A better Picture Styles choice = better RAW histogram?

[canoli]

I'd like to run this by you guys - here's a quote from the 40D Field Guide:


"If you're shooting RAW capture, you can set the Picture Style setting to a lower contrast and get a better overall sense of the RAW histogram and reduce the likelihood of clipping (or discarding image pixels)."


I always routinely push up against the saturation point in the HLs. So anything that can help prevent clipping is a good thing for me.


But does this work? The in-camera histogram is a .jpg conversion, one possible rendering of the RAW data. According to the Field Guide when you tweak your Picture Style to show a lower contrast version you're seeing a better histogram. What are they basing this conclusion on?


Do the Picture Styles actually change the capture? If I specify more saturation do the pixel wells, the receptors, really fill up faster?


I always thought the RAW data is what it is, based on the exposure only, and it remains that way no matter what combinations of in-camera sharpness, contrast, saturation, hue (the Picture Styles) you use. The Picture Styles show different renderings, that's all. They affect the .jpg output but not the RAW data. I thought that was the whole point of RAW capture. If I'm right, then choosing a low contrast render may actually be a less accurate version of the RAW data.


Have I got this all wrong?


Will someone shed a little light here for me? Do these Picture Styles affect the RAW data?

[Daniel Browning]

I'd like to run this by you guys - here's a quote from the 40D Field Guide:
"If you're shooting RAW capture, you can set the Picture Style setting to a lower contrast and get a better overall sense of the RAW histogram and reduce the likelihood of clipping (or discarding image pixels)."
I always routinely push up against the saturation point in the HLs. So anything that can help prevent clipping is a good thing for me.
But does this work? The in-camera histogram is a .jpg conversion, one possible rendering of the RAW data. According to the Field Guide when you tweak your Picture Style to show a lower contrast version you're seeing a better histogram. What are they basing this conclusion on?

I don't know what the author is basing it on, but he is at least partially correct.


Do the Picture Styles actually change the capture? If I specify more saturation do the pixel wells, the receptors, really fill up faster?

No, and I don't think that's what the author meant in your book. I think he meant that it gives you a more accurate histogram, so that you can change the exposure and get a more optimal exposure for raw.


I always thought the RAW data is what it is, based on the exposure only, and it remains that way no matter what combinations of in-camera sharpness, contrast, saturation, hue (the Picture Styles) you use. The Picture Styles show different renderings, that's all. They affect the .jpg output but not the RAW data. I thought that was the whole point of RAW capture.

That's all correct.


If I'm right, then choosing a low contrast render may actually be a less accurate version of the RAW data.

As you said, the in-camera histogram is based on the preview jpeg, which is often quite different than the actual raw histogram. If you want to try to make the histogram more similar to the raw file, there are many settings you must change:
Picture profile (tone curve, saturation, contrast, etc.)
White balance (this is a *big* one, research "UniWB")
Color space
Auto lighting optimizer
Highlight Tone Priority
Exactly what the settings need to be depends on each camera model. Some are more accurate with sRGB (like your 40D), others with AdobeRGB. If you have the contrast too high, it will make the histogram clip even when the raw file isn't clipping. If you have the contrast too low, it will make the histogram show no clipping, even when the raw file is clipping. What you want to do is find the settings that most closely approximate the raw histogram. Usually this means a picture profile with no tone curve and contrast set to 0 (not -4 or +4.
Hope that helps

[canoli]

definitely helps Daniel, thanks.


So the way to
make the histogram more similar to the raw file is to setup a shot on a tripod, shoot a RAW capture, then shoot the various Picture Style combinations that seem likely to produce a histo that most closely resembles the RAW histogram. Assuming the light remains stable, WB, HTP, etc. should be identical for all captures - exposure too of course.


Sound like a plan?

[Chuck Lee]

Sound like a plan?


Yes, that's a plan but still doesn't help in all situations. I just read this same advice over at TheLuminous Landscape ("http://www.luminous-landscape.com/tutorials/right-hista.shtml)and tried it out. Looked good on first couple of attempts but missed severely on last couple of attempts. The raw was off by 1 ev which is not acceptable.


An easy way to do a accurate raw histogram is to shutter a shot with the lens cap on. Then use that image to set the white balance. Your photos on the LCD will be green but your histogram will be correctly displayed for the raw image captured.


I've looked for the source for this info but can't remember where I read it. Maybe Daniel knows where this tip is from. We've discussed it before in another post.


This black image white balance does work well but the green screen output is very annoying. I find with digital that in a pinch if I'm really unsure about my exposure, I'll bracket and shoot 3 frames.


My method currently is toset proper white balance for subject. That usually displays the best histogram for raw on my 5D. As long as my images are well exposed I find little to do in post. I would rather expose for the jpg histogram and loose 1ev of raw overhead than to try and always fill that dynamic space. The newer bodies like my 40D have up to +2ev of headroom but my 5D has only about +1. If I do get by chance some blown jpeg highlights I have a better chance at highlight recovery by saving the +1ev overhead than if I had filled it and overshot the highlights. When you uae a black image WB histogram this is precisely what you are doing and it can lead to undesirable results.


When shooting for $ and precious memories I'd rather be safe than sorry.

[Daniel Browning]

I've looked for the source for this info but can't remember where I read it. Maybe Daniel knows where this tip is from. We've discussed it before in another post.


It's from Vincent Van Gogh. He invented the "UniWB" technique we use for accurate raw histogram. It has that green screen output you talk about:


http://img2.pict.com/49/b2/30/1569871/0/firstuniwb.jpg


[;)] OK, actually I heard about it from Iliah Borg, Guillermo Juijk, and Gabor Sh. (panopeeper). Here's one link on it:


http://www.guillermoluijk.com/tutorial/uniwb/index_en.htm ("http://www.guillermoluijk.com/tutorial/uniwb/index_en.htm)


And here's some more info:


http://www.dpreview.com/forums/read.asp?forum=1021&message=31582853 ("http://www.dpreview.com/forums/read.asp?forum=1021&message=31582853)



When shooting for $ and precious memories I'd rather be safe than sorry.


Agreed. Better to have noise than be completely clipped.


Personally, I don't mind the green review image at all, so I always use UniWB, and only switch to "normal" white balance to check color every once and a while. Having a closer histogram is more important to me.

[Chuck Lee]

It's from Vincent Van Gogh. He invented the "UniWB" technique we use for accurate raw histogram. It has that green screen output you talk about:


Absolutely hilarious! LOL....