[Octave-bug-tracker] [bug #63060] imread returns logical data for 8-bit TrueColor PNG image

[Rik]

Update of bug #63060 (project octave):

                  Status:               Duplicate => Confirmed              
              Depends on:                         => bugs #51264            

    _______________________________________________________

Follow-up Comment #8:

Actually, this is not a duplicate of bug #51264.  In that bug, the potential
fix would be in the read_images() function.  In this case, a fix would need to
occur in the get_depth() routine before read_images is called.

There are numerous bug reports about imread getting confused by data that
looks like it is either binary (0/1) or graylevel.  Alas, this is all about
the Imagemagick libraries which sometimes seem to mis-identify files, but
don't give programmers enough visibility to countermand those decisions.

Either we just live with this situation or we have to create a lot of point
exceptions for certain file types with certain configurations of bit depth,
color profile, etc.  None of it is very good.

There may be some correlation between when a file is DirectClass or
PseudoClass (indexed) and whether it is truly 1-bit data.  I will post a patch
for that here.  But it would need a fair bit of testing on multiple platforms
with different versions of Magick libraries.


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?63060>

_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/