Re: DPX woes with gm1.2 unreleased Q16

Hi! 

> > However, the image I get is very blown out and
"quantized" -- i.e. 
> > what should be a smooth grey ramp shows
"stairstep" aliasing on 
> 
> The basic problem is that the colorspace for a TIFF
file is 
> assumed to be a computer (similar to sRGB) space. So
when you 
> specify '-colorspace CineonLog' it converts from
computer RGB 
> to Cineon Log, which is not what you want since your
file was 
> already in the desired colorspace.  If you don't
specify 

OK... I think that explains the "blown out" bit...
I would have
converted it with CineonLog, then when displaying it in
Fusion,
converted it *again* for display...

So I guess what I should do to test this properly is:

 * Read in the testpattern .dpx file into Fusion,
 * Apply a CineonLog Log2Lin conversion, # Image is now
viewable.
 * Save as 16 bit TIFF, # Simulates output of render.
 * Convert 16 bit TIFF to DPX with gm,
 * Load the new DPX file into Fusion, 
 * Apply the CineonLog conversion as above,
 * And I should have two files that match!

And I just tried it and it seems that it does work. The DPX
files
themselves look a bit different, but once they've been
brought into
linear colourspace for display they seem to match. The big
difference of
course is that the reconstructed DPX file is clipped in the
same fashion
as the TIFF was when I did the conversion...

I guess that I shouldn't expect that the 16 bit rendered
TIFF files to
span the full range of DPX because they can't and still look
good on our
eight bit displays. (We're doing all our work with IFF/TIFF
and then
converting to DPX for Scratch -- and boy, do those TIFF's
look good! 

Does that seem correct?

> A feature film did successfully use GraphicsMagick to
convert 
> from 16-bit TIFFs to DPX in Cineon Log (matching
Technicolor 
> exactly) but in that case it seems that the TIFFs were
in 
> sRGB space.  The person who did it also made a small
tweak to 
> the colorspace transform to use a different gamma.

That feature film was "Everyone's Hero", and it
all worked very well! I
believe that you were kind enough to advise my predecessor,
Steve Sloan,
on it. Unfortunately, we're dealing with a different film
out house on
this new one and their calibration requirements are a little
different.
However, I'm hoping (once I get all the math stuff sorted
out in my
head) that it will wind up being simpler -- these guys seem
to stick a
bit closer to the Kodak/Efilm standards.

Thanks very much!

Jos'h

------------------------------------------------------------
-------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the
chance to share your
opinions on IT & business topics through brief
surveys-and earn cash
http://www.techsay.com/default.
php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Graphicsmagick-help mailing list
Graphicsmagick-helplists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/gra
phicsmagick-help

On Wed, 11 Apr 2007, Jos'h Fuller wrote:
> So I guess what I should do to test this properly is:
>
> * Read in the testpattern .dpx file into Fusion,
> * Apply a CineonLog Log2Lin conversion, # Image is now
viewable.
> * Save as 16 bit TIFF, # Simulates output of render.
> * Convert 16 bit TIFF to DPX with gm,
> * Load the new DPX file into Fusion,
> * Apply the CineonLog conversion as above,
> * And I should have two files that match!
>
> And I just tried it and it seems that it does work. The
DPX files
> themselves look a bit different, but once they've been
brought into
> linear colourspace for display they seem to match. The
big difference of
> course is that the reconstructed DPX file is clipped in
the same fashion
> as the TIFF was when I did the conversion...

The reason for the clipping is that the dynamic range of
computer RGB 
is less than Cineon Log.  Computer RGB does not support
"whiter than 
white" or "blacker than black".

> I guess that I shouldn't expect that the 16 bit
rendered TIFF files to
> span the full range of DPX because they can't and still
look good on our
> eight bit displays. (We're doing all our work with
IFF/TIFF and then
> converting to DPX for Scratch -- and boy, do those
TIFF's look good! 
>
> Does that seem correct?

Yes.

>> A feature film did successfully use GraphicsMagick
to convert
>> from 16-bit TIFFs to DPX in Cineon Log (matching
Technicolor
>> exactly) but in that case it seems that the TIFFs
were in
>> sRGB space.  The person who did it also made a
small tweak to
>> the colorspace transform to use a different gamma.
>
> That feature film was "Everyone's Hero", and
it all worked very well! I
> believe that you were kind enough to advise my
predecessor, Steve Sloan,
> on it. Unfortunately, we're dealing with a different
film out house on
> this new one and their calibration requirements are a
little different.
> However, I'm hoping (once I get all the math stuff
sorted out in my
> head) that it will wind up being simpler -- these guys
seem to stick a
> bit closer to the Kodak/Efilm standards.

I would appreciate a copy of your magick/colorspace.c file
so I can 
see the exact changes that Steve made.  I am pretty sure I
know what 
he did, but a copy of the file would help verify.

A way to specify the current colorspace of an input file
(other than 
via attached ICC CMS profile) is very much needed.  I will
see how to 
implement it most efficiently.

Bob
======================================
Bob Friesenhahn
bfriesensimple.dallas.tx.us, http://w
ww.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMag
ick.org/


------------------------------------------------------------
-------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the
chance to share your
opinions on IT & business topics through brief
surveys-and earn cash
http://www.techsay.com/default.
php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Graphicsmagick-help mailing list
Graphicsmagick-helplists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/gra
phicsmagick-help