Loading ...
Sorry, an error occurred while loading the content.

Re: About conversion from equirectangular to cubic faces (rectiliniar)

Expand Messages
  • dorindxn
    Hi Jim thanks for reply, yes you re absolutelly right. Maybe even shoots pointed to the corners of the cube, however that was given as an example, what I
    Message 1 of 17 , Dec 7, 2008
      Hi Jim thanks for reply, yes you're absolutelly right. Maybe even
      shoots pointed to the corners of the cube, however that was given as
      an example, what I intend to use conevrsion formulas to adjust the
      cubic faces at the edge to make the seam less visible/distracting,
      but with the cubic faces generated direct from PTgui/Hugin of course
      after careful optimise/align with correct lens parametter, so on,
      using a precise pano head, in order to jump over one interpolation,
      as I said.


      --- In PanoToolsNG@yahoogroups.com, Jim Watters <jwatters@...> wrote:
      > There is always some error when aligning images to create a
      panorama. By
      > blending these images together the errors are hidden. The methods
      > have to of blending two images together will put those seams at
      > different places if the two images are cropped in different ways
      even if
      > the overlap does not change. If the blending tool is not told that
      > seams at the left and right of a equirectangular should blend when
      > panorama is displayed in a viewer that seam will be visible and
      > If you just crop out the 90x90 image from the 6 input images to
      form the
      > cube there will be very visible seams at every edge. There are 12
      > edges. If you crop the images to leave a little overlap (95X95)
      and let
      > the viewer do the blending then you are removing overlap that might
      > prevent partial ghost from showing up. But you are still running
      > images through the optimizer and extracting the image to be used.
      > just allowing the view to possibly show blending errors at the
      > The examples you are displaying with DevalVR without stitching
      shows the
      > difference.
      > http://dorin.devalvr.com/no_stitch/indoor_6shoots/tokina10FF.html
      > If you still want to blend the images together, but you want to
      > first generating a single full spherical image, then the best thing
      > to make sure that each edge of the cube is covered by a single
      > To mostly do this. Instead of pointing the camera to each side of
      > cube. For the 4 around point to to the edges. This will hide some
      > the seams but not all of them. In the long run it will likely take
      > longer to longer to generate. the stitcher and blender will be
      > the same image multiple times.
      > Having the entire sphere in a single image for post processing is
      > easier. The equirectangular image works well for this.
      > Jim Watters
      > dorindxn wrote:
      > > Hi Erik, now is my time to not understand you both times because
      > > my English I'm sure :(
      > >
      > > Again is about stitching directly to cubic faces, like, for
      > > one shoots 6 images with fisheye each shoot in direction
      > > with each cube face then remmap and crop those as rectilinear
      > > then to blend, mean to adjust the near border region of each to
      > > the seam invisible at displaying.
      > >
      > > Dorin
      > >
      > > --- "Erik Krause" wrote:
      > >
      > >> ....
      > >> However, to blend cubefaces might be a bit easier, since there
      > >> less anisotropic distortion. In a cube you have 12 joints
      > >> cubefaces.
      > --
      > Jim Watters
      > jwatters @ photocreations . ca
      > http://photocreations.ca
    Your message has been successfully submitted and would be delivered to recipients shortly.