RGB Values

[Von Thomas]

Ok, so I come from the digital still side of town. In most still RAW conversion software, the RGB values are 0 for black no detail, 127 mid gray or 18% gray, and 255 white not detail. Will RED software ever have RGB values to represent this? It would make sense to me, as RED as I'm to believe started with a digital still in mind. It would be totally in line with those transitioning from the still world.

Also why is the histogram the tool most overlooked by those using the RED.

 

[Von Thomas]

I've attached a file from a Canon 1DS, an 11MP camera. The image was processed in Phase One's C1 Pro raw capture software. In photoshop the gray RGB value is 120-120-121, which is very neutral, and also tells me that my exposure is pretty good. I'd love for RED Alert /RED Cine to utilize this method. I know it won't happen over night, but think of how great it would be.

The second image is a screen capture from Phase One's C1 Pro. As you can see, the RGB values in the top center represent the gray portion on the card the girl is holding.

 

[Paul Leeming]

Von,

the Red histogram is number agnostic, so white is still on the far right side and black is still on the far left.

But the 0-255 values derive from an 8bit JPG image (usually). Since Red RAW footage is 12bit I think that's why you get more RGB values per channel (and more accurate colour as a result).

I imagine that even when working with Canon RAW images though, most photo software retains the 0-255 moniker more for consistency than accuracy.

HTH

Paul

 

[Von Thomas]

Thanks Paul, so for my edification, can you tell me what are my equivalent values for white (255), gray (127) and back (0).

What I'm saying is that the digital still world has been using there values to represent RGB for years, and it burned into out minds. RED is no different than any still camera, with the exception that it can capture motion.

All the pro digital still cameras have bit depth much higher than RED, and the industry still maintains those values.

 

[Paul Leeming]

Well, 2^12 is 4096 values per colour channel, so:

White = 4095,4095,4095
Grey = 2047,2047,2047
Black = 0,0,0

Don't worry, I use Photoshop a lot so I understand your frustration

HTH

Paul

 

[Von Thomas]

Thanks Paul.

In looking closer at one of my clips, I see 2346,2315,2286 (gray card). Now this is after I click pick WB in RED Alert. The blue channel in the scene is much lower than the other two values. Is that due to being shot with tungsten light and the blue channel not receiving enough illumination?

But wait a minute, wouldn't it be the other way around,

Black = 0,0,0
Grey = 2047,2047,2047
White = 4095,4095,4095

But wait again, I'm confused as yes Black = 0, but White should be more like 9999, this is not making sense to me, I want my 0, 127, 255 back.

 

[Paul Leeming]

Oops, my bad, you are absolutely correct. This is what happens when you post technical answers at 0300 in the morning!! :cursing: Original post edited.

The maximum value for white is 4095 because it represents the highest discrete value possible in a 12bit system (using 0 as the first value). 9999 would only be a valid value for 14bit systems and up.

This is all to do with the base2 (or binary) way of numbering, since a digital system can only ever have two states, on (1) or off (0), quantum physics notwithstanding!!

For example, the value of 3580 is the same as 110111111100 in base2 nomenclature, which is how the computer inside Red sees it. 111111111111 is the highest possible 12bit number which equates to 4095. 000000000000 is the lowest 12bit number which of course equals 0. All other values can only fit within that framework.

9999 is only possible in base2 using 14bits minimum: 10011100001111

As you can see though, each stop (or bit) of precision gives you vastly more colour values to work with for more accurate reproduction to our analogue eyes. This is why we are always seeking out higher bit depth LCD panels for example, as we get less colour banding visible to our eyes, but the tradeoff is the higher precision is more difficult to manufacture and process, so 10bit panels cost a lot more than 8 or 6bit panels.

So whilst I understand your wanting to keep things simple, you are actually limiting your ability to hit a precise colour when you can only adjust a 4096 step colour channel in 256 steps. This is why Red tools give you the ability to adjust to the finest degree possible, because they are professional tools not limited for simplicity's sake.

HTH!

Paul

 

[Paul Leeming]

Incidentally, this is also why underexposing by 1 stop halves your colour precision and doubles your noise potential.

You get 2048 possible values in the highest stop, from 1000 0000 0000 to 1111 1111 1111 1111.

But underexpose by one stop, and suddenly the highest value possible is 0111 1111 1111, fully half of the possible values had you exposed to the right. Each stop further down the chain halves the values again, giving you less discrete steps between white and black (or red, green and blue since we are talking three channel colour).

The lowest stop gives you just 2 possible values for each colour channel: 0000 0000 0001 or 0000 0000 0000. That means it's either full red, green or blue, or black, with nothing in between. Visually this manifests as chroma noise, since the sensor hasn't got enough precision to get it right and has to err one way or the other with nothing in between. So the real colour value is lost and ugly noise results.

Exposing a digital sensor to the right means you maximise your colour channel precision and thus keep chroma noise to an absolute minimum. You can always crush the blacks later on by decreasing their binary values mathematically (and thus more accurately), but you can't gain precision in the binary space from values which weren't there in the first place.

I hope this helps explain the colour numbering system a bit better, and why the ETTR method with digital sensors is the best way to shoot for post.

Cheers,

Paul

 

[Paul Leeming]

After all that though, remember that in the end it's about the art of colour, not the science....

Cheers!

Paul

 

[Tai Wah Lim]

After all that though, remember that in the end it's about the art of colour, not the science....

Cheers!

Paul

Paul, the images are beautiful. Lim