Pages: [1]   Go Down

Author Topic: Colorthink - unexpected results from export color list  (Read 1981 times)

shewhorn

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 537
    • http://
Colorthink - unexpected results from export color list
« on: June 10, 2010, 01:41:11 pm »

This is a bit of a cross post from Chromix's colorforums.com but I don't think that gets nearly as much traffic as LL so I wanted to see if anyone here has any experience with this. I've left a message with support but I figured the community here might have suggestions for alternative methods of getting a list of colors from a target. It's really easy with targets created for ProfileMaker as the reference file is in RGB and that can easily be pulled into a spreadsheet. Same goes for Colorport, just pull the .tab file into a spreadsheet as it uses the RGB values. Argyll however uses L*a*b* to specify target patches. I visited Bruce Lindbloom's site to get the equations to convert from Lab to XYZ and then to RGB but... yeah I'm just not that smart.   I figured I could just use ColorThink to extract the values from the TIFF file but in doing so I noticed WAY more colors in there than should have been. With other targets you can do this and clearly see some patterns emerge but it doesn't work nearly as well with Argyll targets. Perhaps it would work better if I uprezzed the file (haven't tried that) as there may be less sample points falling on transitions. Anyhow, here's the original post:

_______________

I'm running ColorThink 2.2.1b5 on OS X 10.6.3. I'm trying to create a color list that I can look at in a spread sheet so I can see what Argyll is doing when generating patches. The first attempt I noticed that there were a lot more colors than there really were (this is a 1728 patch target) so I took a closer look and noticed that there is a 1 pixel transition between some, but not all of the patches. I used a fixed size marquee in Photoshop to delete those transitions and tried again but still I'm getting very unusual results. In the spreadsheet I've sorted by ascending R,G,B values and have filtered out the duplicates. I should in theory have 1728 discrete colors but it's showing 4200 discrete values. (I generated a generic target using targen -v -d2 -G test1728, and printtarg -v -iSS -pA4R -t test1728). The target definitely only has 1728 different colors in it.

I also tried this on a target which was generated by Monaco Profiler. MP is generating predictable evenly distributed patches. You can see some of that yet the color list is showing FAR more colors than are actually there in the target.

As a test I created a file that was 1200x800 pixels. I filled 400 pixel increments with red, green and blue respectively, saved and then extracted the colors. What I would expect would be 6600 samples but only 3 different colors that were 255,0,0; 0,255,0; and 0,0,255. What I ended up with (after filtering out the duplicates) was:

0,253,0
5,169,77
16,0,230
85,169,0
255,0,0

I guess I expected that "export color list" would generate a list of every discrete color in an image. I guess in a regular photograph that might be impractical as that could be quite a large number so after thinking about it for a while... it seems as if it may be just taking samples at some pre-determined x,y spacing but with an aperture that's larger than 1 pixel and this is why I'm seeing more than 3 colors in this particular test.

Would that be accurate?

Cheers, Joe
Logged

shewhorn

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 537
    • http://
Colorthink - unexpected results from export color list
« Reply #1 on: June 10, 2010, 02:09:26 pm »

Well, a few seconds after I posted my phone rang. Chromix said basically yes, that is the case (and btw, uprezzing the target actually did work to filter out some of the transitions). Colorthink Pro could actually solve the problem. It's a bit tedious but it allows you to use an eyedropper to define points for creating a color list so that's one option.

Cheers, Joe
Logged

Czornyj

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1948
    • zarzadzaniebarwa.pl
Colorthink - unexpected results from export color list
« Reply #2 on: June 10, 2010, 02:15:50 pm »

Please excuse me, but I simply don't get it - what is the sense of using L*a*b values to specify patches of target, that is suppose to profile a RGB output device?
Logged
Marcin Kałuża | [URL=http://zarzadzaniebarwa

shewhorn

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 537
    • http://
Colorthink - unexpected results from export color list
« Reply #3 on: June 10, 2010, 03:15:34 pm »

Quote from: Czornyj
Please excuse me, but I simply don't get it - what is the sense of using L*a*b values to specify patches of target, that is suppose to profile a RGB output device?

It doesn't make any sense at all, and upon closer inspection, it definitely isn't L*a*b* (thanks for shakin' up the brain bucket :-) ) The labels in the ti1 and ti2 files are RGB_R, RGB_G, and RGB_B which would make sense except for the fact that white is not represented by 255, 255, 255. I didn't think to look for white before but I verified where it falls in the .ti1 file and upon closer inspection white is represented as 100,100, 100. Seeing no values in the vicinity of 200 and up lead me to believe that it was maybe Lab but the a and the b values go positive and negative +/- 127 and there was nothing above 100 or below -100.

With Argyll you first specify a target and it generates the data for that, then you actually run a second command that generates that target. That target can be 8 bit or 16 bit so representing RGB values as a percentage kind of makes sense as you can derive an 8 bit or a 16 bit chart from that set of data (although you could do the same by having the traditional 0-255 values as floating point data and then just round... not quite that simple but that's the basic idea). So I think that's what's going on... the RGB data is being represented as a percentage.

SO... with that I can say multiply by any RGB value by 255 and then by .01 and finally round off to create a target that is close to what the Argyll targets I've generated are. It'll take about 30 seconds to do in a spreadsheet.

Looking at the Argyll numbers they are definitely very different than what the Atkinson/Colorport/MeasureTool generated targets do (they aren't just linearly spaced colorimetric values). Whether or not they're any better I'm not sure. I can definitely say I've gotten some nice results out of Argyll but I'm not quite happy yet. It'll be interesting to see how much of a difference (if any) an Argyll generated target makes when using a different application as the rendering engine to build the profile.

Cheers, Joe
Logged

shewhorn

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 537
    • http://
Colorthink - unexpected results from export color list
« Reply #4 on: June 10, 2010, 03:42:33 pm »

Quote from: shewhorn
(and btw, uprezzing the target actually did work to filter out some of the transitions).

Ack... well how it actually works and what I observed are two different things. In theory uprezzing won't make any difference at all because either way Colorthink downrezes the file anyway. It may be that the dataset I chose to work with the second time around to test the uprezzing thing had a slightly different alignment, blah blah blah so that it appeared as if it worked better but in reality there's no possible way that could have made a difference.

The response from Chromix for anyone interested is here:

http://www.colorforums.com/viewtopic.php?p=3565#3565

Cheers, Joe
Logged
Pages: [1]   Go Up