RoMad

I'm reading along about important bike stuff and up pops a picture of a scantily clad beauty next to the post. I am glad to see the Ad department is doing well, but not sure if any of us will be buying our swimwear there. Not complaining, just wondering.

Yen

Yes. I just resized the browser a bit so she wasn't showing on the page.

Digital Gee

Bummer. You buy the red star, and you miss the good stuff.

maddmaxx 

We have advertisers?

Tom Bombadil

Hey, I thought buying the red star was supposed to be a good thing!

Is there anyway to turn advertising back on?

megaman

Where are they at?

solveg

Someone was telling me tonight that their place of employment actually has a sensor that reads how much skin tone is on a given web page. If it exceeds a certain amount, the Central Scrutinizers are alerted in the main office!

robtown

I think you can use the logout link up top.

maddmaxx 

The Central Scrutinizers have long had me on their lists. It has something to do with all those phone calls and money transfers out of the country. I now leave notes in my spreadsheets for them.

Tom Bombadil

Could it monitor all skin tone colors?

solveg

Surprisingly, there is very little difference in even the broadest range of skin tones when compared to the RGB gamut.

I have video color correction software that actually has a little tiny range marked for "skin color" on the color wheel. As long as you're in this range, it's a human color. It's probably a 3 degree range.

Tom Bombadil

This is a 3 degree difference?

solveg

Yes. In the whole RGB gamut. 3 degrees is actually generous. I can't find a screen shot of what my software looks like, but you'll get the idea from this image, and remember that it is actually a 3D ball.



The tools I use for video color correction are more like this, though:

solveg

And, actually, there are very subtle differences between the hue of the skin. The color ratios for print are very similar, but darker. I mean, when you're talking about the whole spectrum. In reality, we are very sensitive to skin color and can easily detect a few extra percentages off in the ratio.

A fair-skinned pinkish baby could be as light as 15% magenta, 16% yellow. Most Caucasians fall in the range of 5-20% more yellow than magenta. A fair-skinned Caucasian adult could be as low as 20% magenta, 25% yellow. A bronzed Caucasian could be as high as 45% magenta, 62% yellow.

We find that it's easy to oversaturate African-American skin, so be careful there. Yellow and magenta values should be fairly close. We agree with Lee Varis, in that you might find a yellow bias in African-American skin, which will not look good in print.

Asian and Hispanic skin will typically have 10-20% higher yellow than magenta.

Tom Bombadil

According to this:
https://www.swpp.co.uk/professional_i...skin_tones.htm

The range may be around 3% for light skinned Caucasians, but is much wider than that across all skin types.

Brian

Buying the red star is a good thing, it's just less critical to maintaining the server now that a company, rather than an individual, owns the site.

Other than logging out, there is no way to turn the ads back on. Some day, we hope to have a plugin that will allow donating members to turn the ads on and off at will.

As far as objectionable or inappropriate ads, if anyone has a complaint, don't hesitate to let me know. While I do see all the banner ads before they get loaded, the larger ad box and header ads are part of a different program. If there's an ad that doesn't belong on the site, I'm more than happy to take it up with the owners, you just need to let me know what the ad is for.

Thanks,

Brian
Forum admin and humble servant to the elderly

solveg

Nope. That refers to the amount of melanin in the human skin.

Seriously, think about it. The color wheel, which includes, green, purple, orange, etc. Ranging from pure white to pure black. The tones that human skin fall into are very, very small. I have a cheat sheet written somewhere for tones to "start with" when color correcting. When working with skin, even 1% cyan will ruin it to the human eye, so you're right...humans are highly sensitive to that tiny area of the color wheel. We also "know" grass green, sky blue and some other colors which you need to be very careful with.

But a computer looking for naked bodies would be able to plug in a pretty small range to monitor.

Tom Bombadil

That was one set of numbers, but it also contained a few graphs of RGB & other settings that showed a much wider range.

solveg

Not really. It was seeming wide within the small gamut of a printing press. I'll drag out one of my photoshop books called "skin" and see if they explain it better.

Basically, and I can only explain this well in CMYK terms, the ratio for caucasion skin is you take the yellow reading, and magenta should be about half of yellow, and blue should be about half of magenta. The exact ratios vary a lot, so that's really loose.

For the darkest shadows, you may add a combination with is a neutral darkener. These colors basically cancel each other out, visually. You don't really use black ink in faces, so you try to get a neutral effect using cyan, magenta and yellow, but you adjust it so it's not "grey" because that's offensive to humans.

Anyway, while you're adding the neutral darkener, you shift the colors slightly to accomodate different skin tones. But because you're adding color equally across CMY, you can't really take out the M and say, "Wow! We added 30% magenta to that face!" It's all proportional to the original numbers. So if you add 7% to the color 10/20/40/80, your new number will be 10.7, 21.4, 42.8 and 85.6.

Most people don't realize they have to do that math, and will just at 7 to all the numbers, which would be...wrong.

Anyway, so for a dark african skin tone, you're going to have a lot* more magenta and cyan in proportion to the yellow, but we're still talking about primarily a reddish yellow with some cyan in it.

In RGB the gamut of colors is bigger, so the skin tone range takes up even less of the total colors.


Look at this color wheel as if you were a computer. How hard would it be to pick out a page which may have a lot of nudity on it, no matter what race? (remember the color space is really 3 dimensional, with darkness being its own value. Take a caucasion skin, make it very dark without changing the hue of it. You will find it will take a much smaller move than you think to make it african skin. The example you chose is extreme. You could probably catch 95% of internet nudity if you took just the middle of that range.)

Tom Bombadil

Aha! Your reference cited in Footnote 5 that there was a way to distinguish human skin and that it fell into a narrow range. I found the article referenced in that footnote:
https://repository.upenn.edu/cgi/view...ext=cis_papers

They found that using RGB values was a poor way to identify human skin and instead used skin spectograph data which measures the reflectance of light off of skin. Turns out that human skin has some unique qualities there, which are not affected by the color of one's skin. Thus their process doesn't even bother with collecting or analyzing skin color, but just the reflective properties of surfaces within a narrow band of light, a "wavelet."

They found that their process had a very high success rate in distinguishing human skin from that of a mannequin having the same color surface as human skin, based upon the unique reflectivity properties of human skin. Very technical & interesting research. It doesn't say how quickly this data could be collected and analyzed, so I don't know if this method could be used for real-time monitoring of computer screens.

Tom Bombadil

Scary stuff indeed. Big Brother is alive and well. At UW-Madison we are very careful to not get into anything like this. Avoid it like the plague.

solveg

Ah, I had deleted that link because I thought it was saying that things like recognizing eyes and stuff was more important to camera recognition.

But I did find this example which shows how similar the actual proportions of color are among skin tones. When you get rid of the shadows and highlights, you're left with a pretty similar base. I mean, it's hugely different to our eyes, but not necessarily to a computer.

solveg

Interesting!!!!!

See, knowing this really helps me. It's placed additional emphasis on skin texture and light, which I knew was important already. But the more I know, the better I can do my job. All this science stuff is very important.

I can't tell you how many times I've tried to figure out things like how a leaf ages and dries or how light refracts on melting ice. It's the key to making retouching invisible.

maddmaxx 

Now..........on to the nonvisable frequency bands.........................

solveg