Cameras: Are on-camera histograms trustworthy?

[willdoak]

According to Deke McClelland (Photoshop CS2 Mastering Camera Raw, "Shooting Considerations" if you subscribe to lynda.com), on-camera histograms can't be trusted when you're shooting raw, because they "measure the results of JPEG-type presets [that would be] applied to the raw image, not the raw data itself."

So when shooting raw files, you might not be clipping highlights even though the on-camera histogram indicates you are. He notes that some cameras are more accurate with their histograms than others.

I've attached his example of an in-camera histogram he superimposed over a well-exposed raw file, showing how far off the in-camera histogram can be.

Any thoughts on this subject? Anyone done any empirical testing of their cameras' histograms?

I'm using a Nikon D80, FWIW. Maybe I'll give it a shot (so to speak).

Cheers,

Will

[TommyO]

Re: Are on-camera histograms trustworthy?

This would be consistent with other authorities on the subject, as well as common sense. With the limited processing & RAM capabilities of the camera, one would expect only the higher end camera processors to handle a raw histogram properly. Most vendors also have to spend more money to work with third parties to develop the algorithms, whereas they have already made the investment for JPEG's.

[dmrdm]

Re: Are on-camera histograms trustworthy?

I'm no expert at all, but I've read & heard the same thing. I run the D-80 myself...and don't pay attention to the histogram because of what I read about that and shooting RAW. I have the blinkies turned on. Guess I've shot with the D80 long enough now to know when it clips, which it does alot.

My husband has the Nikon D300 and it's senors are so much better than the D-80. Believe it has what is called a scene recognition system. So shooting the same say landscape shots with the 2 cameras, his RAWs come out of the camera a lot better than mine. (wouldn't be sad at all if I dropped my D-80 someday and had to get the D300)

[Craig Walters]

Re: Are on-camera histograms trustworthy?

basically, RAW is exactly that, raw. it's supposed to be data only, no header information, at least in its most pure form. it's just straight out of the sensors data. there is no format, no rhyme, no reason, just raw data. therefore, you'd have to build a headered image within the camera from the raw and do your histogram on that. and that's what a jpeg/raw combo does. so, there is no way to do a true histogram on just raw. you have to convert it to something first.

so, depending on how your camera builds the jpg or whatever it's using, the histogram is based on that and will only be as good as 1. the sensor capture, 2, the raw storage, 3, how the jpeg is built- how well it's done, and 4, how good the histrogram algorithm is.

the simple answer is, you cant build a histogram directly on the raw because the raw isnt interpreted yet. it has no meaning. it's just 0's and 1's. you have to give a significance to the 0's and 1's by making it a structured file type first.

hypothetically, if you were doing this all in-house, you could build a histogram straight from the raw. it would mean having software embedded in the camera that interpretted the 0's and 1's straight to the histogram, but since most cameras that do raw are also capable of doing jpeg, it's like tommy says, why build a second histogram when you already have the jpeg one?

and the other reason i can see is, why worry about a raw histogram when you have instant visual replay through the lcd and the capability of bracketing shots by ev (exposure value).