Slashdot Mirror
Advances In Cinema Tech Overcoming a Strange Racial Divide
barlevg writes "Since the birth of film, shooting subjects of darker complexion has been a technical challenge: light meters, film emulsions, tone and color models, and the dynamic range of the film itself were all calibrated for light skin, resulting in dark skin appearing ashy and washed-out. Historically, filmmakers have used workarounds involving "a variety of gels, scrims and filters." But now we live in the age of digital filmmaking, and as film critic Ann Hornaday describes in the Washington Post, and as is showcased in recent films such as "12 Years a Slave," "Mother of George" and "Black Nativity," a collection of innovators have set to work developing techniques in lighting, shooting and post-processing designed to counteract century-old technological biases as old as the medium itself."
Is this for real? Hollywood produced plenty of Italian American superstars, as well as Latinos. How did Bollywood manage?
Fuck systemd. Fuck Redhat. Fuck Soylent, too. Wait, scratch the last one.
Anyone else remember when HP's webcam face login program refused to recognize black people and it had to be recalled and altered? This totally reminds me of that. Classic HP.
Film is not "biased" towards people with "light skin." Quite frankly, I don't see how any visual medium that's designed to capture an accurate colour spectrum could be racially biased.
I think this whole article is a trollish attempt to inject a "racial issue" where there is none.
I do not fail; I succeed at finding out what does not work.
. . .Lilies of the Field was done in B & W.
That, and there was less competition for cinematography Oscar bestieness.
You shouldn't be a photographer or work in the film industry.
Every (competent) photographer knows that camera film and sensors have a very limited way of "seeing" compared to the human eye.
Your eyes have an incredible dynamic range (the range of light you can see at any one time) that cameras cannot hope to match, at least not currently. That's why you see no stars in the moon landing photos; and why you *can* see stars and the moon simultaneously when you look up at the night sky.
The funny thing is that film (negative, not slide) has *more* dynamic range and exposure latitude than digital. Getting differing subjects exposed correctly is mostly in the lighting, which has always been possible.
This reminds me of when they were developing the original pilot for the original "Star Trek" series. They wanted to know how the green-skinned Orion slave-girl would look when filmed. They covered her in green makeup and shot some test footage. It came back from the lab with normal pink European flesh tones. So they tried darker makeup. Still pink. They tried the darkest, densest makeup they could find. Still pink. It turned out that the lab was oh-so-helpfully "correcting" the color for them. I think this speaks volumes as to the article's premise...
There was a hack in some early NTSC TV sets which actually did have a bias for white people. NTSC has a luminance channel and two color channels, which are converted to three color channels to drive the CRT. Because the color channel bandwidth was limited and the signal level wasn't that consistent, some early color receivers had a special case for "skin color". When the two color channels, treated as a vector, were in the "skin color cluster region", they were pulled to that value, which was set for "white" people. Even if the other colors were way off, the skin colors would be consistent.
But that hack went out with vacuum tubes.
In other words, they've gotten better at color correction. I worked on color correction for Walt Disney's Heroes Work Here campaign and I spent a long time agonizing over the woman in the stadium. It wasn't because of any kind of racial bias, it wasn't because of any kind of subconscious decisions. It's entirely because of shooting technique and conditions. The problem was making her skin exhibit contrast against the dark background without making her dress completely blown out.
It was a combination of the fact that she wasn't shot with enough lighting to make her stand out against the background, and that digital imaging sensors don't have as wide a range of exposure (dynamic range) as the human eye.
The problem is even further than that. When you get into psychovisual enhancements to allow lossy compression to better do its job that means discarding details, and details we least often notice happen to be in the darker portions of luminance. What's needed there is some sort of more intelligent encoding system that can differentiate foreground objects from background objects.
Rawr
The film bias is real. The January 2006 issue of Popular Photography featured an article about different film emulsions sold outside the U.S. that better capture skin tones that are darker/different than caucasian. They shot a black model using Kodak Portra 160NC and Kodak Ultima 100, a film "tailor-made for shooting Indian weddings." They used the same lighting, adjusting exposure only for the 2/3 stop difference in film speed. I quote:
The negatives were dramatically different. Ultima 100 produced visibly more detail in Dionne Audain's skin than did Portra 160NC, especially on the shadowed side of her face. In matched prints, not only was that shadow more open, but there was a much better sense of texture in her hair and black sweater. The surprising thing is that, despite Ultima 100's higher minimum density, it seemed to have more snap overall than Portra 160NC.
First off, film was *not* designed to capture an accurate spectrum. If you took a picture of bouquet of flowers, and compared the spectrum of that image to the original's, the spectra would be quite different even if the color reproduction was perfect.
That's because color isn't a physical property like wavelength. It is a physiological response to wavelength. This sounds like splitting hairs, but it's not. Two different mixes of wavelengths can produce the same perceived color if they stimulate the cones in human eyes the same way. Birds and reptiles have *four* primary colors instead of three (we know this by studying the cones in their eyes). By avian standards mammals are color-blind to colors we obviously don't even have names for. If they looked at our "accurate" color pictures, it wouldn't look right to them at all. Starlings that look black to us might appear a deep -- something to other birds.
Second, while the goal for film might be to reproduce the same color response in humans as if they were looking at the original scene (although that's debatable, e.g. Kodachrome, Technicolor), in engineering an objective is only as good as the tests you measure success with. Up until the 1990s, movie studios shot images of models (the human kind) holding color strips to help film technicians to establish a consistent color balance (link with "China Girl" pictures). These were inserted into the prints so you could check that the print was developed properly. But since models in these pictures were *white*, the test only ensured good results for white skin.
Finally film is far from perfect in reproducing human perception. How many times have you seen an amazing scene, shot a picture, and have the picture come out "meh"? You have to understand the properties of the recording and playback media, and consciously take them into account to get a controlled result.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
White skin presumably means what Europeans have. With a bit of a tan is close to 18% reflectance.
This is the same as a grey card, but the reason is that lots of other things are this tone too - foliage, brick, weathered wood, old roads. It's also, on a log scale, the midpoint between the lightest and darkest tones film can show.
Article is a load of horsefeathers. Films are no more biased against black people than they are in favour of yetis.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."