I recently started using Hugin to create 360° panoramas of HDR images.
The images I am trying to stitch together are in OpenEXR format.
I noticed that cpfind hardly finds any control points on these HDR images.
If I convert the images to LDR first, then cpfind actually finds quite a few control points.
I poked around in cpfind and found out that this is probably because the HDR images are first mapped
to [0,255] before any features are detected. As this is done using a linear mapping the images end up being
mostly black with only the brightest parts in the image visible. (Usually the sun...)
When I switched the mapping in cpfind to logarithmic in the source code, the results became much better and
more control points were found.
Is this a known issue, or am I doing something wrong?
I am using hugin on ubuntu 16.04 64bit.
I built hugin myself from hg revision:674a88cfacb6468ac185c8e12392ba9c206a7113
-- Torsten