check the options bar. make sure you’re in "Normal" mode, and check the opacity and fill.
David,
Your question is basically the same as the one I raised a few topics earlier "Luminosity in PS", despite your asking about the desaturate mode vs my question on saturate mode. I am working in PS 7, by the way. In BOTH modes, the Sponge tool preserves luminosity and hue.
To prove that hue is fixed, read before and after hue values in the info palette with HSB values selected in addition to RGB.
To prove that luminosity stays constant, select a square swatch and read its mean luminosity from Image > Histogram before sponging. With the Sponge tool active, use a circular brush inscribed in the square, i.e., large enough to occupy most of the square’s area, and click the mouse once. Reselect the square and recheck its luminosity. It will not have changed. But the circular area sponged will be darker (in desaturate mode) or lighter in saturate mode.
With the brush still centered in the square, click the mouse again and observe another step in the saturate/desaturate process, with luminosity still constant. In the desaturate mode you are interested in, the high color channel value drops, the low channel rises, and the middle channel adjusts one way or the other always keeping luminosity and hue constant. The desaturate process continues to step with each mouse click until all channels are equal—that is, FULLY desaturated. And the result is darker as you have noted.
We both want to know why the result looks darker (in desaturate mode) and lighter (in saturate mode) despite LUMINOSITY being constant—since luminosity in PS weights the channel values according to the human eye’s perceived brightness response. Since luminosity doesn’t change when sponging, how come the apparent brightness changes?
George
Thank you for the info, this helped. Please look at <
http://www.davidgraphics.net/photoshop/red_to_gray.html> for my follow-up reply. It contains an illustration which shows what happens to red as the color is gradually removed, and what shade of gray results once all color is gone.
David
At your site you state the "correct" gray value after totally desaturating 255/0/0 is 128/128/128. It’s unclear what brought you to this conclusion.
If luminosity is held constant, the correct gray values after fully desaturating pure red (255/0/0), pure green (0/255/0), and pure blue (0/0/255) would be 76/76/76, 150/150/150, and 28/28/28, respectively, using the R,G,B channel weighting factors of 0.3, 0.59, and 0.11—which PS does use.
George
Dave
"…check the options bar. make sure you’re in "Normal" mode, and check the opacity and fill…"
With the Sponge tool the only "mode" choices on the Options Bar are "Saturate" and "Desaturate". There are no blending options. There is no opacity option either—just a "Flow" adjustment.
Possibly you are thinking of the layer properties, but these are on the layer palette rather than the options bar, and the behavior cited occurs when the layer mode is Normal and its opacity is 100%.
George
George
I concluded that fully desaturating 255,0,0 results in 128,128,128 by going into Paint Shop Pro, and in the color picker window, starting with 255,0,0 and reducing saturation to zero. Sounds like I’m not correct, oh well, wouldn’t be the first time. I don’t plan on checking this particular thread anymore, but if you wish to contact me, I’m at Thanks.
David
correct goerge. sorry for the misinfo. didn’t have ps in front of me at the time.
hmm… interesting. 255,0,0. image> adjustments> desaturate leaves 127,127,127. no further desaturation is possoble with the sponge tool.
yet again start with 255,0,0 and use the sponge tool on it repeatedly until it stops "desaturating" (no feather brush, keep clicking until the image under the brush tip doesn’t change anymore) yields 78,75,75. considerably darker than a straight image> desaturate. using image> adjustments> desaturate at this point show that this is true (in case I didn’t believe the info palette<g>).
too bad dave k. bowed out. this could get interesting. it’s certainly beyond the scope of my ps color abilities to explain, but I’m sure someone here will come up with a reason for it.
dave
Dave,
I’ll be repeating myself, but here goes:
What PS is doing with the Sponge tool is keeping the LUMINOSITY constant (and the hue as well) while desaturating.
"Luminosity" is not an idle, subjective property—In PS it is precisely defined as the weighted sum of the component color channel values. It ranges from 0 to 255.
There is a multiplier for each channel value derived from experimental data on the eye’s perceived "brightness" when stimulated by radiation of given physical intensity (watts). The eye is more sensitive to green, so a lower wattage in green is needed to get the same sensation of brightness as red or blue. The eye is less sensitive to red and least sensitive to blue.
The weighting factors applied to channel color values in calculating luminosity used in PS are 0.30, 0.59, and 0.11 for R,G, and B. Note that they add up to 1, so when all three color values are the same (total desaturation), say 138 for example, the luminosity is 138. When all three are 255 (white), the luminosity is 255. When red is 255 and green and blue are zero, the luminosity is 0.3 times 255 + 0 + 0 = 76. etc, etc. The weighted contributions from each channel are added to get the resultant luminosity. (You can read the luminosity of a uniform swatch by selecting it and going to Image > Histogram.)
Keeping luminosity constant thus entails unequal changes in the channel values as the highest value is lowered and the lowest value is raised. Only those two are needed to control saturation, but the third intermediate channel gets involved to keep hue constant as the highest and lowest values move toward each other when desaturating. The algorithm used to do this is unfailing, and you will find no change in either luminosity or hue when applying the Sponge.
What bothers me is that the perceived brightness changes radically despite the constant luminosity, even when the before and after swatches are compared in grayscale. And that contradicts my notion of real world "luminosity" which is supposed to indicate perceived brightness. That’s where I am seeking solace. 🙂
George
Thanks george. I appreciate it.
What PS is doing with the Sponge tool is keeping the LUMINOSITY constant (and the hue as well) while desaturating.
So what does the the image> adjustments> desaturate command do? and why does it do it differently?
What bothers me is that the perceived brightness changes radically despite the constant luminosity… And that contradicts my notion of real world "luminosity" which is supposed to indicate perceived brightness. That’s where I am seeking solace.
yea, that’s got my panties in a bunch too! 🙂
thanks for the explination (again!)
dave
cont. sorry, didn’t wish to edit…
seems as if (to me, the color noob) that the image> adjustments> desaturate is correct as it’s removing the color and leaving everything else alone, where as the sponge tool in desaturate mode is not only desaturating, but burning/darkening as well, as dk noted in the original post.
just wanted to clarify my muddle-mindedness.
dave
Dave,
"…seems as if…the image > adjustments > desaturate is correct…"
It’s not that either one is "correct". They are different.
Image > Adjustments > Desaturate sets all channel values to the average of the highest and lowest, without regards to the middle channel. It doesn’t gradually decrease saturation—it goes all the way to total desaturation in one click. Preserving hue is not a consideration, since hue is obliterated entirely. And there is no attempt to preserve luminosity, which changes to whatever results from simply averaging the high and low values. This feature is relatively crude and arbitrary.
The Sponge tool in Desaturate mode doesn’t totally desaturate all at once. It lowers the saturation stepwise while keeping luminosity and hue unchanged at all times
It’s not that either one is "correct". They are different.
I guess that’s a factually true statement. What I meant was that the I> A> Desat. gives you (and by that I mean me <g>) what’s visually expected from "desaturation".
The Sponge tool in Desaturate mode doesn’t totally desaturate all at once. It lowers the saturation stepwise while keeping luminosity and hue unchanged at all times
But when you desaturate 100% with a hard brush doesn’t that mean completely desaturated per your previous stmt (following)?
It doesn’t gradually decrease saturation—it goes all the way to total desaturation in one click. Preserving hue is not a consideration, since hue is obliterated entirely.
So I would THINK that 100% desaturation would yield the same results as I> A> Desat., but it doesn’t. It gets darker, as I said, like a burn. Theoretically 100% desaturate in a single click with the sponge tool should take that area to 127, 127,127, then that area is completely desaturated (hue removed, neutral gray… is my definition wrong?) But then click again and it keeps going. And going. And going, ad infinitum (or to black, whichever comes first!<G>)
So in addition to hue, it’s messing with luminosity, as you said. But why? Is it supposed to do that? I understand there’s a corolation to the burn and dodge tools in PS to the photography world, does the same kind of thing exist with the sponge tool? If so, it could be that not coming from that enviornment I just seem to have a different expectation (an incorrect expectation) than what a pro photographer would have.
I don’t want to try your patience george and I appreciate you taking the time to try and suss this out. I’m not really obsessed over this, but it is a bit of a curiosity. I actually hadn’t even noticed it until this post. I’m really deficient in this area and I go by what looks good. I’m not really up on the math behind the calc’s and what relationship they have on the image. So if I say anything foolish, please excuse me in advance. 🙂
Thanks, dave
Correction: you said the sponge tool keeps luminocity. I misquoted you. Sorry. But that’s not what I’m seeing. (Or at least, that’s not what I THINK I’m seeing).
Dave
I think the term "desaturate" means to make a color less saturated,i.e., to move toward but not necessarily reach total desaturation. Total desaturation is achieved when all channel values are equal, whether that be 255/255/255 (white), 0/0/0 (black), 37/37/37 (a very dark gray), or whatever. All colors comprised of RGB components that are equal are equally and totally desaturated no matter what the value. So a totally unsaturated color (gray) can be of any brightness or luminosity.
A color can never be totally saturated unless one of the component colors is zero. If one (or two) color values are zero, the color is 100% saturated no matter what the other color(s). So a fully saturated color can also be of any brightness, but can never reach 255 in luminosity, because one component is missing. If the zero component is blue, and red and green are 255, the luminosity is 255x(0.3+0.59+0) = 227 and that is as high as luminosity can get for a 100% saturated color. Note that "brightness" in PS is the value of the highest single component. 255/0/0 is 100% bright to PS but not at max luminosity. In fact its luminosity is only .3 x 255 = 76 on a scale from 0 to 255.
The Sponge tool in desaturate mode stops doing anything as soon as all channel colors are equal. It does not continue until black is reached, as you indicated. In saturate mode, the sponge tool stops its action completely as soon as any one channel reaches 0 or 255. So if one channel is at 0 or 255 to begin with nothing (almost nothing) happens. It is curious that I had to throw that caveat ("almost") in because there is some small change at the outset but it is soon damped out.
I really don’t think the sponge action was ever intended to go all the way to the unsaturated state. I think it was meant to be a "tweaker" with the scenarios in which invoked not comparable to I>A>Desaturate. Note also that I>A has no action corresponding to the Sponge in SATURATE mode. Any correspondence is between the Sponge in DESATURATE mode and I>A>Desaturate.
George
So a totally unsaturated color (gray) can be of any brightness or luminosity.
Hmm… that makes sense. Ok.
The Sponge tool in desaturate mode stops doing anything as soon as all channel colors are equal. It does not continue until black is reached, as you indicated.
No, that’s true, but not what I said. What I said was that applying the sponge tool a second, thrid and n-th time (repeated clicking the same area) will continue to darken an already completely desaturaed area until it goes black. If the area is already 100% desaturated after the first application (click), how can succesive clicks have any effect on it?
I really don’t think the sponge action was ever intended to go all the way to the unsaturated state. I think it was meant to be a "tweaker" with the scenarios in which invoked not comparable to I>A>Desaturate.
I’m guessing that too. One practical outcome of this discussion is my thinking that the sponge tool isn’t the best way to defeat red eye. I mostly use the channed mixer method anyway, but sometimes I cheat and hit the quickie with the sponge tool. Now I see I may be making the eye darker than it really should be.
Wonder if and Adobe engineer could pop in here and tell us what the sponge tool is SUPPOSED to do. I guess what I’m seeing is normal, I’d just like to know why it’s normal.
thanks george, dave
"…What I said was that applying the sponge tool a second, thrid and n-th time (repeated clicking the same area) will continue to darken an already completely desaturaed area until it goes black…"
(1) Change "UNTIL IT GOES BLACK" to "UNTIL IT GOES GRAY" meaning all channels EQUAL. To say black means all channels are not only equal but zero as well. Black (0/0/0) will never be the final product because the Sponge tool brings the tones toward each other so, even if the lowest channerl starts at zero it will move upward as the high channel moves lower and the channels will eventually meet at some common, non-zero (non-black, therefore gray)value.
(2) Delete "will continue to darken an already completely desaturated area until it goes black". As soon as the area becomes "completely desaturated" it is GRAY (all channels equal—the definition of totally desaturated)and all further sponging action stops.
Dave, there is no question what the Sponge tool does. An Adobe rep is not going to change the fact that the sponge holds luminosity and hue constant while altering saturation. That’s exactly what the sponge does, period . Let’s take it from there.
George
100x100px
fill with 255,255,255
set hard edge 80 px brush for sponge tool
set to desaturate. set at 100%
first click info shows: 220,15,15
second: 192,27,27
thrid: 170,37,37
etc.: 152,46,46
137,51,51
125,56,56
116,60,60
108,63,63
102,66,66
97,68,68
73,69,69
90,70,70
87,71,71
85,72,72
84,73,73
83,74,74
82,75,75
81,75,75
79,,75,75
78,75,75
hmm.. you’re right. stops there. so why is that gray so much darker than the I>A>Desat. gray, which yeilds 127,127, 127 on the same image described above? and why doesn’t the first click NOT yield the final value of 78,75,75 when i desaturate at 100%?
I think the answer to my question lies in my not understanding exactly:
the sponge holds luminosity and hue constant while altering saturation.
checking the same image with hue/sat adj. layer. turning down the sat to -100 (on the dialog’s scale) and not touching the lightness or hue also yields 127,127,127.
what is it that accounts for the rising g and b values when using the sponge tool… is it just a matter of they rise while the r falls and where ever they meet is "desaturated"?
one more (ok 2! <g>) test:
0,255,0 with the sponge tool bottoms out at 149,152,149. I>A>Desat. gives 127,127,127.
now again with 0,0,255. sponge tool goes to 25,25,28. I>A>Desat. gives 127,127,127 again.
just not sure what to make of all this. i can understand (and it makes more sense to me) that to fully desaturate any color may yield a different color (shade of gray) than desaturating another color, so maybe i’m looking at this from the wrong side. why does I>A>Desat. always seem to yield the same value? but if that’s so (which it can’t be) why does a full color image with a wide range of colors that you run I>A>Desat. on not become fully gray at 127,127,127? Is it becasue I’m playing with "pure" colors here? *
ok, answered my own question here. yes. it’s because I’ve been dealing with "pure" colors. 34,234,255. run I>A>Desat. and you get 144,144,144. the sponge tool multiclick method gives: 177,177,180.
still finding the differences a bit odd… i SEE what it’s doing, i just seem to have a block about WHY it’s doing it. as I said, i’m not losing sleep over it, but it’s interesting. i’ll understand if you want to drop out, if you think i’m getting in over my head. 🙂
thanks, dave
Dave,
Let’s take your example 34/234/255
Without doing anything, I can tell you what will happen with each tool
(1) Sponge Tool in Desaturate mode:
Luminosity = .3(34) + .59(234) + .11(255) = 10 + 138 + 28 = 176
Since luminosity stays constant no matter how many times you click the sponge, it will still be 176 when all channels are equal to some ultimate gray value we’ll call V. Then
176 = .3V +.59V + .11 V = V
But each channel equals V, so the final result is 176/176/176. This is precisely what you found.
(2) Image > Adjustments > Desaturate
This tool ignores the middle value. It looks only at the highest and lowest values. It takes their average. It makes all three channels equal to that average and that again is some value V. In this case:
V = (34 + 255) / 2 = 289/2 = 144. Again, this is precisely what you found.
Do you begin to get the picture? When you do, all the other questions in your last posting should be answered.
Hang in there. I’m not going away.
George
Dave,
You lied to me 🙂
You didn’t start that 100x100px swatch by filling it with 255/255/255
If it ended at 78/75/75 then you started with 255/0/0
For a starting color of 255/0/0, the initial and final luminosity would be .3(255) = 76, and the final gray would be 76/76/76. Close emough to 78/75/75 for me! Even at that, the luminosity of 78/75/75 is still 76:
L = .3(78) + .59(75) + .11(75) = 75.90
Sponging an initial color of 255/255/255 (white) will do absolutely nothing because, as I have said in an earlier posting, the sponge in desaturate mode STOPS when all colors are equal (fully unsaturated). So, in that case, the sponge stops before it even starts. It won’t do anything at all to white, black, or ANY shade of gray. Why would it, since the color is fully desaturated to begin with?
OKAY, OKAY. I know the 255/255/255 was a typo. You meant 255/0/0, didn’t you.
George
Luminosity = .3(34) + .59(234) + .11(255)
Where do the .3, .59 and .11 come from?
You didn’t start that 100x100px swatch by filling it with 255/255/255
DOH! Wow, good catch. I DID start with 255,0,0 as the original poster did. I mistyped it. Damn good catch!
This tool ignores the middle value.
What? It ignores green? The value you’re talking about (i think) is the green value in an rgb triplet! I guess that would explain the diff between the menu command and the sponge tool, but that seems kind of whacked to me, to just discard the whole green value.
Do you begin to get the picture?
yes, i think i do, what I’m missing is my question above about where you got those decimal values.
OKAY, OKAY. I know the 255/255/255 was a typo. You meant 255/0/0, didn’t you.
no. I was testing you! 😉 Of course the typo was evident from the decending r channel value following, but it was a stupid one.
thanks, dave
Dave
Sorry, I should have forseen the possible misinterpretation of "middle" value. Middle does not refer to G as in RGB. It refers to the value between the highest and lowest. For example the middle value in RGB 100/50/250 would be the RED value, in RGB 50/100/250 the GREEN value, and in RGB 250/50/100 the BLUE value.
Where do the luminosity weighting factors come from? That is explained in reply #8. I believe thousands of people were asked to compare red green and blue swatches as the light from them was individually varied in intensity in units of radiant power. They were asked to say "when" when the swatches appreared to be equally "bright", subjectively, of course. The averaged results produced the relative channel weights normalized so they sum to 1.
Some people might contend that a different set of weights should be used. Which set is correct is not really important. What IS important is the weighting set used by Photoshop and there is no question about that—PS uses 0.3/0.59/0.3.
George
oops! I’m also entitled to a typo I guess. Change tha very last phrase in my response #21 to "PS uses 0.3/0.59/0.11".
hmm… ok, so now I see how, but still not why. going back to your post #8, i’m also left with:
What bothers me is that the perceived brightness changes radically despite the constant luminosity, even when the before and after swatches are compared in grayscale. And that contradicts my notion of real world "luminosity" which is supposed to indicate perceived brightness. That’s where I am seeking solace.
I believe thousands of people were asked to compare red green and blue swatches as the light from them was individually varied in intensity in units of radiant power.
who asked? adobe? nobody checked w/me!!! 🙂
how did they decide to drop the middle value? seems that if they took all 3 and devided by 3 they’d be closer to the result you get from the sponge tool – though not exact, i guess due to the weights.
I’m also entitled to a typo I guess. Change tha very last phrase in my response #21 to "PS uses 0.3/0.59/0.11".
the scary think is I caught that and knew what you meant. 🙂 I guess that means I’m close to "getting" it. <G>
getting closer, thanks, dave
Dave,
Further defining the Sponge tool action at a pixel site (with a hard brush):
Let H = highest color channel value, L = lowest color channel value, and M = middle color channel value (between the highest and lowest)
Then H-L is the difference between highest and lowest values
In DESATURATE mode, the Sponge tool
(1) DECREASES (H-L) by 20% times the flow, so the decrease is 20% at 100% flow, 10% at 50% flow, etc.
(2) Uses this decrement for each mouse click, while holding luminosity and hue constant.
(3) Stops decrementing when H-L = 0 (digitally), i.e., when all channels are equal (nearly)
In SATURATE mode, the Sponge tool
(1) INCREASES (H-L) by 40% times the flow, so the increase is 40% at 100% flow, 20% at 50% flow, etc
(2) Uses this increment for each mouse click, while holding luminosity and hue constant
(3) Stops when one of the channel values reaches 0 or 255
In both SATURATE and DESATURATE modes, the middle value is never overtaken. It has to change in order to hold hue constant and in doing so always remains between the high and the low values even as H and L are approaching each other in desaturate mode.
I have verified this action on only one sample patch (200/150/100), from which I deduced the incremental changes. Try it on some other arbitrary patch—I’m getting weary of proof-testing.
George
Dave,
For more on the origin of the luminance weights, see Appendix B of Wayne Fulton’s "A Few Scanning Tips" www.scantips.com.
George
I hope the original poster pops back in here to see what’s become of his thread. I say good show george. Maybe I’ll drop dk a note tonight when I get home. Then if he’s still interested he’ll at least get the same benifit I got out of his thread! 🙂
Thanks for the edjamakation george!
dave
Dave
And thanks for lending an ear. It sure helps to have someone to talk to. My basic question is still up for grabs, but in the dialog with you I clarified a few other things for myself that I would not otherwise have thought through.
George
My basic question is still up for grabs,
That’d be this, right?
What bothers me is that the perceived brightness changes radically despite the constant luminosity… And that contradicts my notion of real world "luminosity" which is supposed to indicate perceived brightness. That’s where I am seeking solace.
Thanks, dave
Dave,
You’ve got it!
George
I am repeating here the conclusion I have reached on the basic question I raised and posted on an earlier, related thread "Luminosity in PS"
In mulling over this issue, I have changed my thinking to be consistent with the facts. Here’s how I squirm out of the dilemma:
One is being much too literal to expect perceived brightness to remain constant by simply constraining changes in the 3 channels to keep the invented parameter "luminosity" (.3R + .59G +.11B) constant. Such a constraint might be expected to mute the eye’s unbalanced sensitivity, but hardly to do so perfectly. It is likely impossible to find any set of constant weights that will do this job over the entire range of intensity variations. Using the chosen weights HELPS but cannot do the miraculous.
Therefore, in applying the constraint that luminosity be held constant, one should not expect perceived brightness to be unchanged. Just not changed so much over limited ranges. It’s better than not trying at all to weight the values to compensate for the eye’s favoritism. To be sure, one can keep the defined luminosity constant in an editing operation such as sponging—and this is indeed what PS does—but look only for compensation not total elimination of perceived brightness changes. My mistake is looking for perception to rigidly mimic the parameter’s constancy.
George
Can you please repeat that in English? 🙂
Dave
You are taking me back xx (secret) years to English class and I’m re-agonizing the "D" grade on my assignment to explain how a notebook-paper hole-puncher works and, worse than that, having to rewrite and resubmit. All I need is a "C" grade. Here goes:
The same radiant intensity in R, G, and B portions of the spectrum does not produce the same sensation of brightness. If we wish to maintain that same sensation as we manipulate the component values, we have to compensate for the eye’s uneven response. We do so by assigning weighting factors to each of the three channels. Perfect weighting factors to define perceived brightness cannot be found. We incorporate a best set into a "luminosity" parameter we invent and we hold that parameter constant in our manipulations. Since the set is imperfect, the perceived brightness does not stay the same, but it behaves better than with no weighting factors at all. When it’s crudely applied over broader ranges than ever intended, we don’t shout and wave it about.
George
Thanks. Much better! B+ 🙂
Thanks, Dave, for the grade, but shouldn’t Chris be the professor?
Yea, I was juat grading the english. I understood the rewrite a little more.
