Per recent discussion in this topic I was persuaded look into using “scene referred (Agx)” rather than the “scene referred (filmic)” workflow that I’ve been using. I notice that setting both white & black relative exposure is done in both. I have no reason to think the criteria is any different between the 2 modules.
In my case, I started using filmic rgb quite some time ago (might have been Version 3 of Darktable). Back then I seem to recall reading articles that indicated that the EV exposure distance between the white and black setting ought to match the range associated with the camera lens. This is something I’ve been doing for quite some time even though I have noticed that both settings have an eye dropper associated with them that could be used for picking a value.
Given that Agx is completely new to me I decided to follow the advice given in the Recommended Workflow section of the User Manual which says to use the “auto tune levels” picker to set these values. I was surprised to see a dramatically different range, around half as much, than what I’ve been using (which has been done by first using the picker to set “white relative exposure” and then the dynamic range of the lens to set “black relative exposure”.
I’d like to know what people with more knowledge than I have think about setting exposure range in Darktable. Especially if there is any different between Agx and filmic.
It is totally different. AgX does not really set black and white.
Filmic calculates a norm, and uses the norm to control black and white. It acts on (applies the curve to) the norm value, and then calculates RGB values based on the mapped norm value and the original RGB.
AgX works on reach RGB channel separately. It finds a lower cut-off point (which I called black on the UI, but which is not black, just the lowest of all RGB values) based on the minimum of the RGB values, and the maximum RGB channel value is used for white. In practice, this works well, but it’s not truly ‘black’ and ‘white’.
I have no idea what this means, though:
I always used the auto pickers with filmic rgb, too.
I always used the auto pickers with filmic rgb, too.
[/quote]
My mistake. Not sure what I was thinking when I said “lens”. It is rather the camera’s sensor which, I think, means that is maximum range in the raw data that can be associated with that sensor.
Afaik, a lens has no associated range. At least nothing to do with the dynamic range of the image, which is the EV distance between black and white reference in filmic. If that range should match anything, it would be the dynamic range of the intended output medium… (note: “intended output”: screen and paper don’t have the same dynamic range).
Even the sensor’s dynamic range no real relevance for filmic’s white and black references, which is geared towards the output, not the input. Given the editing done before filmic (or AgX), the sensor range can be heavily modified, through use of modules like “exposure”, “tone equaliser”, “color balance”, …
Out of curiosity I decided to make a duplicate on the first image I’ve tried to develop using Agx. Then the only thing I changed was the “black relative exposure” which I reduced to the point where the range (difference between white and black point) of exposure equals the range of the camera’s sensor for this type of shot.
First, an image based on using “auto tune levels” to set both “white relative exposure” and “black relative exposure” follows:
Then, an image based on changing only the “black relative exposure” to provide a range matching that of the camera’s sensor for landscape photographs follows:
In this case, the range for the original development is 6.91 whereas for the duplicate (i.e., camera based) is 12.39, which I’d say is pretty significant.
While it is quite reasonable to question whether fine points associated with the development of raw files might get lost in a .jpg format development I would offer that I can’t tell any difference on my display. While I wouldn’t say my display is the best available it has been calibrated and is 27" with pretty good color rendering.
I wouldn’t overthink the technicalities and the mathematics involved in either filmic or AgX. I find that auto tune levels picker in AgX works really well for 99.99% of my images. It sets the highlights and shadows really well in my view. When I used filmic I nearly always had to pull back the white relative exposure slider, but this is not the case with AgX.
These two modules have some similarity in the UI look but I would not try and do a direct comparison. What is really powerful in AgX is the shoulder (highlights) and toe (shadows) power adjustments. I view these as ‘contrast’ sliders for the shadows or highlights. Keeping the toe at default usually gives good shadow details while increasing toe power adds contrast and impact to the shadows but some details may be lost to blackness. It is all about the look that is best for the image. I like dark shadows often in my images.
The shoulder power slider can often be set very high without adverse effects. If either the shoulder or the toe region is shown as orange then the power slider is ineffective. It is not necessarily a problem that needs to be corrected. In filmic if any part of the curve went orange it indicated a problem that needed correction. That is not the case in AgX so again be careful of direct comparisons.
It is also worth looking at Sigmoid too as a filmic replacement. Although I usually use AgX.
Thanks for the advice. As you know Agx is new to me and I do think I should check out Sigmoid as well. I should point out that I also developed the image shown above using filmic rgb (i.e., my former method). The results were pretty similar but I’d say the Agx was slightly better (i.e., a little sharper and the sharpen module was not used on any of them).
My real motive in posting this topic had to do with checking on the validity of my former method of setting white and black relative exposure. Based on this discussion, I now think there was no validity to what I was doing.
While the results of this initial experiment turned out pretty good I still like the idea of being able to check the black and white point and make adjustments if necessary. The exposure module supports black point but there is nothing in these trials that does white point. What about the idea on enabling the shadows and highlights module and setting both shadows and highlights to zero correction? Note: I do plan on also experimenting with tone equalizer as a replacement for shadows and highlights adjustment but I don’t think it provides any capability to adjust white point.
Remember too that in the scene referred you are using exposure to set the “middle” gray anchor…then DNR is mapped around that …thinking of black and white points as boundries is a sort of display referred mind set… Free yourself a bit… I have found that using the picker in the exposure module on an area that I want well exposed and to be the focus will often make adjustments that I like but I might not have gone that high… If I dont like it I still have a ballpark for exposure and I can bring it up or down manually. Then just let the tonemapper do its work and map your DNR around that…you get to decide how the stops are distributed around the anchor…
Since I adopted kofas AGx Sigmoid Matched presets I usually don’t have to worry much about the “white point”. AgX “needs” a healthy dose of Shoulder Power (which the current default lacks) to properly use the upper range of the output-dynamic range.
Also there is a multitude of ways to control the highlights and point where they clip. Tone Equalizer, Colour balance RGB with brilliance grading, AgX Slope control, Shadows/Highlights module, local contrast module. None of those are inheritly wrong or right. If you look through some playraws on here you will see how different people use darktables toolset.
I personally try to use few modules after AgX in the pipeline but that isn’t inheritly better then any other way :-).
I would advise strongly against the slope control. That, lift and saturation are only present because I had a bad understanding of AgX in the beginning. slope influences the whole image, shifting mid-tones as well (and so does lift).
Using darktables tonemaooers doesn’t work with that “set black and whitepoint” usually done in several different raw or image editing tools.
for darktable tonemappers it’s more important to set the midtones correct. (keeping the tonemappers default settings is completely fine in this phase). There’s no “absolute” midtone - you decide, what’s the import part of your image which shouldn’t be part of tonality compression.
Once thats done you can use the black/white and curve controls to bring the ranges below and above this midtones into a proper shape.
btw; hopping between different tonemappers doesn’t help - they share common concepts but are quite different on the controls (same name doesn’t imply exactly same function).
Better start with the default (AgX recommended) and learn how to best make use of these controls.
Then maybe check filmic rgb, if you expect a different color rendition.
Why not use it? If one is cautious of those caveats it can really help bring in some pop or final polish IMHO. Even though I rarely use it. Midtones are not an absolute truth to me, I go by gut feeling.
(I wanted to bring in some tooling analogy, but the only one I could come up with was using a chisel top open paint cans but…thats just wrong. But probably also how you feel about people using slope and lift ).
Because a control for highlights (or shadows, in case of lift) should ideally not influence the whole tonal range. These controls are not part of AgX, they were simply part of the first GLSL shader that I ported – but that shader did not have any other adjustable parameters, such as black/white point, contrast, toe/shoulder power etc. Also, scale can easily produce values > 1, which break the guarantees offered by the tone mapper (mapping to display 0…1).
You’re free to use them, of course. I was just voicing my opinion.
I am wondering why you are so focused on white point. When I edited images in GIMP which were usually jpg I was focused on setting balck and white point in levels to maximise the contrast without clipping. Now I am using raw files in DT I just rarely give much thought to black and white points and rather consider highlights and shadows.
I think you hit on it. I use Darktable (DT) for developing raw files. My normal approach is to create a file in .tif format that preserves the camera sensor’s full resolution (i.e., do not omit any pixels). The idea being to produce an image file that doesn’t loose any quality. Once I have that file I typically use GIMP to produce the finished product which can differ quite a lot depending on purpose. In that, printing which includes different sizes and adding text or maybe borders. Those that go to my website are converted into low resolution jpg format.
You mention the Levels tool in GIMP I find that using the auto level feature very often makes improvements I want to keep no matter how good a job I did developing the raw file.
My idea on black and white point is simply inspecting the histogram and trying to use the full range without clipping. My idea is that avoiding clipping is also a good idea in DT.
I respect the math and for sure you know the module inside and out, and yet like @mino I will use them from time to time…early on I found that images with just a little bit too much crush would be visually improved by 2 or 3 click of the mouse scroll wheel of lift. Similarly with slope again used with extreme restraint sometimes I found again just a touch would brighten the image in just the right way…probably I could and should have used some perceptual brilliance the way that Boris always does but I have found that I often try just a touch of slope and sometimes its just that little bit of contrast that seems to bring the image alive… So they may be mathematically or operationally flawed but yet I can convince myself that just some minor tweaks using them often makes worthwhile improvements to my image with very little effort invested so I am glad you kept them in the module…
In GIMP I too often like the auto levels tool and sadly auto levels in DT is no where near as good in my hands. That is auto levels in the levels module, but the auto tune levels in AgX works well for me in giving a nice shadow and nice highlight.
I am probably not convinced that this is using DT to its full capabilities. Since using DT I have all but abandoned GIMP, which is my bad because GIMP has some useful options. I have rediscovered GIMP for some ‘healing work’ in my panorama stitching when the software has a problem with alignment. I have always used GIMP for photo restoration, but it has little role in my standard digital camera edits because DT is fully capable in this regards.
I mention panorama stitch which means I at times have to be working with a tiff file or png file. I feel handcuffed when working with this file format but I have no option. So reading that you deliberately handcuff yourself by converting the raw to tiff and then doing edits is hard for me to understand. GIMP is a pixel editor for graphic design in my view and a photographic editor as a by-product.
But ultimately you need to be true to yourself and do what gives you enjoyment in editing photos and achieving the look that you want.
Maybe the fact that I had been using GIMP for quite some time prior to discovering Darktable (DT) has something to do with it. This goes back to before I even knew what raw files were. In fact, my interest in acquiring cameras capable of producing raw files was triggered by seeing the kind of adjustments that could be performed using software for post-processing. Prior to discovering DT I used both the software provided by Canon, the manufacturer of my cameras, and Rawtherapee for developing raw files, which did get good results from my perspective. When it comes to my overall workflow DT is a substitute for those programs.
Is there anything wrong with cropping files (e.g., for different size paper) using GIMP? How about adding text? Even for scaling and converting to other formats I suspect there is little or no difference in the final result.
When it comes to editing I also know how to use GIMP to do things like removing a bland sky and replacing it with a better version. Can I do that with DT? Even if so, I would intend to preserve the results of the raw file development prior to making such changes.
Many users on this forum use RT and I will not knock that program. For me I chose DT because of the drawn and parametric masks that allowed me to localize my adjustments. Prior to DT I would use RT but then have to go to GIMP to do some localized adjustments using layers and masks. But the masking in DT effectively replaced most of my need for GIMP so I mainly use DT. In DT I can export in various formats. I can also add simple borders and captions for my pictures. I also downsize images for posting on the internet.
GIMP has many uses still, but DT is the superior photo editor because that is what it is designed for and only for. GIMP is a pixel editor designed for graphic artists but can be used as a pixel editor for photographs. So replacing skies, adding cats to photos, making collages and preparing composite scientific images for publications is certainly where it excels.
If you are comfortable with your workflow using GIMP and get the results you want then that is all that matters. If you prefer working with RT then there is no reason to use DT as both are good programs.
I use the canon software only for printing since I have a Canon printer, but it is reasonable software for photo editing. Certainly better than the software supplied by Nikon.
BTW, sky replacement can be achieved using the composite module in DT but I personally would use GIMP for such a task. I also sometimes take advantage of the healing tool found in GIMP for photo restoration. The restore module in DT is very good but is a bit clunky and slow when you have extensive amount of spots and scratches to retouch.
Be careful with that kind of thinking. No doubt many a perfectly good photo has been ruined by an insistence on always stretching the histogram. Use it to verify no clipping is happening, but otherwise trust your eyes that the photo looks good.
That depends. When setting exposure, adjust for mid-grey and don’t worry about clipping, since you can always bring the highlights back later. Up until the tone mapper, it is in fact impossible to clip them in a way that they can’t be recovered.
).