So do you advise to use a tone mapper in all cases, even if it is adjusted to not do much (visually) as a safe guard to ensure the scene is mapped to display ?
If I use the auto pickers in Filmic RGB to set the white and black points, do the set EV values confidently tell me the actual dynamic range of my RAW file ?
The actual dynamic range is defined by ISO as being the range between when the signal-to-noise ratio = 1 and when the raw data is just below saturation.
I donāt know if dt can measure those points or not but the ISO method is clearly defined and discussed in depth here:
http://dougkerr.net/Pumpkin/articles/ISO_Dynamic_range.pdf
Kerr says:
THE ISO 15739 DEFINITION OF DYNAMIC RANGE
The concept behind the dynamic range definition given by ISO 15739 is based on the ratio of the maximum luminance that receives a unique coded representation (the āsaturationā luminance) to the lowest luminance for which the signal to noise ratio (SNR) is at least 1.0.This is based on the very arbitrary assumption that detail recorded with a SNR of 1.0 or above is useful and that recorded with an SNR less than 1.0 is not.
I personally doubt that setting the white and black levels in anything equals the ISO method of measuring DR to an International Standard.
Not necessarily: if the image has no values above 1 (for example, a misty landscape), or only in areas you donāt mind blowing out (featureless, bland sky; a white wall in a high-key image; the Sun), then itās OK not to have it. Alternatively, masked exposure or tone equalizer can also be used to manage brightness. Note effect of simple curves (such as sigmoid, filmic, tone equalizer without preserve details, rgb curve, base curve, tone curve) on local contrast when you adjust global contrast, too (reducing/enhancing global contrast does the same to local contrast, but you never get halos; masked exposure and tone equalizer with preserve details can preserve local contrast even if you reduce global contrast, but you may get halos). See Tone equalizer vs tone curves - #15 by kofa.
However, note that both filmic and sigmoid also affect colours (see the highlights saturation mix when filmicās color science is set to v7, the and preserve chrominance + extreme luminance saturation controls when color science is v6 or below; or the color processing, preserve hue and primaries controls in sigmoid). Clipping can also affect colours, usually in a rather ugly way (when not all channels get clipped, you get serious distortion).
@cedric @kofa Thanks for the links, more bookmarks. Looks like I have got some reading to do.
@kofa I see you only mention letting highlights clip, do you not need to consider crushing the blacks, or do you only need to consider highlights when deciding wether or not to use Filmic RGB ?
Shadows donāt lie outside of the range that 0ā¦1 covers.
If middle grey is at 18%, diffuse white is 100%, and light sources are above that. I find processing easier by first dealing with the main subject in the image (usually around mid-brightness), and then take care of the extreme (low and high) luminosities. If you have a pipeline that clips in early stages, you have to fight to avoid unwanted clipping all the time, while also developing those mid-tones.
I mentioned Histogram Equalization which is a known technique in image processing. Here is your raw image converted by RawDigger (the soft one) but then processed with just one click of Histogram Equalization in ImageJ ā¦ no doubt darktable has an equivalent function:
Ahh, I never considered that. Obvious now you mention it. That nugget is really helpful.
I believe Filmic RGB is meant to tone map by compressing and not clipping, in a way that mimics how film behaves.
I am going to play around with tone mapping without Filmic RGB, try and put all this new knowledge from this thread into practice.
My display profile does have perceptual intent selected, so on a side note can I throw in a related question, which I think is related to tone mapping in general;
The display profile handles the out of gammut colours. If I show gammut clipping in Darktable, I find it very difficult to make adjustments to bring those values back into gammut in a way that gives the same result as that achieved by the display profile. How do you handle out of gammut colours, or do you just let the display profile handle it (with perceptual intent for example) ? If you do, is it common practice to simply ignore gammut clipping and just concentrate on how the image looks on screen ?
@cedric Thatās not my image, but I get your point. I donāt use Rawdigger or ImageJ, I think I am more interested in how you guys handle tone mapping conceptually, even better if it is done within Darktable off course 
I donāt think Darktable has a histogram equalization module, not sure.
There is a nuance and perhaps a reason to gamut indicators in DTā¦ if you choose the actual gamut clipping icon it will show you gamut clipping based on what you have set for your softproofing profileā¦so that might be srgb or a printer profile or somethingā¦ you can also monitor gamut using the overexposure warning toolā¦It has the ability to check or indicate full gamut, lightness only, saturation only or I think also if any one rgb channel is out of gamutā¦ but this uses the histogram profileā¦so you need to decide what tool and what profile to set as your colorspace for your gamut testingā¦
The display profile does not affect exports.
You can find profile for output here: https://www.color.org/srgbprofiles.xalter
See Color management issue for specific image - #23 by kofa
@priort Now thatās interesting about the softproof profile. When I read your reply it came back to me that I recall having read that somewhere before. So much to remember! I donāt use the softproofing because I donāt print anything (not yet at least).
You can set the softproof profile and the histogram profile, as I export to JPEG do you think it would benefit me to set them both to sRGB ?
@kofa thanks for the links, going to take a look at them now. Same question to you about soft proofing, if the display profile doesnāt affect the export profile, should I really be making more use of the softproofing ?
I donāt know, I donāt use it. My display is wider than sRGB, but not by much, so what I see in the darkroom is close to what gets exported anyway. If I see flattened-out regions in saturated areas, I check if reducing saturation helps. Thatās all.
Iām much like @kofa my display is not exactly sRGB but close and I donāt print at this point in time. So I donāt bother checking too much. I do use an export profile from color.org as it has the necessary tables to do rendering intentsā¦the DT profile is a matrix profile and canāt do rendering intents and I dont see much difference if any when softproofingā¦
I guess if you wanted to experiment you could set it up as followsā¦ I leave histogram profile set to rec2020 and so monitor clipping of the working space with that. I then trust my display and output profiles to color manage the conversions to the display using my calibrated display profile icc and my export or output profile of which I use one of the ones found hereā¦
I use the preference or appearance one but more often the preference one as I like the way the relative renders look when I need that extra contrast and strong blacksā¦
When I set one of these profiles in the softproofing profile and enable softproofing I can see differences from my display profile and the rendering intent setting of the display profile does impact what I see for the preview when softproofing as wellā¦ This could be used to make adjustments to match output but what you see and the extent to which you see a differences would likely depend on how far away from sRGB your monitor isā¦but you could always experimentā¦
To the best of my knowledge, matrix profiles support relative and absolute colorimetric intents. They do not, indeed, support perceptual and saturation intent.
relative is, as I understand it, clipping at the gamut boundary; in-gamut colours are not modified. This can lead to a kind of āflatteningā. perceptual contains instructions to gradually compress colours as the gamut boundaries are approached; thus also modifying in-gamut colours.
Actually for me perceptual always seems to look flatter but this can be in a good way when the colors are quite saturated and there is the potential to go out of bounds. THe same when black point compensation is used you get a sort of flatter looking blackā¦ I never really considered absolute as it is rarely used but I suspect your are correct as always as you wouldnāt need to remap itā¦nevertheless I have never seen any change in these while using the DT profile that makes or leads to any visible changeā¦
Agreed. Although many matrix profiles such as āDisplayā come set to perpetual intent, what you actually get is relative colorimetric and therefore the possibility of gamut-clipping.
Gents, interesting about the export profile. I am using an export profile from color.org, set it up over a year ago but I had a very low understanding at that point. I have it set to perceptual, I am going to experiment with this more following your advice.
I like the idea of setting histogram profile to the working space as well, going to play around with that.
My display calibration profile indicates 90% sRGB (I will double check). Canāt afford a new monitor just yet, so this is just what I have. So am I right in thinking I do want a good export profile with a suitable intent, probably perceptual ?
I have been wondering lately if one of my issues with general tone mapping is a misguided attempt to retain all detail, everywhere. I might be making my images too flat, not on purpose, simply lack of practical post processing experience.
May be a silly question this late in to the threadā¦ But what is tone mapping?
The term is commonly used but ācolor mappingā is more descriptive, IMO. https://en.wikipedia.org/wiki/Image_color_transfer
