That is something I have been meaning to test. The perceptual rendering from i1Profiler is fantastic mostly. I would love to see if it can be bettered.
You test profile quality the same way you have been testing your profile conversions with Bill's balls and Andrew's picture. Plus make some prints. A lot of prints. Soft proofing helps a lot to catch the main behavior without wasting too much paper and time. Color geek software can be useful to provide some quantification of your results. You have posted a lot on that; I believe you know what to look out for, the usual suspects. It is nice to know the behavior of the various profiling engines, but no one is publishing this information, and it is ridiculously time consuming and expensive to test. I have been doing a lot of that last year, but i1Profiler has updated their color engine again.
Funnily enough, softproofing in Photoshop is not very precise but I don't know why no one has pointed it out before. Soft proofing in the Canon 16 bit plug-in is more accurate. Often I would see gradients soft proof as smooth in Photoshop, but print with slight banding. Sometimes there is tone and color shifts not present in the actual prints. Soft proofing in the Canon plug-in shows the banding, and the correct rendering of tone and color. There are other issues too, but that is a whole different topic...
Yes, Andrew's balls can be of some use for this sort of test alright. Still - it ends up being very specific and in the case of the balls we're dealing with 100% saturated colors in ProPhoto, so using an intelligent mapping isn't going to be much better than using the CMM (if at all). The real benefit will be when using a workspace like ProPhoto, but with colors that are not too far out of the destination gamut.
I think i1Profiler uses a combination relative/perceptual mapping for the perceptual mappings, possibly with colors not too far out of gamut being mapped perceptually and colors well out of gamut being mapped colorimetrically ... or something along those lines. It's entirely up to the profile maker what sort of algorithm to use and unfortunately XRite don't tell us how they do their perceptual mappings.
I'm not sure if I posted this elsewhere, but here is a Perl script to do the perceptual mapping using Argyll (call the script
perceptual.pl):
use strict;
use warnings;
my $inprofile = $ARGV[0];
my $outprofile = $ARGV[1];
my $intiff = $ARGV[2];
my $outtiff = $ARGV[3];
my $gam = $intiff; $gam =~ s/\.tif*$/.gam/; my $linkprofile = $intiff; $linkprofile =~ s/\.tif*$/.icm/;
print "Creating Gamut file\n";
#`tiffgamut -v -f90 -w -pj \"${inprofile}\" \"${intiff}\"`;
`tiffgamut -v -w -pj \"${inprofile}\" \"${intiff}\"`;
print "Making Link File\n";
`collink -v -dmt -qm -G\"${gam}\" -ip \"${inprofile}\" \"${outprofile}\" \"${linkprofile}\"`;
`del "${gam}\"`;
print "Making Perceptual TIF Image: ${outtiff} with embedded profile\n";
`cctiff -e \"${outprofile}" \"${linkprofile}\" \"${intiff}\" \"${outtiff}\"`;
#print "Making Perceptual TIF Image: ${outtiff} WITHOUT embedded profile\n";
#`cctiff \"${linkprofile}\" \"${intiff}\" \"${outtiff}\"`;You can call this script in Windows with this batch file (call the batch file
perc.bat):
rem Convert image from source to destination using an intelligent perceptual mapping.
rem Parameter 1 is the source profile (for example ProPhoto). The profile extention must be .icc (do not specify it)
rem parameter 2 is the destination profile (for example HPZ3100-CansonPhotoHiGloss). The profile extention must be .icc (do not specify it)
rem Parameter 3 is the image name. It must be a tif image. Do not specify the extension
rem Example: 'perc ProPhoto PZ3100-CansonPhotoHiGloss testImage'
rem the converted image will be called 'imageName-perc-destinationProfile.tif'
perl perceptual.pl %1.icc %2.icc %3.tif %3-perc-%2.tif The script takes around 2 minutes on my PC, so it's not exactly quick, but it does a lot of processing!
Robert
