Taming the red channel

rawtherapee

#1

Hi, I am new here. Kudos to the Rawtherapee devs, I find it so far the easiest to get the colours I want with my Olympus m43 gear.

I have a scenario where it feels the reds are boosted heavily with most curves I try, By default I use the auto matched curve with Adobe DCP file for the camera, but will use bundled profiles here to illustrate.
eg.

I just screenshotted these 3 from preview as I had too many images (new user 4 image restriction):
image

Resulting image clips the reds heavily (screenshot):
image

This helps to control the clipping:
image

I use this as a sample because the histogram shows it very clearly, even though I seem to not lose too much applying the RGB curve. In another one I need a more extreme curve and visible colour and luminance issues (I had them all on here but that number-of-images restriction). Somehow the Olympus jpg outputs are managing a better job than me, although with a little bit of clipping on some of them.

Any suggestions how to compress the red channel and maintain colour and luminance across the rest? Or am I too pedantic about clipping?


#2

Try enabling highlight reconstruction and apply some highlight compression.


(Shreedhar Inamdar) #3

In the LAB module of the exposure tab, the CH curve is ideal to manipulate the saturation of hues. Have a look. I find it extremely useful. See:


#4

Thanks James, I did this write up a little after actually doing this, so forgot to mention that even at full 500 Highlight compression and 0 threshold (see below pictures) I still have quite a bit of clipping. None of the highlight reconstruction settings make much difference.

image

image


#5

I should note the histogram from the Olympus jpg:

image

It’s as if they really just managed to compress the red channel.

It may just be that I misinterpret the histograms but from my original post it looks like the red channel seems to be boosted more with the tone curve than the green and blue, or I am missing something.


#6

Thanks @shreedhar. I relooked at this because I had tried the CH curve. Basically I have to turn it down quite a lot and the saturation drops significantly and I can’t seem to put some back without blowing it out again.


#7

Looks like it isn’t the channel per se but the red jacket. Maybe try one or both of

image.

Tweak the values to see if it helps. Or try masking the jacket and only concentrating on it.


#8

Thanks @afre. HDR tone mapping (I am on version 5.4 linux mint) needs extreme values to take away clipping and Tone Mapping just worsens it at any setting.


#9

If highlight compression isn’t giving enough effect, you can try reducing the exposure slider and using a tone curve to compress the highlights, whilst compensating for the reduction in exposure for the darks and mids. Something like this
image

Also, as @shreedhar reduces chromaticity of reds with the L*a*b* CH tool, you can also reduce the lightness with the LH tool. Also, although not targeting just the reds try CC and CL tools. CC can take quite a strong curve, something like this
image

Also, if you are using tone curves, try switching their mode to standard rather than film-like, film-like increases the saturation.

Does the blown red channel matter? If it still looks OK with the other two channels providing variation in the red jacket, then don’t worry about it :smiley:


#10

Most probably, you want a combination of most of these techniques each applied subtly, to keep the image looking natural.


#11

I am using the dev version of RT.


#12

Those flourescing garments and accessories can cause such ruckus… We’ve had play_raw threads with fluorescing jackets and/or dog collars. Every interpretation of the fluorescent colors is different.

The raw histogram neutral profile looks decent, you can still see the red pop out above everything else, but you still have headroom. Whatever you’re doing to obtain the clipping (like using the Auto-Matched features), perhaps do less of that. Use a tone curve(s) or the L*a*b* LC/whatever curve to bring out what is going to look much darker than the fluorescent colors.


#13

This [PlayRaw] 3: Phil has a bright jacket. Maybe look at what people have tried.


#14

@James

I can get ok results starting from your tone curve and adjusting, definitely possible to tone down the clipping.

I have never tried CC and the extreme curve would not have been intuitive for me, thanks for that. The LC does a good trick of it though:

image

And yes the clipping is not too bad, I can live with it here, but I want to grow and deal with it well in the ones that matter, so cut my teeth on these :slight_smile:

@HIRAM

Ok it’s good to know it’s not just me! That these fluorescents are a bit of a pain…

@afre

Thanks for the link it would indeed be useful to go through it… Lots of attempts! :slight_smile:

Overall I am working hard on my standard profiles so I can do more work now and hopefully much less later. I was hoping there was something I could make part of my std curve or at least a special profile for this situation. Especially for just the homelife documentary stuff like this.

Learning about the fluorescent issues others also have helps me be realistic about it.

Thanks so much everyone for your time, I really appreciate it!!


#15

I forgot to mention changing curve to Weighted Standard and upping Saturation also worked, although not sure about the overall colour profile…


(Jacek Gozdz) #16

Just a question: are You using dcp color profile with tone curve checked in? If so - turn it off and use Your own tone curve in Exposure tool (I personally use weighted standard, and then compensate saturation with CC curve from LAB tool, this boosts saturation without oversaturating).


#17

Thanks @cuniek

I always have tone curve off and use auto matched curve. So I played around a bit and found it really hard to use auto matched curve with lab CC curve (easily clips, before saturation looks good), but if I do my own rather simple curve it seems to work well. In fact I think the skin tones are coming out a bit better than other approaches.

Definitely another approach in the toolbox… I will also try to see how well this works as a general profile, but I often find something works well in a narrow range of scenarios, and then it is lots of compensatory tinkering to get something relatively general.