Skin Beauty: Software Batch Correction of Skin for the Most Pleasing Tone

In a previous thread here, Ben shared the novel idea of using a profile for the 5D from a Nikon D700 to optimize skin tones. The related posts including iPortrait Professional provide compelling alternate methods of getting pleasing skin tones in portraits or wedding pictures. So they were moved here to start a new thread devoted to software approaches to skin beauty.

With software one can remove contaminating hue referencing a neutral gray and only then apply the desired adjustment for the beautiful skin tone applied to the skin only. Other parts of the picture unaltered and the whites are perfect. Asher



I'm with Jack here as a wedding photographer, asthetic taste over accuracy every time! What is important for me is that all the other tones corresspond with the overall chosen WB. For example in ACR I could never get the faces to go right. Forget accurate which wasn't pleasing in most cases, the facial tones didn't fall into place nicely when the asthetically correct WB was chosen. I have mostly managed to fix that by mapping the Nikon D700's colour onto my 5D files using the DNG Profile Editor. It's very subtle but the saturations and depth of the facial colours is now (more) correct relative to the chosen white balance. I'm hoping that using the Gretag chart will give me this same help when shooting in complicated lighting. You can get the white balance correct but it won't help unless the software knows what to do with all the other tones relative to the WB. I find fluorescent lighting to be the worst by far to get pleasing skin tones for as everything looks dead under it! Mix it with flash and it's a nightmare to get right. You can have a perfect WB by using a Whibal but it still doesn't look 'right'.

Canon & Nikon skin colors, correcting color, WhiBal™ & colored reference patches!

Ben Rubinstein said:

I'm with Jack here as a wedding photographer, asthetic taste over accuracy every time! What is important for me is that all the other tones corresspond with the overall chosen WB. For example in ACR I could never get the faces to go right. Forget accurate which wasn't pleasing in most cases, the facial tones didn't fall into place nicely when the asthetically correct WB was chosen.

Click to expand...

Well, what Drew's device and other colored reference cards appear to do is to provide a nearly neutral reflecting material but doped with a dye that absorbs more of the blue-green colors so the resulting image will be boosted in the red-yellow, (if I am interpreting things correctly form the limited information presented). If it works for different skin types in a batch process, then it might be great time save! I just wonder what it does to the other delicate colors. Also how will one choice for one skin type effect various other racial types represented in the same photograph! What will it do to delicate nuances of color in floral arrangements or the bride and bridesmaids' dresses. These questions are just that, but not criticisms.

Ben Rubinstein said:

I have mostly managed to fix that by mapping the Nikon D700's colour onto my 5D files using the DNG Profile Editor. It's very subtle but the saturations and depth of the facial colours is now (more) correct relative to the chosen white balance. I'm hoping that using the Gretag chart will give me this same help when shooting in complicated lighting.

Click to expand...

This sounds interesting. Could you post an example of what your correction does in a new thread? I wonder whether you might be able to share a photoshop curve that I could overlay on my 5D portrait files? I know some people prefer the look of skin on the Nikon and Fuji cameras.

Ben Rubinstein said:

You can get the white balance correct but it won't help unless the software knows what to do with all the other tones relative to the WB. I find fluorescent lighting to be the worst by far to get pleasing skin tones for as everything looks dead under it! Mix it with flash and it's a nightmare to get right. You can have a perfect WB by using a Whibal but it still doesn't look 'right'.

Click to expand...

Well what this claims to do is give a pleasant look. The proof of the pudding is yet to come, but no doubt will appear before long!

Asher

Asher Kelman said:

Well, what Drew's device and other colored reference cards appear to do is to provide a nearly neutral reflecting material but doped with a dye that absorbs more of the blue-green colors so the resulting image will be boosted in the red-yellow, (if I am interpreting things correctly form the limited information presented). If it works for different skin types in a batch process, then it might be great time save! I just wonder what it does to the other delicate colors. Also how will one choice for one skin type effect various other racial types represented in the same photograph! What will it do to delicate nuances of color in floral arrangements or the bride and bridesmaids' dresses. These questions are just that, but not criticisms.



This sounds interesting. Could you post an example of what your correction does in a new thread? I wonder whether you might be able to share a photoshop curve that I could overlay on my 5D portrait files? I know some people prefer the look of skin on the Nikon and Fuji cameras.

Well what this claims to do is give a pleasant look. The proof of the pudding is yet to come, but no doubt will appear before long!

Asher

Click to expand...

Hi Asher,

I don't yet have the gretag chart, it's in the mail. What I do have already is a profile for use with the new Adobe beta profiles which is the D700's colours mapped onto a 5D's colour profile. Basically apply it to a 5D colour to get more nikon like colours. I'll send it to you if you like.

Ben Rubinstein said:

Hi Asher,

I don't yet have the gretag chart, it's in the mail. What I do have already is a profile for use with the new Adobe beta profiles which is the D700's colours mapped onto a 5D's colour profile. Basically apply it to a 5D colour to get more nikon like colours. I'll send it to you if you like.

Click to expand...

Thanks Ben,

That's something I've wanted for some time. One can similarly get all these desired skin appearances by batch addition of one or such "correction" curves. That's what Pictocolor iCorrect Portrait does! I just received an email from them:


So it's all ready for use and an batch process too with perhaps as many skin tones as you might imagine. This allows correction of individual faces in a special picture by selecting that person and choosing the tone that most applies to them.

Still, for an easy one click batch process, and picking just one tone, the colored card sounds something to look into.

Asher

Hi Ben and Asher,

These are all most intersting appraoches to the "skin tone" (and other similar) issues.

Do I assume that in applying these techniques, we would first apply white balance color correction (using whatever best technique we have available there)?

Best regards,

Doug

yup!

Here's the link. www.studio-beni.net/profiles.zip

There are two there,

A 5D profile mapped from D700 (ACR 4.4 Profile)
A 5D profile mapped from D700 (Adobe Standard Beta Profile)

They go into documents and settings/username/application data/adobe/cameraraw/cameraprofiles/

Then log off and back on again (no idea why it's needed for adobe to recognise them). When you open ACR or LR and go to the camera profile tab and choose the appropriate new profile.

Please note this will ONLY work if you have camera profiles installed. It will also only work with 5D RAW files!

I have no idea where to put the files on a Mac. Sorry guys!

I downloaded and mapped the D2X profile as well, it's said to have the best colour of any nikon, however what we are playing with is Adobe's interpretation of these cameras colours and they seem to have a default interpretation of Nikon, there was zero, zero difference between their D2X profile and the D700 profile.

Try it and let me know what you think. I personally find the depth and saturation of the facial tones with the D700 profile are better than the canon ones. Please compare like to like, 4.4 to 4.4 and beta to beta.

This so far is only mapping the Adobe interpretation of the D700 onto their interpretation of the 5D. When I get the gretag then I'll have a chance to make profiles specific to difficult lighting conditions. Adobe's standard profiles are based on a profile shot in tungsten and daylight and everything else is averaged between that.

Asher Kelman said:

Also how will one choice for one skin type effect various other racial types represented in the same photograph! What will it do to delicate nuances of color in floral arrangements or the bride and bridesmaids' dresses. These questions are just that, but not criticisms.

Click to expand...

Asher:

First and foremost, I use a *profile* for my camera when under known lighting, and a profile eliminates all the aforementioned extraneous concerns. However, for mixed location lighting like at a wedding, profiling is generally not a viable option, so here we have to default to a white-balance methodology applied to the best profile for that situation we have. I point this out because I believe most color balance issues stem first from cameras with poor profiles; this includes most DSLR's I've tested, and especially Canon. In effect what Ben has described above is he "profiles" his Canon to better meet his criteria for accurate color.

Assuming the camera now renders more properly to begin with, what I find -- and I am sure other folks opinions will vary -- is that in practical application, most humans are far more tolerant of slight color-shifts in whites, blacks and florals than they are of skin tone. IOW, most people accept a broad range of white as "white" even though it may range several hundred degrees warmer or cooler than true neutral white. Ditto for blacks, pastels and florals. However, when it comes to skin color, go a few points cool, warm, green or red and they notice it as "wrong" right away; IOW humans seem more acutely tuned in to what accurate skin color should be...

*I* find that usually only slight changes to color temperature or tint are required to make the skin look pleasant again, and these relatively minor color adjustments do not move the whites, blacks or florals far enough off to cause concern. Moreover, I also find that in almost every case, the change required to make Caucasian, Asian, Black, Hispanic or Mediterranean skin look "right" improves the other skin tones surrounding it at the same time --- IOW if the target skintone is too cool, usually so are the rest. To summarize, I thus focus my adjustments on the person's skintone who's paying me, and let all the others fall in where they may --- and most of the time, where they fall is truer and thus usually acceptable. While no method will be perfect from a purely scientific perspective, this is a case where in actual use, close enough usually is And here is where I think a tool like Drew's is of benefit to photographers...

PS: My result of course is one gleaned from use in a real-world shooting environment, and as such is not perfect science and therefore may not sit well with color theorists who would prefer to debate the specific effects of the resultant delta-E's on color plots.

PPS: Note that profiling tools vary and each manufacturer has to choose between a different set of evils to create the final profile -- and thus, not even a perfectly executed profile will be "perfect," and depending on the tool chosen some will be better in given situations than others.

Cheers,

How come this became it's own thread? Bit confusing, only found it by looking at my recent posts in my profile!

Hi, Jack,

A nice essay, and very useful. Thanks for the insights.

Jack_Flesher said:

is that in practical application, most humans are far more tolerant of slight color-shifts in whites, blacks and florals than they are of skin tone.

Click to expand...

Indeed.

In fact, what I like to speak of as the "dirty little secret" of the NTSC image "coding" system (the original standard for North American color television) is that, in partitioning the bandwidth allocated to the chrominance ("color") signal, the choice of orthogonal axes for the chrominance plane was made to give the highest-resolution discrimination between chromaticities along the locus of typical (1940-era Caucasian) skin tones, to small differences in which observers (presumably within a Caucasian-centered culture) were known to be most sensitive.

Best regards,

Doug

Doug_Kerr said:

Indeed.

In fact, what I like to speak of as the "dirty little secret" of the NTSC image "coding" system (the original standard for North American color television) is that, in partitioning the bandwidth allocated to the chrominance ("color") signal, the choice of orthogonal axes for the chrominance plane was made to give the highest-resolution discrimination between chromaticities along the locus of typical (1940-era Caucasian) skin tones, to small differences in which observers (presumably within a Caucasian-centered culture) were known to be most sensitive.

Click to expand...

Interesting tidbit Doug, I had no idea that was the case. But I guess it does make sense -- or at least maybe did for that era...

Best,

Jack_Flesher said:

Doug and Bart:

I think it would help us all if you each posted one of your personally-produced images that illustrates the superiority of your methods of balancing skintones.

Click to expand...

Okay Jack,

First I want to re-iterate that skincolor reproduction with a digital camera is a process that varies with the color temperature of the dominant illuminant, and with the amount of in camera IR filtration.

The following graph shows how the near-IR spectral reflection by (my) skin is 2-3 times higher than at the blue end:


That means that a lightsource with a higher IR bias (lower incandescent color temperature) will be even more critical relative to the camera's response.

Here is a (horribly lit) scene which could be considered to have a correct general White Balance:


While the grays are neutral, and the overall color reproduction of the sensor array and Raw converter are calibrated, the skincolor is the only thing that's quite a bit different from the accurate color.

We could try and change the skin color by using a different White Balance when converting the Raw data, as is the implict recommendation by Drew's method. This is what the 'accurate' color of the skin would require:


The drawback of such an approach obviously is that ALL colors, except for the skin, will be rendered with a cast, and the skin color is still only accurate to a degree because WB alone cannot nail it exactly.

The preferred approach, which I'm advocating, is to use the correct White Balance for the scene, and only adjust the deviant skin colors with a Color Balancing correction. The reason for that being that skin color will deviate most from accurate in many Digtal camera / Raw converter combinations. The result is shown here:


Selecting skin colors (for a mask) is relatively straight forward with a color range selection, and the colors can be changed at will with a Color Balancing filter (layer). That would also allow to Color Correct the camera's response to a mix of different skin colors in the same image. It can even be used in an action if speed of operation is more important than absolute perfection.

With this method we can also apply a partial correction, or if we want to add a tan or sunburn to an existing image, which is often considered to be more pleasing (to a customer) than accurate skin colors.

Bart

Bart_van_der_Wolf said:

Selecting skin colors (for a mask) is relatively straight forward with a selective color selection, and the colors can be changed at will with a Color Balancing filter. That would also allow to Color Correct the camera's response to a mix of different skin colors in the same image. It can even be used in an action if speed of operation is more important than absolute perfection.

With this method we can also apply a partial correction, or if we want to add a tan or sunburn to an existing image, which is often considered to be more pleasing (to a customer) than accurate skin colors.

Bart

Click to expand...

Bart,

This is obviously a great approach and one that Picto color uses. You are suggesting using the color selection tool. How would you set up a batch process. Would you do one skin type at a time and can one store such selections as one would have to specify tolerance sufficient to distinguish a caucasian from an Asian or Black skin, each with sub types.

Asher

Asher Kelman said:

This is obviously a great approach and one that Picto color uses. You are suggesting using the color selection tool. How would you set up a batch process. Would you do one skin type at a time and can one store such selections as one would have to specify tolerance sufficient to distinguish a caucasian from an Asian or Black skin, each with sub types.

Click to expand...

A selected color range can be saved, and one can add multiple skin colors in a single selection (the dropper tool with a plus, in the Select Color Range dialog). One could save such a range with a generic name, e.g. "SessionSkinColors". That would allow to use a generic script which loads that specific color range for the session at hand. The selection will act as a mask when the action creates a Color Balance layer with some highlight color adjustments for the specific camera.

Bart

Bart_van_der_Wolf said:

A selected color range can be saved, and one can add multiple skin colors in a single selection (the dropper tool with a plus, in the Select Color Range dialog). One could save such a range with a generic name, e.g. "SessionSkinColors". That would allow to use a generic script which loads that specific color range for the session at hand. The selection will act as a mask when the action creates a Color Balance layer with some highlight color adjustments for the specific camera.

Click to expand...

What are the syntax limitations for how the name is written?

Would one do this for a series of White Balanced skins?

I'd like to see an action in practice and see if it picked up a part of a chair or else an incomplete face?

Asher

Asher Kelman said:

What are the syntax limitations for how the name is written?

Click to expand...

I can think of no other restrictions than those imposed by one's Operating System.

Would one do this for a series of White Balanced skins?

Click to expand...

I'm assuming that the general White Balance is correct. Then one could sample a few different skins, in the typical session setting. Alternatively, one could use a predefined palette* of typical skincolors and adjust the tolerance to cover a large base.

* As a hint, the skincolor at the back of my (currently sun starved) hand averages to Lab = 59.77, 5.67, 12.13 as measured by a spectrometer at 5000K, 2 degrees observer angle. This would resemble the ProPhoto RGB colorspace (also 5000K), convert to RGB coordinates for your particular working colorspace. One could assemble a larger sample database and calculate (or sample) shaded and highlighted tone variations. Such a palette could serve as a basis for color range selections under a given dominant illuminant driven White Balance color temperature. Correct White Balancing remains important, due to IR contamination for a specific camera model.

I'd like to see an action in practice and see if it picked up a part of a chair or else an incomplete face?

Click to expand...

The wider one sets the tolerance, and the more shadows are included in the samples, the larger the chance of picking up non-skin colors. As with all fully automatic approaches, there will be unintended side effects. With a little more work, one could employ a selection of a skincolor range, with an intersection of an actual coarse selection of faces/hands/arms.

Bart

well, except no...

You don't need to mask for a correction like this, though. All you need is the right overall balance.

Now, there are a lot of (what I assume are) neutrals in that shot, and you could correct to any of them. Then you correct for skin, and measure it any way you like.

This is a 1 minute curves job in PS. Of course, when processing hundreds and hundreds of shots, I don't want to spend that minute on each shot and would like to get it done in the RAW converter, which is where correction in WB, contrast, and profiling comes in...

Still, here's what I mean...

Here's one of Bart's original shots, where he says that the shot has been corrected for skin with white balance.

First of all, that can't be right, because even without measuring I can see the cyan cast on the entire shot, including the skin. If you had a hand that looked like this in that kind of light, then you'd be seeing the doctor, and quickly. Or it may already be too late
View attachment 83

Now watch what happens to the same shot with a white point and black point corrected (I chose the chart patches to force to neutral, but you could have chosen whatever you like). Then, on the curve, I selected a patch of skin and corrected for that "to measure"...

Voila! Good skin, and perfect neutrals (according to the RGB measurement) in the chart (but not on whatever that cap thing is, unless it really is blue).
View attachment 84

See, once again "theory" gives way to practice You just shouldn't get caught up in claims of "accuracy" when the result is clearly and demonstrably wrong to human eyes.

See--this is not as "subjective" as Doug or others might think. It is, however, a qualitative thing, and you can't always measure qualitative differences that simply.

What I've just done in PS in less than a minute with no mask is essentially what Drew's device does too, if you know what you're doing with it...

BTW--given the bad source, here's the curve in PS necessary to correct both neutrals and skin tone:

View attachment 85

James Roberts said:

even without measuring I can see the cyan cast on the entire shot, including the skin. If you had a hand that looked like this in that kind of light, then you'd be seeing the doctor, and quickly. Or it may already be too late
View attachment 83

Now watch what happens to the same shot with a white point and black point corrected (I chose the chart patches to force to neutral, but you could have chosen whatever you like). Then, on the curve, I selected a patch of skin and corrected for that "to measure"...

Voila! Good skin, and perfect neutrals (according to the RGB measurement) in the chart (but not on whatever that cap thing is, unless it really is blue).
View attachment 84


BTW--given the bad source, here's the curve in PS necessary to correct both neutrals and skin tone:

View attachment 85

Click to expand...

Jamie,

Yes, it's better, for sure, but our patient still goes to the hospital, cyanotic!

What's wrong with just making the neutrals neutral? The skin then is a nice pink. (Bart, don't take that as any compliment on your looks!) What values did you set the skin to?

Asher

James Roberts said:

You don't need to mask for a correction like this, though. All you need is the right overall balance.

Now, there are a lot of (what I assume are) neutrals in that shot, and you could correct to any of them. Then you correct for skin, and measure it any way you like.

This is a 1 minute curves job in PS. Of course, when processing hundreds and hundreds of shots, I don't want to spend that minute on each shot and would like to get it done in the RAW converter, which is where correction in WB, contrast, and profiling comes in...

Click to expand...

That's exactly what I did in shot 1 (accurate WB), and shot 3 (accurate WB and Skincolor). Shot 2 shows what happens when WB alone is (ab)used for skincolor.
Doing a Color Balance (or curves/hue saturation/filter layer/other adjustment) will allow to start from an accurate scene, and allows to transform towards accurate or even pleasing skin color.
If that adjustment throws the scene color balance too far off, one can mask for skin only and leave the rest of the scene alone in correct WB. Whatever the circumstances dictate.

Here's one of Bart's original shots, where he says that the shot has been corrected for skin with white balance.

First of all, that can't be right, because even without measuring I can see the cyan cast on the entire shot, including the skin. If you had a hand that looked like this in that kind of light, then you'd be seeing the doctor, and quickly. Or it may already be too late

Click to expand...

The scene is turned Cyan due to the required WB to get the skin color accurate. I agree, it doesn't look 'pleasing' (especially in Cyan surroundings) but it is accurate.

Voila! Good skin, and perfect neutrals (according to the RGB measurement) in the chart (but not on whatever that cap thing is, unless it really is blue).

Click to expand...

I wish my sun starved cold skin looked like that in the winter at 10 degrees Celcius (50 Fahrenheit), in fact it doesn't even look like that in the summer.

See, once again "theory" gives way to practice You just shouldn't get caught up in claims of "accuracy" when the result is clearly and demonstrably wrong to human eyes.

Click to expand...

You seem to have misinterpreted the examples, I was not demonstrating 'pleasing' skincolor, I was demonstrating the flaws of achieving a certain skincolor by abusing the WB to get there. In fact you demonstrated that, even for more pleasing skincolor, a Color Balance can be used to achieve it without getting the whole scene out of balance. And you could have used the first example to get there, it already starts with a correct overall WB.

Bart

Asher Kelman said:

What values did you set the skin to?

Click to expand...

In shot #2, I aimed for the same color as read by a spectrophotometer, approx. Lab 60, 6, 12 in ProPhoto RGB colorspace, then converted to sRGB for the Web version. The scene was unfavorably lit by rather frontal direct light.

I compared my skin color with outdoor light falling on it, next to my (shielded) calibrated display, and the color looked surprisingly similar.

Again, in #2 I was aiming for accurate, not pleasing, because pleasing is a moving target whereas accurate is accurate (within accuracy of the spectrophotometer).

It would be most interesting to learn e.g. Andrew's reading of his skin (back of hand, trying to miss the veins/arteries and hair, averaging several readings). Others, like Michael, with a spectrophotometer please join in in buiding a better sense of what accurate can mean. Asian skin and African skin colors are also most welcome to our small experiment.

Before we wander into more pleasing color directions we need a startingpoint for our vectors.

BTW as far as CO 4.51 is concerned, my skin color in the sample shots is characterized as light pink (not cyanotic), and I can get a reasonable WB by using that. Which also demonstrates how poor our judgement of color is, and how tricky a concept of pleasing is.

Bart

Bart_van_der_Wolf said:

In shot #2, I aimed for the same color as read by a spectrophotometer, approx. Lab 60, 6, 12 in ProPhoto RGB colorspace, then converted to sRGB for the Web version. The scene was unfavorably lit by rather frontal direct light.

I compared my skin color with outdoor light falling on it, next to my (shielded) calibrated display, and the color looked surprisingly similar.

{Snipped}

Click to expand...

You guys just don't get it, I fear...

Bart--all this says is that the spectro is as human-unfriendly as the camera. I could have started with shot 3, I guess, but to say that shot 2 had "pleasant" skin colour as you do is ridiculous. So you might have abused white balance, but it's abused all the way

Asher--none of the shots Bart posted are close to being pleasing, if by pleasing we mean "following assumptions for printing that most people would assume look good."

Setting the neutrals and letting the skin tone be pink I suppose would be ok, except in our culture (NA) most people don't see their skin as pink (too magenta) but rather as a shade of brown with more or less pink overtones.

FWIW, in the corrected shot I posted, the neutrals on the chart were neutral. The other "neutrals" were not And while I don't claim accuracy for the shot I posted it's much closer than any of the others.

If I can get "there" in RAW as opposed to in PS, I will do it.

But my only point here was to show 1) you don't need a mask 2) neutrals are subject to interpretation 3) skin tone is pretty consistent

IOW, in a way Bart's right: you need to know where you're correcting *to* before you start correcting. In my view, the spectro mis-used is just as unreliable as a completely subjective interpretation.

Just for kicks, then...

Here's Bart's "white balance is ok, skin tone is ok shot (shot 3)"... skin is still way too cyan regardless of what a spectro read:

View attachment 86

Ok, now here is the skin tone corrected for a better caucasian RGB (or, in my workflow, still CMY) ratio. Since you said your skin is pink, Bart (or was that Asher?) I left a lot of magenta in there :

View attachment 87

Now, notice that the white and black patches on the chart are now neutral, but the WhiBal is most emphatically not neutral Neither is the background. This is a too red for my liking (and the browser is making it worse--this is better in PS), but heck--you guys want pink skin

If you think the contrast has increased too much due to my setting a white and black point (the points on the chart) then you can desaturate the result without losing the skin tone ratio:

View attachment 88

The WhiBal--and whatever that thing is in the foreground--is still not neutral.

James Roberts said:

You guys just don't get it, I fear...

Click to expand...

Agreed Jamie, and thanks for posting this. I had decided to leave it alone and forgo the arguments... I think the most important point is that both skintone and neutrals are subject to interpretation, and what is most pleasing may not not necessarily be the most technically accurate. And given the choice, I'll lean towards having a pleased customer even if it means the white tablecloths were a tad warm...

Jack_Flesher said:

Agreed Jamie, and thanks for posting this. I had decided to leave it alone and forgo the arguments... I think the most important point is that both skintone and neutrals are subject to interpretation, and what is most pleasing may not not necessarily be the most technically accurate. And given the choice, I'll lean towards having a pleased customer even if it means the white tablecloths were a tad warm...

Click to expand...

Hi Jack,

Actually we all share the same views just that there is a triangle of three sets of color

• The Camera's
• Reality
• What's pleasing

I see this discussion as useful. I'm looking forward to dealing with skin color more effectively. We'll all learn from this. Jamie is so talented as is Bart. Together we'll get better at seeing how color might be approached for different purposes and more efficiently. There's actually more work going on in the background than disclosed here and it's all very generous and friendly.

Asher

James Roberts said:

You guys just don't get it, I fear...

Click to expand...

Hi Jamie,

I'm afraid you are right.

Bart--all this says is that the spectro is as human-unfriendly as the camera.

Click to expand...

Exactly! I never said it produced pleasing color, it just established an accurate baseline for further tweaking. I also shows that a general WB correction for pleasing skincolor will throw off the WB of the majority of the scene (unless a mask is used).

I could have started with shot 3, I guess, but to say that shot 2 had "pleasant" skin colour as you do is ridiculous.

Click to expand...

Accurate, not pleasing!

Setting the neutrals and letting the skin tone be pink I suppose would be ok, except in our culture (NA) most people don't see their skin as pink (too magenta) but rather as a shade of brown with more or less pink overtones.

Click to expand...

Yes, and that is (as we established in the other thread) a cultural thing. We like to fool ourselves, which is fine.

All we need to establish as photographers, is how to get from accurate to 'pleasing' without sacrificing the overall scene. For that we need insight into a valid baseline, and a number of vectors to modify an accurate recording from our camera+Raw converter's profile into whichever tones are deemed 'pleasing' in a certain cultural setting.

If I can get "there" in RAW as opposed to in PS, I will do it.

Click to expand...

Obviously one would try to avoid anything besides Raw if significant color or tonality shifts are to be expected. I think there is also a huge underestimation of what an accurate camera profile can do. On top of that, Raw converters should have facilities to tweak skincolor with ease. Not just the White Balance, but skincolor. That will remain an issue until we finally stop guessing, and start injecting some simple science into the equation (Baseline plus vector, effect restricted to skintones).

IOW, in a way Bart's right: you need to know where you're correcting *to* before you start correcting. In my view, the spectro mis-used is just as unreliable as a completely subjective interpretation.

Click to expand...

Any instrument misused is unreliable. The spectrophotometer just established the *from* position, and the vector (defined by position, direction and magnitude) is what we are trying to establish.

Bart

Asher Kelman said:

Hi Jack,

Actually we all share the same views just that there is a triangle of three sets of color

• The Camera's
• Reality
• What's pleasing

Click to expand...

And how do you define "reality?" IOW is it what you see, what I see or what the client sees, or is it what we measure with a colorimeter? The issue here is we all see "color" differently and colorimeters are not perfect. While I know that red is red, I don't know exactly what red it is, though I can easily distinguish a one point change in 8-bit mode if I set two patches next to each other on my monitor; I can distinguish the hues, but do not have perfect enough visual pitch to identify it accurately. Nor do I know if you see red in the exact same hue I do -- in fact, I'd assume you don't. Moreover, if I look at red and cover one eye, it looks a tad bluer than if view it only via my other eye...

Thus I would lean toward the fact that we as photographic technicians are put in the position of taking camera color and turning it into pleasing color --- and pleasing color may in fact only be achieved if it is accurate color, as in say a corporate graphic.

Cheers,

Jack_Flesher said:

And how do you define "reality?" IOW is it what you see, what I see or what the client sees, or is it what we measure with a colorimeter? The issue here is we all see "color" differently and colorimeters are not perfect. While I know that red is red, I don't know exactly what red it is, though I can easily distinguish a one point change in 8-bit mode if I set two patches next to each other on my monitor; I can distinguish the hues, but do not have perfect enough visual pitch to identify it accurately. Nor do I know if you see red in the exact same hue I do -- in fact, I'd assume you don't. Moreover, if I look at red and cover one eye, it looks a tad bluer than if view it only via my other eye...

Thus I would lean toward the fact that we as photographic technicians are put in the position of taking camera color and turning it into pleasing color --- and pleasing color may in fact only be achieved if it is accurate color, as in say a corporate graphic.

Cheers,

Click to expand...

Hi Jack,

You've covered the fuzzy meaning of "reality" very well. Color that you or I perceive my well be slightly different.

Skin color, as with all color, is an experience of the human mind in how we perceive illuminated objects. However we can readily confirm that we refer to the same surface area and take a color reading with a meter. Our perception of that object's color can then be referenced to color charts to determine observer variation. So these together, spectrum of light from an object and our joint perceptions might approach something agreeable.

The point of all this is that except in museum, gallery, forensic or scientific work, we rarely want to achieve color reproduction fidelity. Rather our color choice represent culture and changing taste and social values. This is a challenge since we have a library of values of what's desirable that can be so varied. People do not, in fact, want to see in a picture what they can see in the mirror for free! They are so influenced by what appears in magazines and movies, often retouched and fanciful. That's the challenge. Except for unblemished children, what "is" is not always what's really wanted.

That's where the fun is!

Asher

Asher Kelman said:

The point of all this is that except in museum, gallery, forensic or scientific work, we rarely want to achieve color reproduction fidelity.

Click to expand...

To push a bit further, my point is more that reality is irrelevant. What is relevant is what pleases the viewer, be they a paying client or casual observer. In the case of the paying client, if they want reality -- whatever their version of reality is -- and we deliver it, they will by definition also be "pleased." To my mind, reality as it respects color serves as a good topic for endless debate among academics that have nothing better to do other than debate.

They are so influenced by what appears in magazines and movies, often retouched and fanciful.

Click to expand...

Precisely! And if we deliver a product that makes them look like that, they will likely be pleased LOLOLOLOL!!!

Jack_Flesher said:

To push a bit further, my point is more that reality is irrelevant. What is relevant is what pleases the viewer, be they a paying client or casual observer. In the case of the paying client, if they want reality -- whatever their version of reality is -- and we deliver it, they will by definition also be "pleased." To my mind, reality as it respects color serves as a good topic for endless debate among academics that have nothing better to do other than debate.

Precisely! And if we deliver a product that makes them look like that, they will likely be pleased LOLOLOLOL!!!

Click to expand...

Still, Jack, what we want to do, and why we tolerate such long, long seemingly endless discussion of the reference patches offered, is that we do want:

to get a sense of reality and pleasing skin color. Without the former, the second does not work. That's what it's about. ​

For that reason, I'll be doing work on this and look forward to getting the new device currently on its way from Drew to both Doug and myself. I prefer software solutions, but anything that work is welcome to me!

In spite of what may well seem contentious, pointless and challenging, Drew and Doug have some magnet attraction and respect for to each other. Drew has valued Doug's opinion and quotes his work! Yes, that's a paradox. What I want most is batch software processes that are user friendly. I hope, at the end of this, we'll be better able to get both great skin in a realistic setting.

Asher