Pages: [1]   Go Down

Author Topic: Question about the fuzziness slider in Photoshop's color range tool  (Read 1382 times)

Jeffrey Saldinger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 257
    • Jeffrey Saldinger

I am aware that when the fuzziness is set to 0%, the selection has hard edges at 82 and 164 (for the example shown in the screenshot), and as the fuzziness setting is increased, the range of the image’s tones over which the selection goes from 100% selected (white) to 0% selected (black) increases.

My question is: What does the slider measure a percentage of, exactly?  Or, to put it another way, for a given setting of the fuzziness slider (e.g. 33%, 67%, or 100%), how can I know the image’s tones that are the “first” (on either side of the range setting) to be 100% unselected?

Perhaps I should have add that my interest is in making selections of BW photographs to use for making layer masks for certain effects.  I gather that in a color image, the different channels might fall into 100% unselected at different distances from the given 100% selected range, but this doesn't interest me now.
Logged
Jeffrey
Astoria, New York
www.jeffreysaldinger.com

nirpat89

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 661
    • Photography by Niranjan Patel
Re: Question about the fuzziness slider in Photoshop's color range tool
« Reply #1 on: July 23, 2018, 12:46:00 pm »

By my measurements on 8 bit, R=G=B:

0%:     82-164 selected
33%:   57-189 selected
50%:   44-202 selected
67%:   31-216 selected
100%:  3-243 selected

Now you make sense out of that.... :)
« Last Edit: July 23, 2018, 01:14:47 pm by nirpat89 »
Logged

Jeffrey Saldinger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 257
    • Jeffrey Saldinger
Re: Question about the fuzziness slider in Photoshop's color range tool
« Reply #2 on: July 24, 2018, 06:58:37 am »

Thank you, Niranjan. Can you please take a minute or two to describe how you made the measurement?

I can make some sense of your numbers by observing that 82-3=79=243-164, so if you take e.g. 50% of 79 (=39/40 if only integral values are used) and subtract that from 82, you get 43/42, which is close to your measured 44, and 164 plus 39/40 is 203/204, which is close to your measured 202. I gather that in your 50% example, 43 and 203 are the first tones in the mask that would be fully black.  For the other fuzziness percentages the arithmetic would be the same.

So the fuzziness setting seems to measuring the fraction of the distance (measured in tone steps) between a bit lighter than 0 and the low point of the range which, in the special nearly-symmetric case I used, is the same as the distance between the range’s high point and a bit darker than 255.

By the way, when I tried doing color range midtones 120-136 with fuzziness 100% on a black-to-white gradient, the transition in the mask from white to black (just viewed on the screen, not measured in any way) was a long gradual gradient, which points to the implicit assumption in the preceding paragraph that it’s not the width of selected range that determines what the fuzziness percentage represents but rather the distance of the low and high points to black and white, respectively.

This is a useful result for me.  I feel an emerging curiosity about what happens with asymmetric ranges, which maybe I can look into more carefully before too long

Thanks.

Jeffrey
Logged
Jeffrey
Astoria, New York
www.jeffreysaldinger.com

nirpat89

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 661
    • Photography by Niranjan Patel
Re: Question about the fuzziness slider in Photoshop's color range tool
« Reply #3 on: July 24, 2018, 10:39:59 am »

I used a 256 steps stepwedge.  For a measurement, make the intended selection, then add a Curves layer so the selection becomes a mask.  Modify the Curves layer with one of the colors jacked up high, say R channel.  Then using use the color dropper to see where the input R not same as output R from each side.  If you use 16 bit on the info palate, you will get better accuracy.  There is probably a better way to do it.  This is cumbersome if you were to do a detailed study of it.  You can also try just using the shadpws or highlights where the fuzzying would be only on one side to gain a better insight.  I would have thought they used some sort of Gaussian function centered around the selection edges. 


(Google Peter Mhrar, he has a downloadable stepwedge that you can use.)


:Niranjan.


Edit:  On a second thought, an easier way would be to go into quick mask mode after selection and just read the mask densities.  No need for a Curves layer.
« Last Edit: July 24, 2018, 11:29:51 am by nirpat89 »
Logged

Jeffrey Saldinger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 257
    • Jeffrey Saldinger
Re: Question about the fuzziness slider in Photoshop's color range tool
« Reply #4 on: July 24, 2018, 06:27:53 pm »

Niranjan, this is terrific.  Thank you.

I did download the Mhrar stepwedge (256 steps) and was playing around in Ps along the lines you described.

Using a color fill layer with the mask based on the color range settings gave a very clear reading of when the mask turns to full black or white.  I just rode the info tool cursor over the squares until the info panel showed black or white for the first time.

Your surmise that a Gaussian distribution is at play here undercuts my original hope that the answer would be something readily calculated algebraically from the range and fuzziness values set in the color range tool, but your method with the Mhrar stepwedge is a wonderful alternative.

Thanks again.

Jeffrey

Logged
Jeffrey
Astoria, New York
www.jeffreysaldinger.com

nirpat89

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 661
    • Photography by Niranjan Patel
Re: Question about the fuzziness slider in Photoshop's color range tool
« Reply #5 on: July 24, 2018, 08:59:24 pm »

Niranjan, this is terrific.  Thank you.

I did download the Mhrar stepwedge (256 steps) and was playing around in Ps along the lines you described.

Using a color fill layer with the mask based on the color range settings gave a very clear reading of when the mask turns to full black or white.  I just rode the info tool cursor over the squares until the info panel showed black or white for the first time.

Your surmise that a Gaussian distribution is at play here undercuts my original hope that the answer would be something readily calculated algebraically from the range and fuzziness values set in the color range tool, but your method with the Mhrar stepwedge is a wonderful alternative.

Thanks again.

Jeffrey

Jefferey, You are welcome.  I had wondered about this issue myself some time ago.  So your query got my curiosity going again.   

I made a graph of what happens at one of the edges - looks half a Gaussian (or some other non-linear function) to me.  The selection was mid-tones 100 to 200 @50% fuzziness.   Even though it is non-linear I guess you can still come up with an empirical relationship as you were trying to do.


:Niranjan.





« Last Edit: July 24, 2018, 09:34:28 pm by nirpat89 »
Logged

Jeffrey Saldinger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 257
    • Jeffrey Saldinger
Re: Question about the fuzziness slider in Photoshop's color range tool
« Reply #6 on: July 25, 2018, 08:51:36 am »

So perhaps it’s the case that for a given range and fuzziness “all” one would need to do is put the values into some Gaussian function and get the answer.  Maybe that’s what’s happening under the hood after all, to get the mask we get.

But your solution with the Mhrar stepwedge strikes me as simpler and more fun, although I can imagine that others might enjoy figuring out whatever variables and constants one would need to make the calculation’s result match the observed stepwedge result.

Jeffrey
Logged
Jeffrey
Astoria, New York
www.jeffreysaldinger.com
Pages: [1]   Go Up