This site is supported by the advertisements on it, please disable your AdBlocker so we can continue to provide you with the quality content you expect.

Is this Parallax error?

Discussion in 'General X Camera Forum' started by waloshin, May 16, 2018.

  1. waloshin

    waloshin Member

    Joined:
    Aug 8, 2016
    Messages:
    73
    Likes Received:
    8

    -Return to Top-

    I am trying to setup my Gigapan Epic 100, with my Fuji XE2S, Nikkor 50mm. as well as my Nikkor 35-135mm f3.5-5.6, Nikkor 70-210 F4.5-5.6, and Fuji 28mm f2.8. This test shot was with the 50mm.

    Please login or register to view links
    Another example: Please login or register to view links
    Is this Parallax error?
     
  2. Frankie

    Frankie Premium Member

    Joined:
    Mar 31, 2011
    Messages:
    3,544
    Likes Received:
    640
    Location:
    The Left Coast

    -Return to Top-

    There is no parallax error in using the X-E1, 2, 2s or 3 [...or any camera with an EVF] with any lens or mount...WYSIWYG 100%.
     
  3. waloshin

    waloshin Member

    Joined:
    Aug 8, 2016
    Messages:
    73
    Likes Received:
    8

    -Return to Top-

    Though when taking panoramas it can happen.
     
  4. Richard_M

    Richard_M Premium Member

    Joined:
    Jan 28, 2014
    Messages:
    850
    Likes Received:
    1,405
    Location:
    Australia

    -Return to Top-

    It does look like it, did you have the camera and lens set at the Nodal point?
     
  5. dem

    dem Premium Member

    Joined:
    May 31, 2015
    Messages:
    811
    Likes Received:
    441
    Location:
    UK

    -Return to Top-

    You are trying to stitch several rows of panoramas ala Brenizer method. It is important to lock all settings (exposure, white balance, focus) and allow for plenty of overlap between the images.

    Please login or register to view links

    Turning the camera around its front nodal point (no parallx point, close to where the front element is) rather than simply panning the camera will make the job of your stitching software easier.
     
    Last edited: May 16, 2018
  6. Arjay

    Arjay Admin Staff Member

    Joined:
    Feb 27, 2011
    Messages:
    13,561
    Likes Received:
    5,307
    Location:
    Munich, Germany

    -Return to Top-

    Your test samples do indeed show stitching errors that are the result of parallax errors. The camera's EVF is WYSYWYG, and doesn't exhibit parallax in itself, but the individual shots do. This happened because you didn't mount the camera over its lens' nodal or "no parallax" point.
     
    Shadowside, Irene McC and fugu 82-2 like this.
  7. Goldingd

    Goldingd Premium Member

    Joined:
    Jan 20, 2014
    Messages:
    935
    Likes Received:
    280
    Location:
    Anaheim CA

    -Return to Top-

    What software are you using to stitch the images together with?
     
  8. Frankie

    Frankie Premium Member

    Joined:
    Mar 31, 2011
    Messages:
    3,544
    Likes Received:
    640
    Location:
    The Left Coast

    -Return to Top-

    Parallax error only occurs when there are two perspective centres in a single view.
     
    lawsofphysics likes this.
  9. Shadowside

    Shadowside Good Glass is Forever...

    Joined:
    Jun 24, 2013
    Messages:
    2,168
    Likes Received:
    1,818
    Location:
    Winnipeg, Canada

    -Return to Top-

    Software won't fix or prevent this, you need to calibrate your rig, for each lens and/or body, to avoid parallax.
     
  10. waloshin

    waloshin Member

    Joined:
    Aug 8, 2016
    Messages:
    73
    Likes Received:
    8

    -Return to Top-

    Gigapans stich software.
     
  11. waloshin

    waloshin Member

    Joined:
    Aug 8, 2016
    Messages:
    73
    Likes Received:
    8

    -Return to Top-

    I am going to shoot a nice flat field today with my nikkor 50, and my fuji 27mm and see what happens.
     
  12. Arjay

    Arjay Admin Staff Member

    Joined:
    Feb 27, 2011
    Messages:
    13,561
    Likes Received:
    5,307
    Location:
    Munich, Germany

    -Return to Top-

    The test samples consist of two individual views stitched together into one. Since the samera wasn't positioned so that the center of rotation during shooting was located precisely under the lens' "No Parallax" point, these individual shots were taken from slightly different viewpoints. This made it impossible for the stitching software to align the individual shots so that the aggregate sum would not show any discontinuities.
     
    lawsofphysics and Shadowside like this.
  13. Arjay

    Arjay Admin Staff Member

    Joined:
    Feb 27, 2011
    Messages:
    13,561
    Likes Received:
    5,307
    Location:
    Munich, Germany

    -Return to Top-

    I doubt if you would be able to solve your problem doing that.

    You need to determine the "no parallax point" by shooting a scene with a grid or any other vertical structure in the foreground and some vertical structures in the background. Take pairs of shots, shifting your camera forward/backward for every pair. Every pair itself should consist of two shots during which you pan the camera left and right for ~15 to 20 degrees. Take notes of the forward/backward positions for every shot pair. You'll have found the correct "no parallax point" if there isn't any parallax error left between the shots of a pair, i.e. if the offset between your foreground and background structures remains constant in a pair.
     
    Shadowside likes this.
  14. Frankie

    Frankie Premium Member

    Joined:
    Mar 31, 2011
    Messages:
    3,544
    Likes Received:
    640
    Location:
    The Left Coast

    -Return to Top-

    Time to expand what I'd wrote...

    Photogrammetry 101:
    1. "Parallax error only occurs when there are two perspective centres in a single view."
    2. If the [lens] perspective centre is aligned with the rotation centre in the same axis, then the error would only be imperfect merging of each slit.
    3. When photographing a flat field in panoramic mode [with camera leveled], each slit further away from image centre has progressively slightly smaller image scale...also image displacement [horizontal keystone'ing] where rectangular slits become a trapezoid [...like a swung open door].
    4. So long as the differential [slit] image scale and the horizontal keystone'ing are small enough...you won't notice the mismatch in a print.
    The slit mismatching could only be corrected by resampling each individual slit with the swing angle and scale factor applied. Software for doing that could be obtained in photogrammetric "ortho-rectification" packages...costs more than any X-system still.

    I won't bother trying the panoramic mode...well-knowing the result is less perfect than what I'd like.
     
  15. Shadowside

    Shadowside Good Glass is Forever...

    Joined:
    Jun 24, 2013
    Messages:
    2,168
    Likes Received:
    1,818
    Location:
    Winnipeg, Canada

    -Return to Top-

    Nodal Adjustments are referenced here @ 1:50



    They also mention another video they do that goes into how to calculate this in greater detail.
     
  16. Finder

    Finder Premium Member

    Joined:
    Aug 31, 2017
    Messages:
    455
    Likes Received:
    698
    Location:
    Florida, USA

    -Return to Top-

    Some of it is parallax. Some of it is from a tilted camera. The results are more to do with your software.

    I have shot tons of panoramas, many of which hand held. I have never seen errors this bad. I would have a go with a trail version of Photoshop. I would also make sure the WB is equal for each image.

    These are five-frame pano swung at the center of the camera body:

    [​IMG]

    [​IMG]
     
  17. Frankie

    Frankie Premium Member

    Joined:
    Mar 31, 2011
    Messages:
    3,544
    Likes Received:
    640
    Location:
    The Left Coast

    -Return to Top-

    I'd watched the GigaPan video...missing was how to align off-set tripod mount to the GigaPan horizontal rotation axis. [Yes...I know you could buy an RRS Arca base plate for some X-camera models.]

    Also tricky is aligning the lens perspective centre [...they call it "nodal" point]...which differs from lens to lens. Using the diaphragm plane is a useful trick but not visible in the lens barrel...:(

    There are surveyors' techniques in aligning both...another long story...
     
    Last edited: May 16, 2018
  18. Goldingd

    Goldingd Premium Member

    Joined:
    Jan 20, 2014
    Messages:
    935
    Likes Received:
    280
    Location:
    Anaheim CA

    -Return to Top-

    Hmm, apparently I was computer challenged this morning, following was not posted.

    Some points:

    In first image, clearly your tripod is not properly leval, resulting in stepping.

    In second image, I question if you had enough overlap between images. Actually perhaps in first image as well. You need at least 1/4 overlap, but most would go for 30% overlap. Stitching software can have big issues without enough overlap
     
  19. Goldingd

    Goldingd Premium Member

    Joined:
    Jan 20, 2014
    Messages:
    935
    Likes Received:
    280
    Location:
    Anaheim CA

    -Return to Top-

    So, getting back to your posting. Specifically looking at your second example. The curved edges, very apparent at the bottom is an issue with parallax, although even fantastic lenses perfectly place on your rig can still show this to some extent. The odd stitching, where parts do not align, is not (IMO) but is caused by lack of overlap in the individual shots. In your second example it looks like two rows by three columns, total of six shots, with no or very little overlap. The lack of overlap prevents the software from being able to find enough comoaality between shots to properly stitch them.

    Perhaps you can post the individual shots to demonstrate this, or to show that my assumption of how little overlap exists is wrong.
     
  20. Goldingd

    Goldingd Premium Member

    Joined:
    Jan 20, 2014
    Messages:
    935
    Likes Received:
    280
    Location:
    Anaheim CA

    -Return to Top-

    By the way, invest in a good tripod leveling base to make leveling the tripod easier and prevent that stepping issue. The apparent stepping seen in the second example, the patio, could have been handled by the software, although you would wind up cropping the result, also indicating lack of overlap.

    Also, yes, I have done panorama work, both with full frame Canon 5 D, and Fujifilm X-T2. With the X-T2 I have even gotten away with handheld panorama (RAF post processed in Lightroom), and no way can one properly pivot on the nodal point handheld. Parallel errors can give me issues with horizontal objects at the edges (like a bloody railing at the Grand Canyon) but what I see in your examples occurs when I fail to properly overlap.
     
    Last edited: May 16, 2018

Share This Page

  1. fujix-forum.com uses cookies to help personalize content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice