users Country Flag
    Join Date: Aug 2008
    Location: Netherlands
    Posts: 1,741

    Hi John,

    A bicycle ride seams to blow the head free for new power. I would not have thought to check the focal length of the pics. But from now on I'll go for this too. I always check my zoom setting before shooting. But may be a good tip for Johnny to tape zoom and focus ring.

    Heinz
  11. #11

    Wow thank you guys so much. I am blown away by your helpfulness :) Much appreciated.

    For the sake of completion here are my main LRS calibration tests... www.johnnylawlor.com/JL-LRS-calibration.zip
    I didn't follow the instructions exactly but the results seem to clearly indicate teeth at 56 & 58, whereas 57 seems optimum.
    The UPS were included in my first post.

    Thank you both for clarifying the need to orientate the down shot with the anchor shot. I hadn't imagined this would be necessary.

    I have been using "lens shift + heavy" and running the optimiser on the nadir patch all along.

    Heinz, thank you for alerting me to the "Delete worst control points" option. This has certainly helped. I can only assume that the small misalignment at this stage is as John noted re the focal length of my lens in my down shots. This has certainly taken me by surprise as my lens does not feel in any way loose, but of course the evidence is damning, they are clearly at 11mm when i check them! This could certainly explain the inconsistent results, supposing my lens was being nudged arbitrarily. More tests needed then.

    Question... I notice that you guys refer to aligning the main images THEN importing the nadir shots and re-aligning, masking etc. Is there a reason for this? I have been importing all my images and aligning them together in one go inc the nadir shots, then masking, re-aligning etc. I occasionally imported the patch shot separately but only as a trouble shooting step.

    John, can you explain how you assign individual lens parameters? I can only see the option to uncheck the 'Automatic (use EXIF data from camera)' option and force it to take the default 10.45mm for the whole project. This didn't seem to make a difference and i can't find anywhere where i can manually adjust the focal length setting for individual photos.
  12. #12
    Users Country Flag
    Join Date: Aug 2008
    Location: Netherlands
    Posts: 1,741

    Johnny,

    I only do this to analyze the problems of a set of pics to get to know this step by step. Having a good stitch with set around, zenith and nadir1 I make a template and store it for the next project.

    Concerning the calibration your pics are what you looking for: 56mm cut to the right, 57mm round circle, 58mm cutting to the left. So 57 is your LRS. Well done.

    Heinz
    Last edited by hindenhaag; 12-08-2013 at 11:11 PM.
  13. #13

    Quote Originally Posted by JohnnyLawlor View Post
    Thank you both for clarifying the need to orientate the down shot with the anchor shot. I hadn't imagined this would be necessary.
    That is not the requirement. All images should preferably be supplied to PTGui in the same orientation - i.e. all in the native landscape orientation of the camera, or all rotated from that orientation by the same angle. So for portrait orientation, rotate all the images clockwise by 90° or all counterclockwise by 90°. This ensures that the lens shift parameters shared by all the images are applied correctly. PTGui can cope with mixed orientations, but the processing then becomes much more complicated with the necessary use of individual lens parameters.

    Question... I notice that you guys refer to aligning the main images THEN importing the nadir shots and re-aligning, masking etc. Is there a reason for this? I have been importing all my images and aligning them together in one go inc the nadir shots, then masking, re-aligning etc. I occasionally imported the patch shot separately but only as a trouble shooting step.
    People gradually work out a workflow of their own that suits them. While automatic operation works well enough in many cases, PTGui permits manual intervention at every stage of the process and experience will help you to recognize when things are not going as they should and what you can do to rectify the problem. The optimizer is the key program, and this works by improving the current state of alignment by making small steps, moving images slightly each time to get the control points to align better. Starting from scratch, the optimizer can sometimes come up with bizarre results, particularly when a few bad control points are present. When you know the values of the lens parameters and all the angles the shots were taken at, it greatly helps the optimizer if you give it all this information on the appropriate tabs so that it has the best possible starting position. Applying a template is a convenient labour saving way of doing this.

    As a general rule, all images that are shot with the camera in the NPP position can be directly optimized together. An offset nadir shot is best dealt with separately so that it cannot skew the alignment of the main set of images. After aligning the main images, you can uncheck their y,p,r parameters on the Optimizer's advanced tab and also the lens parameters, and then optimize the nadir shot in position without affecting the alignment of the already aligned main images. So either add the offset nadir image separately, or add in all the images together but use the "Include control points of.." list on the Optimizer tab to include/exclude the nadir shot from the current optimization.

    John, can you explain how you assign individual lens parameters?
    In advanced mode, go to the Lens Settings tab. There you will find the "Use individual lens settings for.." list. Images not checked there will use the normal global lens parameters. Individual lens parameter values can be entered on the Image Parameters tab, and there will be extra lens parameter boxes to check on the Optimizer tab.

    John
  14. #14

    Ok got it. So then which settings did you adjust to compensate for the incorrect focal length? I can only imagine a bit of trial & error with the fov parameter???
    Of course i will shall be keeping an eye on this error source (lens nudge) in the future, but i wish to use the images i made a couple of weeks ago as i have the apartment full of christmas decorations now and i don't want to take them down or wait until after xmas to put together my vr tour!! Also, it's good for helping my understanding of ptgui.
    Hi5's and xmas wishes to you both :)
  15. #15

    Quote Originally Posted by JohnnyLawlor View Post
    Ok got it. So then which settings did you adjust to compensate for the incorrect focal length? I can only imagine a bit of trial & error with the fov parameter???
    The "bit of trial and error" is performed automatically by the Optimizer. That's its job. It will vary the individual fov parameter (if checked) to achieve a good alignment.

    John

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •