The Science Of Green Screener
How does it, like, work?
Put simply, it takes your device camera and breaks the green channel (or red, blue or luminance) into 4, 8 or 16 bands. We correlate that to roughly 2-Stop, 1-Stop or ½ Stop resolution between the bands. This article explains the correlation.
The rule of thumb in the industry has been that green screen should be even to within 1 Stop of difference on a spot meter. That’s not terrible advice, although it’s not really accurate. But let’s first take a moment to appreciate that the person responsible for shooting the green screen is the Director of Photography, and since his tool is a light-meter, the advice has been given in light-meter terms.
In reality, the keyer doesn’t know anything about Stops, all it sees is RGB values (256 levels in 8-bit color spaces), and all it cares about is the difference in RGB values. Ideally, we’d like there to be less than 32 RGB values (1/8th of the range) in total variation. This is the real absolute.
That depends on how much latitude the camera has. If your camera has 8 Stops of latitude, one might assume that this would map to 1/8th of the RGB levels, or 32 values. If your camera has 14 Stops of latitude, 1 Stop could mean 18 values. Or if your camera has 6 Stops, it could mean 42 values.
But the conversion from Stops to RGB levels also isn’t linear, and this is the real fuzziness here. Here is an exposure test with Green Screener on 16 Bands (Hi setting), courtesy of Ari Golan at Atomic Imaging Studios in Chicago.
And here is a graph of the correlation between RGB values and F-Stops / Foot Candles:
The F-Stops x 10 column only serves to scale the value for the graph. The Band column represents 16 bands of 16 RGB values each (1/16th of the total RGB range).
It’s clear to see that both the Foot Candles and the F-Stops drop exponentially, while the RGB levels decrease linearly. And this produces an interesting conclusion: That 1 band on the Hi setting represents roughly half a Stop of exposure, or that 1 band on the Mid setting represents roughly a full Stop of exposure… but only at key .
Thus, if you intend to place the green screen at the very top of your exposure range, you can trust that Lo/Mi/Hi represent almost exactly 2 Stops, 1 Stop or ½ Stops of exposure.
But even though DPs often intend to place the green screen 1 Stop under key, looking at the evidence (the green screen plates we’ve seen through history), the green screen often ends up 2-3 stops under key. And suddenly, 1 Stop represents a whole other range of RGB values, maybe 50 to 100. And since RGB values is all the keyer knows, the green screen has now dropped dramatically in quality.
This is never-minding the fact that when you use a light meter to measure a green screen, even a meticulous 1-Stop variance, in practice, turns into several Stops more from all the spots that didn’t get measured. And once you have talent in there and you change the camera angle (which changes the reflectance angle and level of the green screen), your green screen is essentially all over the map.
The moral here is that if you measure 1-Stop variance, you can’t really be sure what you’re measuring. There are too many variables.
The 1-Stop rule evolved because the only tool we had was a light-meter, but in reality, it’s vague. However, it has value if it’s a standard that everyone has calibrated around.
In the same sense, 1 Band on Green Screener doesn’t correlate directly to anything, it’s simply 32 RGB values of an average 6-Stop video camera, but in time, it can become something we calibrate around.
Where before you might have said “If I’m within 1 Stop, I know I’m good”, with Green Screener you might learn to think “If I’m within 1 Band on Mid, I know I’m good”, and in time, that thinking starts to mean something.
Because Green Screener isn’t completely accurate in a scientific F-stop sense. It’s simply RELIABLE in what it does, and it represents what a keyer sees much closer than a light-meter or a waveform-monitor.
To add one more piece to the puzzle, we only measure a single channel at a time, but brightness is a weighted average that uses 71% Green, 21% Red and 8% Blue to produce the brightness. We’re ignoring that we’re basing our calculation on 71% of the true luminance value (like a light-meter would have), because it doesn’t matter. And this assumes pure green. If the green screen is paint-based, it’s polluted with Red and Blue, and suddenly, reading just the Green channel IS accurate after all, because now it’s a proxy for the other channels. Confused yet? And a light-meter isn't exempt from problems with weighted average calculations.
What matters is simply that we learn what 1 Band on Mid or 1 Band on Hi means in terms of the quality of the key. In time, we might learn that we still find the results acceptable if we stay MOSTLY within 1 band on Mid, but dip into a neighboring band once in a while.
This is a much easier, and ultimately much more reliable way to think about green screen.
Finally, a small but important issue is the difference between Normal Mode and Target Mode.
In Normal Mode, the bands are simply divided harshly at 4, 8 or 16 boundaries. Taking Mid (8 bands) as an example, this means that it’s possible to have a perfectly lit green screen that appears as though it’s using two bands, because you’re on the edge of a boundary.
However, as soon as you’re in Target mode, the band is CENTERED on the value of where you tapped. This is important, because you can now trust that if you’re within a single band, you’re LITERALLY within a single band.
Written by Per Holmes, with measurements from Ari Golan, Atomic Imaging Studios in Chicago.
Green Screener is $9.99, and is available for iPad, iPhone and Android phones and tablets. Simply download from one of the App Stores below:
"This app does exactly what it is designed to do....help you light your green/blue screen so keying works perfectly. Found our lighting problems within 1 minute of purchasing this app."
Conover Company (via App Store)
"TI have been looking for something like this my whole career (25 years as a gaffer). My light meter works, but its slow to check then change then check again. We set this up on the set today. I then had my guys adjust the lights until it was even. Fast and easy. The key was perfect. Thank you for making programs that work."
Bananagrip (via App Store)
"If you work with green screens then you know they should be lit evenly. This nifty tool shows over and under exposed areas. Boom."
Tarpaper (via App Store)
"I love the one star reviewers' comments - "can't film with it or add any effects. Stupid app". Oh, you SO don't have a clue what this app is for. But professional videographers do and this app is brilliant. It takes all the guesswork out of chroma key, especially when I'm out in the field with a portable green screen. No more worrying about what I've got until I get back to the studio. Bravo!"
Cedar Park Tx (via App Store)