[argyllcms] Re: LUT versus XYZ monitor profile

  • From: Nikolay Pokhilchenko <nikolay_po@xxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Wed, 23 Dec 2009 00:25:19 +0300

Graeme Gill wrote:

> Nikolay Pokhilchenko wrote:
> > I confirm that generally XYZ-LUT profiles are tend to significant banding 
> > in real use.
> > The banding is preventing the use of -ax key for display or camera 
> > profiling. The
> > quality of such profiles is unacceptable (ArgyllCMS V1.0.4. I've didn't try 
> > -ax with
> > latest versions). IMHO, much more smoothing is needed for XYZ data in the 
> > darks, may be
> > in overall.
> How are you testing this ?
> Can you reproduce the problem using imdi/cctiff ?
> (ie. are you sure it's not a CMM issue ?)
> Graeme Gill.

OK. I've found my display data and have done an experiment with 

1. timage -t -s -x -r300 TST.tif
2. colprof -v -ax -qu vaio
3. profcheck -k -x -w -e vaio.ti3 vaio.icm
   errors(CIEDE2000): max. = 0.059043, avg. = 0.010457, RMS = 0.011356
4. cctiff -v -p -ir sRGB.icm -or vaio.icm TST.tif TST-ax.tif

5. colprof -v -al -qu vaio
6. profcheck -k -x -w -e vaio.ti3 vaio.icm
   errors(CIEDE2000): max. = 0.660557, avg. = 0.132991, RMS = 0.161808
7. cctiff -v -ir sRGB.icm -or vaio.icm TST.tif TST-al.tif
8. Comparing the resulting images

Take a look at *.jpg attached. At the top - the inverse XYZ-LUT result, at the 
bottom - Lab-LUT result. Both are desaturated equally to fit any display gamut.
The error values between profiles are differs more than an order (x10) with the 
same instrument and the same device. I didn't find appropriate display and 
instrument noise in -ax profile error check results. The smoothing for -ax may 
be insufficient. IMHO, the smoothing in perceptual Lab-space is quite good 
thing, but the same smoothing in non-perceptual XYZ-space may playing a bad 
trick on resulting wedge banding. IMHO.

Attachment: -ax(top) vs -al(bottom).jpg
Description: JPEG image

Other related posts: