hugin-devs team mailing list archive

[Seth Berrier <seth.berrier@xxxxxxxxx>]

I have run it against the devel tip I think (see comment #10) and you
can generate the input images from the files in #2 and #3 (plus
ptodummy) but to make things super easy here's as minimal of an example
as I could come up with.  Note that the resolutions are carefully tuned
to ensure the use of the coarse mask.

Use the attached images and then run enblend like this:

enblend -f 2200x1100 -o Zenith-BadPixels.tif -- Input0000.tif Input0004.tif
enblend -f 2200x1100 -o Wrap180-BadPixels.tif -- Input0000.tif Input0001.tif Input0002.tif Input0003.tif
enblend -f 2200x1100 -o Both-BadPixels.tif -- Input0000.tif Input0001.tif Input0002.tif Input0003.tif Input0004.tif

All three will give you bad pixels in one form or another (either near
the zenith, the 180 wrap or both).  Note that these will all be 32bpp
tif files.

You can work around this problem by adding --fine-mask as a command line
option (which disables the course mask and also disables seam
optimization in the devel build).

I also found that you can eliminate MOST (but not all) of the white
pixels by using NFT as the primary seam generator along with the fine
mask.  This still does seam optimization which gets disabled
automatically when you use the fine mask with graphcut.

Anyways, hopefully this will give you the same problem on your end.

** Attachment added: "Images.tbz"
   https://bugs.launchpad.net/hugin/+bug/1356551/+attachment/4189302/+files/Images.tbz

-- 
You received this bug notification because you are a member of Hugin
Developers, which is subscribed to Enblend.
https://bugs.launchpad.net/bugs/1356551

Title:
  White pixels at 180 wrap and zenith for 16/32bit HDR Pano

Status in Enblend:
  New
Status in Hugin - Panorama Tools GUI:
  Invalid

Bug description:
  When stitching my final pano for an HDR workflow I get a column of
  blank/white pixels near the equator at the 180 wrap and an entire row
  of blank/white pixels at the top.  Here's the basic pieces of my
  workflow:

  - Multiple exposures at 7 angles with samyang 8mm + pano head
  - Combined exposures to 7 HDR images using pfstools (sometimes exr files for 16bpp, other times tiff for 32bpp)
  - Use hugin to: mask out pano head arm, find control points, optimize positions, barrel and view and set output options
  - Stitch with a script using nona+enblend to equirec HDR image

  Can view example of the white pixels here:
  - you're facing the 180 wrap when it first comes up
  - zoom in at zenith to see the missing top pixel
  http://vr.rollerblading.es/pano/file?path=https://dl.dropboxusercontent.com/u/3340541/Panos/MuralLobby-Pano-LDR.jpg

  When I look at intermediate files I'm pretty sure it is enblend that is the source of the problem.  I've tried several different versions:
  - The stock version from the OSX binary on the source forge web site (v 4.1.1)
  - Enblend 4.1.2_2 compiled with macports
  - Enblend 4.1.3 compiled with macports (using my own updated portfile to grab the 4.1.3 tar ball)

  All three versions show the same problem.  I'll attach a screenshot
  pointing out the truant pixels as well as a minimal .pto file and the
  script I use for stitching.

To manage notifications about this bug go to:
https://bugs.launchpad.net/enblend/+bug/1356551/+subscriptions