Dreaming of color balance

[Raul Quintanilla]

Just to clear my previous post: I thought it would be better if we didn't have such a notorious difference between different color balances. I know my eye adapts as well, but it never feels that extreme in comparison to a camera. I don't know how to explain it, I just wish the "color range" could expand just like dynamic range, so there wasn't so much difference from what the eye perceives.

And, I love Eyes Wide Shut. I never said we should not have that kind of colors available, but I think it would always be better to start with a realistic color balance and then, if you decide, you can turn it up just like Kubrick did. Options!

I wasn't really thinking of Auto white balance, it was more of an expanded range, not changing colors, that I can already do with my DVX100 and Canon 7d.

 

[Scott Crawley]

The question you were meant to ask yourself after looking at the Kubrick example was; how would an automatic system deal with that frame? Would it average both the warm and cool temps out? Would it pick one temp and normalize the other? How would it know which is "correct"? Either way it would have ruined the desired effect.

The other point was this: The colors that we are getting, though disperate are accurate, and accurate is best.

 

[David Mullen ASC]

It's very hard to simultaneously create a system that sees distinct subtle color tones with good saturation and minimal crosstalk -- and yet is less sensitive to color temperature differences.

Again, how is a camera to know that an object is grey and lit with cyan light versus painted cyan and lit with white light? So if you put an orange and cyan object next to each other, do you really want the camera to minimize the difference between them? If you put a cyan object next to a blue object, do you want the camera to fail to see the green in the cyan object and render it blue?

The camera just receives color wavelengths bouncing off an object, it doesn't know whether those wavelengths were originally coming from a light with the same color or if the color was created by the color of the object. So while you can do an overall color temp shift towards one bias away from another, lowering the sensitivity to conflicting colors means basically rendering an extremely pastel image with little color separation. Which may be great for minimizing color temp mismatching, but is otherwise fairly useless as a color imaging system.

 

[Andrew Wilding]

Again, how is a camera to know that an object is grey and lit with cyan light versus painted cyan and lit with white light? So if you put an orange and cyan object next to each other, do you really want the camera to minimize the difference between them? If you put a cyan object next to a blue object, do you want the camera to fail to see the green in the cyan object and render it blue?

for static scenes, perhaps the camera could be tethered to a strobe that emits a known white light that would tell the camera what color was what under the fixed light temp. For moving shots, the flash could be triggered during the pulldowm and that colo data applied to the next frame. I'm not saying that this is a good idea or that I in anyway would want my camera to do anything differently then it does now white balance wise. Just thinking out loud.

 

[David Mullen ASC]

Not sure what the camera could do with that information though -- once it determined that one area in the frame was lit with a light with more green in it, it can't add magenta to the frame to correct because the other areas would shift to magenta. And if it subtracted green saturation way down instead, any green objects in the frame would suffer.

 

[Andrew Wilding]

If the white strobe ws strong enough to overpower the rest of the lights and give an accurate reading of colors across the frame, couldn't the camera or more likely the post software then deduce from this the relative color temperatures and tints across the screen and using that informtion automtically creat masks over certain areasbtjt could be corrected differently?

Say you had ten tennis balls in a row and lit from one side with daylight and the other side with flourescent. The flash that strives between frames takes an accurate shot of all of th colors. From the frame the software can then see that all balls are equally colored, and might then deduce that the extra green that is on the balls lit with flouros is a color cast from the lighting. It can draw a mask over whatever part if the frame has a higher green value than the refernce image. Then you could attenuate it, if you wish, without affecting the other parts of the image. Like a secondary in color correction, but without the hues of the objects in the shot making things difficult.

Is this ridiculous? I'm not sure, myself. if it would even work, the strobe continuously flashing in peoples faces would be an obvious downside. And it would only really b a benefit when working in spaces with mixed color temps that you can't control. Those tend to be full of people. People that do t want to look into a bright flash many times a second.

 

[David Mullen ASC]

Yes, but what happens when a person in the "white" light crosses the area of the frame where the background objects are under the colored light? That would have to be a very precise window around them, and tracked.

Even if you save all of this for post (can't imagine a camera doing frame by frame tracking Power Windows with secondary corrections, etc.) ultimately in most shots, human decision-making will be required since we are talking about moving pictures after all, things cross in front of other things, etc.

All this to say that if the solution isn't simpler, then it isn't any better than how we currently handle it.

 

[Raul Quintanilla]

I'm really enjoying this post. This is what I sometimes talk with some friends, discussing the dream camera and then trying to figure out how it could actually happen. But hearing the experts talk makes me think that my idea just isn't possible. Oh well, Scarlet still is my dream camera

 

[michael zaletel]

Maybe a 3D Beam-Splitter Rig with both cameras synched fixed at same point rather than 3D. This will be used for HDR and could be used to solve some of the problems you are thinking about. In other words, maybe the solution isn't one camera but two with different settings, filters and etc. filming the exact same image and then combined in post.

-michael zaletel

 

[David Mullen ASC]

If the point is to help the low-budget filmmaker working fast to deal with mixed color temp issues, a 2-camera beam-splitter rig shooting two different color temps followed up with moving split-screens and rotoscoped mattes in post to combine select areas & objects from each version... well, that doesn't strike me as being more practical and cost-effective, than, let's say, changing some light bulbs on location.

However, a single-sensor HDR design, if it somehow involved combining two different captures, could be used to record two different color temp versions of the same exposure rather than getting extended dynamic range from two different exposures. However, you'd still have the same problem in post of combining bits and pieces from two different images to build a single one. Not sure that gains you much over just using normal color-correction tools to correct areas in a single image.

 

[Antony M]

How about just filming the scene in black and white?