Just wondering what peoples thoughts are on what makes for the best kind of diffuse skin maps in terms of colour, brightness & saturation.

Lets say as a point of reference we're talking in terms of an average untanned caucasian skin, for use with a PP2012 SSS material setup.

Looking through the limited number of maps I have in my runtimes there's quite a variance in all three aspects. Most are made using photo resources and have at least some shadow or specular that I usually photoshop out, but besides this there's an overall colour, tone & saturation to a map. Some have mostly a yellow bias, some an orange/red bias. Some are very bright and saturated whilst others are really quite grey.

One theory I read suggested that for most diffuse textures (probably with the exclusion of skin due to it's complexity in how it reacts to light) what you're looking to have is the material as if it had been photographed in bright sunlight. This may well be bogus, I don't know.

In any case I sometimes find it quite tricky to guage how far to tweak a map in a shader setup using the HSV node for example - so easy to blow out one or other aspect, like lips or nipples, whilst trying to achieve that perfect skin tone.

Blue eyes, deep blue lipstick, and pink nipples on fair skin.

That help?

Frankly because we are talking about something with a bit of  variance, as long as you are within normal bounds, it doesn't matter all that much.

Four renders of V4 under the same lighting scheme (EnvSphere with HDR map plus 1 inf light) - bear in mind there's some colour coming from the environment lighting.

Each uses the same material setup with the exception of slight variations to the HSV node settings between the scatter node & diffuse maps. From left to right theres a slight decrease in hue (reddening the output), an increase in the saturation, and decrease in value. It seems subtle but overlay one atop another and flick between them and there's quite a difference.

Which looks more natural?

I get what you're saying about being withing normal bounds, but to me the first (unadjusted map) looks too yellow and under-saturated.

...I've been looking at this too long haven't I?

To my preference, the right figure looks the most natural.  My eyes are more attracted to the texture details than the others.  I have a definite problem with the left figure, something unnatural in her appearance.  Too pale, perhaps? 

I think the second one gives more balance, not to saturated yet not too reddish. 

I come across this problem with some of my renderings of textures.  My characters can become way to red or way to yellow.  I like balance in my textues, but I can't see to get a decent effect as what your going through.

I wish I know more about HSV, is it helpful?

The HSV node is really just a means of adjusting the output of the map in this shader setup, not dissimilar to adjusting an image in photoshop using Hue/Saturation. It still depends on the balance and quality of the map though.

They all look about right to me, given the variations you'll see in caucasian skin tones.

If anything, the palest one is most typical of many Northern European tones, since we're all starved of sunlight up here. ;) 

And that might be an underlying problem, as Sam points out.  What is a correct skin tone?  But Dan, you are bringing up an interesting concept to alter and control the output through HSV.  There may be no correct or incorrect result except personal preference.  However, if the map looks like crap, there may not be any other alternative than to rework the texture map or find an extensive material room solution to correct it.

Your eye can only help so much with this. The human visual system is prone to trick you based on expectations, a trait that is often used in making optical illusions.

You should use histograms.

A histogram of a swatch of skin texture will show you the exact proportions of red, green and blue in that swatch. It will also show you the dynamic range of each colour channel.

You can then compare your Poser skin texture swatch with one taken from a photograph of real skin that you want to emulate. Obviously you will need to take care to match the lighting conditions as accurately as possible. The histograms will tell you the truth, your eyes will not.

I think saturation is enemy of skin shaders. Saturation of information.. Photos pick up skin blushing(sss) as well as its translucence. Its alternate highlighting along with its specular .  Reflections are hard to see on a white skin but they are there.

All these should react to your lights or scene and not be part of your texture. Stripping all that out leaves a ton of nodework to get it back in properly. Not for everyone but you could handle it. I see collarbones with shadows on this texture. The white "highlights" painted on her belly...nice 2005 tricks but bad in 2012

I'd like to see a way to ramp in blush. Not there yet.

A good spec map with a clean alfa texture using Snarly's EZ skin..and patience

 Hey Carodan

I like the 3rd or 4th best- the far left seems a little too yellowish for a Caucasian to my eye.

I love using the HSV to adjust-and I also find I need to change the saturation- all this using Gamma Correction- and BB's Lightmeter prop-just so I have a light intensity REFERENCE -since that can change everything.

Eric

Quote - Your eye can only help so much with this. The human visual system is prone to trick you based on expectations, a trait that is often used in making optical illusions. You should use histograms.

A histogram of a swatch of skin texture will show you the exact proportions of red, green and blue in that swatch. It will also show you the dynamic range of each colour channel.

You can then compare your Poser skin texture swatch with one taken from a photograph of real skin that you want to emulate. Obviously you will need to take care to match the lighting conditions as accurately as possible. The histograms will tell you the truth, your eyes will not.

Wow, that was the best advice I've heard today, SG. Did exactly as you suggested and POP. Yes to histograms. Mind you, I'm pretty sure one KobaltKween told me something very similar (cc?) but do I listen? :glare:

Quote - The histograms will tell you the truth, your eyes will not.

+100

An advice to everybody: do yourself a favour and learn at least the basics of image processing.

I'd go with 1 and 2. As Sam said, I think 1 is quite valid for some real people.  For whatever reason, I tend to see many textures as having too much of an orange cast. On Caucasian skin I suppose it can pass as spray tan, but I rarely see an African skin texture that doesn't exhibit the effect to the point where it just doesn't look quite right It may be that my monitor, my eyes or my brain are miscalibrated - probably all three.

I realised my render comparisons don't really illustrate much beyond the use of the HSV node.

Richardson & Snarlygribbly have both hit (in their own ways) on what I think I was trying to get at here.

We tend to think that a photo reference of skin used as a diffuse map will give a (more) accurate render, but as Richardson points out any photo of skin already contains colour & tonal information that we will eventually be trying to simulate using 3d shaders & lights, based on the conditions under which that photo might have been taken.

Using a histogram to examine the colour information of a skin swatch is useful, but surely it will give us results that are only useful to compare with a final 3d render of skin under a similar set of lighting conditions, not the composition of the diffuse component used to achieve that render (that's how I see it at this point in any case).

Now, I'm thinking that bb's shader setups (that most of us are using variants of) partially account for this by not using the Scatter nodes built-in material option and give us very passable skin, but there's still a lot that's being calculated on the basis of colour information that probably shouldn't be in a diffuse map.

It's a simulation of course, but I wonder how much is being skewd by the maps we're using at present, and the shader balancing act employed to deal with that kind of diffuse component.

You're absolutely right Carodan.

A photo can only be taken with light (it's in the name!), and the properties of that light will affect the photo. Change the light and the photo changes.

But lighting can be controlled in a studio to produce very diffuse lighting. On a grey, cloudy day you can get some wonderfully diffuse lighting naturally, with barely a shadow or specular highlight anywhere to be seen. With such lighting the directional component is minimal and with judicious adjustment of the white balance any colour cast introduced by the light source can be obviated.

The histogram will then provide some very useful information which can be used as a yardstick when comparing Poser skin textures.

As an aside, and somewhat off topic, there are a couple of optical illusions here (scroll down to 'optical illusions'), which show how easily your own visual system fools you, while a histogram would tell you the truth :-)

Back on topic, I hope you persevere with this line of enquiry. Understanding how textures need to be made to maximise their effectiveness in Poser can only bring goodness to our hobby! Your attention to detail is renowned, so your conclusions will be invaluable, I'm sure.

Sorry to bang on about histograms (I'm not obessed with them, really!), but they have another major advantage that's worth mentioning: they are not affected by monitor calibration (or lack thereof).

Few people have calibrated their monitors accurately. Most who think they have probably haven't. So everbody looking at your four reference photos will be seeing them with a bias introduced by their monitor display and its properties.

A histogram is immune to those considerations - it reports only the data.

Snarlygribbly - well, I largely operate using the core work done by others, (bb's advancements in Poser shaders & lighting in this arena). I'm about as far from scientific in my approach as you can get, relying on the 'eyeball' technique. As those optical illusions illustrate, this doesn't always hack it, so I try for at least a half-way house of trying to understand the theory pared with comparison to photographic references. Still leaves a lot of room for error.

Incidentally, I noticed you credited me in your EZSkin documentation. While it gives me a warm, fuzzy feeling I should point out that everything I've done with PP2012SSS is based almost entirely on shaders created by bb with only minor tweaks by myself. Just putting that out there. It's a very nice script btw.

Yes though, I want to do a lot more experimentation regarding diffuse and control maps. Time is always my greatest enemy though.

Monitor callibration - omg it's a mine field. Just the difference with an LCD monitor working at night and then during the day makes my head hurt.

Yes, I agree, monitor calibration can be quite a handful with staggering differences from user to user, both amateur and professional, and not to mention the environment surrounding the setup. 

I think the histogram approach may help to set some standards.  However, will we be looking at the overall histogram to base our judgements?  Or will we need to go through the various red, blue, green components to fine tune each stage of the texture?  This may bear some relevance to how the map was generated originally under specific lighting conditions.