Quantcast

troubles with 32bit float...

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

troubles with 32bit float...

Casey Connor-2
I'm having lots of issues with the 32bit tiffs that come out of G'MIC
CLI processing in Linux (Kubuntu 16.10 in my case).

They don't open properly in GIMP, or most any other software. There is
clearly something going wrong somewhere, but I'm not sure where to file
a report as I'm not sure where the root cause may be (GIMP, underlying
tiff libraries, or G'MIC, etc.) If anyone has any insight, I'd
appreciate it.

When I apply a -median_files filter with G'MIC, it generates a 32bit
float file that opens as all-white in GIMP. But I can change the
precision (in GIMP) from 32bit-float to any integer format, and the
image suddenly appears and I can work with it just fine. Here is a
demonstration image: http://caseyconnor.org/pub/image/32bitfloat.zip (145MB)

Can anyone reproduce that? I've asked here before without luck, but I've
seen on other forums that I'm not the only one with this issue. (This
happens in any recent version of GIMP on Linux or Windows for me, but
most recently tested in otto-kesselgulasch repo commit 2944dbd.)

I tried to generate a smaller simple example starting with a 100x100
pixel image, 16bit integer precision (same thing happens with 8bit): I
make a simple gradient, save it to png or tif, process through G'MIC,
and open the result in GIMP. Result: image is all white, and in this
case changing precision does *not* make the real image appear -- I can't
see the image no matter what I do. Here is a zip file with the original
and the gmic processed version that won't open in GIMP:
http://caseyconnor.org/pub/image/smallexample.zip

The KDE image viewer (gwenview) shows the image differently every time
it opens it: sometimes a corrupted version, sometimes a line across a
black background, and the line changes color every time the image is
opened. (I mention this only in case it points at some system library
issue.)

If I open the 32bit float G'MIC images in Hugin, they open fine, but the
resulting 32bit float output from Hugin can not be opened in or
processed by GIMP in linux or windows.

AN important note: I can open all of these images (post-G'MIC, big image
or small 100x100 image) in Luminance HDR: it seems to have no trouble
with them.

Any thoughts? Are there GIMP bugs here to be addressed? Can anyone
reproduce this?

I've had some success using G'MIC in 16bit integer mode (by appending
",ushort" to the file name); my main goal here is to sort these bugs out.

Thanks!


_______________________________________________
gimp-user-list mailing list
List address:    [hidden email]
List membership: https://mail.gnome.org/mailman/listinfo/gimp-user-list
List archives:   https://mail.gnome.org/archives/gimp-user-list
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: troubles with 32bit float...

Ken Moffat
On Wed, Apr 26, 2017 at 06:01:05PM -0700, Casey Connor wrote:
> I'm having lots of issues with the 32bit tiffs that come out of G'MIC CLI
> processing in Linux (Kubuntu 16.10 in my case).
>
> They don't open properly in GIMP, or most any other software. There is
> clearly something going wrong somewhere, but I'm not sure where to file a
> report as I'm not sure where the root cause may be (GIMP, underlying tiff
> libraries, or G'MIC, etc.) If anyone has any insight, I'd appreciate it.
>

The problem with tiff is that it isn't a simple format, it's a
specification which has been described as Thousands of Incompatible
File Formats (thanks, wikipedia).  When I was trying 16-bit images
with gimp-2.9 using ufraw, I used to see lots of messages about
unexpected features in the file format.  In the end I moved to pngs,
and for my amateur pics 8-bit is actually good enough.

So the first question is: can you use a different file format (I'm
only minimally familiar with G'MIC, used as a plugin, and I have no
real understanding of what you are doing) ?

> When I apply a -median_files filter with G'MIC, it generates a 32bit float
> file that opens as all-white in GIMP. But I can change the precision (in
> GIMP) from 32bit-float to any integer format, and the image suddenly appears
> and I can work with it just fine. Here is a demonstration image:
> http://caseyconnor.org/pub/image/32bitfloat.zip (145MB)
>

I recognize that G'MIC is not only for the gimp, but since it works
as a plugin working on opened images - can you use the plugin to do
what you are doing ?  If not, I would be tempted to ask at G'MIC
because they seem to be active and some of their output clearly
works in the gimp.  If they say it's not their problem, perhaps they
can narrow down whose problem it might be.

Sorry this probably isn't a lot of help.

ĸen
--
I live in a city. I know sparrows from starlings.  After that
everything is a duck as far as I'm concerned.  -- Monstrous Regiment
_______________________________________________
gimp-user-list mailing list
List address:    [hidden email]
List membership: https://mail.gnome.org/mailman/listinfo/gimp-user-list
List archives:   https://mail.gnome.org/archives/gimp-user-list
Loading...