raw histogram

why can't camera manufacturers create a histogram that displays the true histogram for raw photos? anyone know why it hasn't been done?

Unknown456 wrote in post #15464208
why can't camera manufacturers create a histogram that displays the true histogram for raw photos? anyone know why it hasn't been done?

A "Brightness" histogram of the Raw data would be pretty garbled. Even though the individual "sensels" do capture the photons, the actual values are discreet R, G and B "captures. The data then has to be "interpolated" in a process that allows an "averaging" of neighboring pixels and then "breaking things down" to individual RGB pixels.

This sensor design and process is known as the Bayer sensor design and interpolation process. It was, I believe, developed in the Kodak labs a bunch of years ago because visual/optical scientists determined that a resulting RGB image could successfully "present" itself to the human visual system. They also determined that more "accuracy" could be achieved by using twice as many "Green sensels" than the R and B sensels.

There is an alternative sensor design called the Foveon design used in some cameras that tries to "get" RGB pixels using "layered" R, G and B sensels, but it's not the designed in use by Canon, Nikon, etc.

Now, the thing to realize about the Raw data is that the Raw data is the actual R, G and B sensel data. The camera has not done the "interpolation", that is up to the Raw processor. Than means a couple things: first, the "averaging" of the neighboring sensels has not been applied. I've never seen a study of what the resulting Luminance histogram would be and how it may differ from what we see after the Raw processing, I can't comment on that. But the huge difference would show up in the RGB histogram, because of the green 'bias" that would show up before processing.

So, there are reasons why the in-camera histogram actually uses a processed jpeg that is "embedded" in the Raw file.

Now a consideration that a lot of us who shoot Raw bear in mind is that this jpeg uses the Picture Styles the camera is set to shoot, including a fair degree of Contrast and Saturation, which to a lot of us is not ideal, and so a lot of us set our Picture Style to Neutral and then dial back the in-camera Contrast and Saturation all the way.

What this does is let us view a histogram that would be closer to what you might see in an app like Lightroom and Camera Raw, and also to push our exposures more "To The Right" without the evidence of "clipping" showing up in a misleading way.

There are also Color considerations that have led some to play with the UniWhiBal technique, but I really haven't delved much into that. If you are wanting to know, you can actually do searches for "UniWhiBal" and get good info, and also ask here and hopefully the guys who know that stuff can chime in!

One way I kind of get a 'raw luminance' estimate is just to look at the green channel and ETTR on that - generally, if you equalize all the channels, the image will be mostly green.

Of course, this doesn't work well if the entire scene is red or blue.

Tony one way to do a "RAW" histogram for checking if any one channel is saturated would be to have 4 of them, Red, Green, Green, Blue. As the Bayer filter is effectively alternating columns/rows of Red/Green and Blue/Green, giving the double quantity of green pixels. The resulting histograms would then show each results for each 2×2 sub-set of four red/green/green/blue sensels. That way you could ETTR and truly know if you have saturated any of the pixels. This could be achieved from the RAW data only and would not need to affect, or be affected by, the preview JPEG that is produced in the camera. Of course most camera owners would be terribly confused by RGGB histograms, but I see no reason not to make this available at least on high end cameras.

Alan

Well, there ya go, nobody ever accuses me of being all "on top of this stuff" but fortunately we have some good contributors!!!

Unknown456 wrote in post #15464208
why can't camera manufacturers create a histogram that displays the true histogram for raw photos? anyone know why it hasn't been done?

There is absolutely no reason or justification for it not being done at least in the "advanced models. (A soccer mom with a Rebel might find it confusing.) There have been several free applications with a Raw histogram (the UFRaw converter, Rawnalyze whose author is now diseased and Raw Digger http://www.rawdigger.c​om/ ) so there is no reason why it can't be in the firmware. Uni-WB can be a useful substitute if you don't mind WBing without an approximately close starting place in the converter. It is not a true Raw histogram because it is gamma corrected and Raw data is linear, but if all you use the histogram for is the placement of the white point, it does a good job.

thank you tonylong for your detailed response. I did not really expect to get a real answer when I asked, you have given me a lot of things to learn about. And gave me a direction to start looking. thank you again. do you know if you would you still use the neutral picture style with contrast and saturation set to zero when using uniwhibal?

Unknown456 wrote in post #15464208
why can't camera manufacturers create a histogram that displays the true histogram for raw photos? anyone know why it hasn't been done?

RAW data cannot have a histogram, because it'd have to have settings to be used during conversion from RAW data to image file. How much contrast, how much saturation, what type of tone curve?

If using a Canon dSLR, set your picture style to "neutral" and that will be with as little contrast and saturation as choosable in a picture style.

Unknown456 wrote in post #15468290
thank you tonylong for your detailed response. I did not really expect to get a real answer when I asked, you have given me a lot of things to learn about. And gave me a direction to start looking. thank you again.

Hey, glad to chip in. Realize, though, that I'm not a physics guy or an optical or vision scientist, or any kind of scientest, I've just researched these things from a layman's point of view...

I always have my Picture Style set to Neutral and have my Contrast and Saturation setting set to -4 because I shoot Raw and prefer the extra "latitude" I get from "pushing things!

HappySnapper90 wrote in post #15468307
RAW data cannot have a histogram, because it'd have to have settings to be used during conversion from RAW data to image file. How much contrast, how much saturation, what type of tone curve?

If using a Canon dSLR, set your picture style to "neutral" and that will be with as little contrast and saturation as choosable in a picture style.

Faithful is a better profile than Neutral. IMO.

Raw histogram also doesn't make as much sense now, because both LR4 and CO7 can reconstruct highlight data if one or two of the color channels clip, this makes it possible to pull highlights by as much as a stop more than previously possible.

Kolor-Pikker I agree that LR4 and I will take your word on CO7, as I have never tried it, are really good when it comes to recovering highlight data when only one channel has clipped. It would still be nice at least on some of the higher end cameras to offer RGGB (or RGBG depending on which way round the 2×2 grid you go) histograms from the RAW data. Yes calculating it may slow the camera a fraction but it colud be very useful to those who like to take maximum advantage of ETTR. As I said before it would not affect, or be affected by, the in camera JPEG processing options, although if you are shooting as far to the right as absolutely possible the in camera JPEG conversion is going to look "pants" anyway, no matter what in camera settings you have.

Alan

HappySnapper90 wrote in post #15468307
RAW data cannot have a histogram, because it'd have to have settings to be used during conversion from RAW data to image file. How much contrast, how much saturation, what type of tone curve?

What's this?

Although I am light-years away from the "get it right in the camera" gang, this is one area in which I try to avoid the computer's aid. The converters are amazing in the reconstructions they can do, but it's still a reconstruction; data from algorithms and not from optics. Just like good deconvolution sharpening like DPP's DLO is amazing, but it still can't turn a dog into an L lens. Nor can the best resampling turn a 300D into a 7D. To a confirmed pixel peeper the best image will be the one where the ETTR hasn't overshot the mark.

tzalman wrote in post #15470159
Although I am light-years away from the "get it right in the camera" gang, this is one area in which I try to avoid the computer's aid. The converters are amazing in the reconstructions they can do, but it's still a reconstruction; data from algorithms and not from optics. Just like good deconvolution sharpening like DPP's DLO is amazing, but it still can't turn a dog into an L lens. Nor can the best resampling turn a 300D into a 7D. To a confirmed pixel peeper the best image will be the one where the ETTR hasn't overshot the mark.

Just saying, obviously it's not something you'd want to rely on, but it's there if you mess up.

tzalman wrote in post #15470122
What's this?

Referring to the histograms in your post #12, I realize that there are twice as many green-filtered photocells as there are red and blue, but why have two green histograms when they're always going to be identical?