B&W Processing with darktable - module choice & ordering

For sure I’ll post a Play Raw if I come across a concrete example. But I’m not really having any difficulties per se, I can usually always get the results I’m looking for. I was just wondering if there were easier or cleverer ways of working with modules for individual colours. I brought the topic up because some of your posts (like this B&W one and the Channel Mixer one) bring up new techniques I’d never thought of before and I find them enlightening. Also, I realized there was no direct equivalent of Lightroom’s colour mixer module, which I remember from my Adobe days. It allows you to play with the HSL on individual colour ranges very quickly and easily. It’s nothing special, but it is very simple.

Is there a reason that you and everyone are making this so complex? Maybe I am missing something? Why not use color zones to do your B&W conversion? Just select the black & white film preset in color zones and then adjust the colors to get the B&W look you want. That is what I do.

Note I am a complete novice darktable user. I am just trying to learn how to use it and experimenting right now. If all goes well then maybe I will leave Lightroom behind. I am using 3.2.1 on Win10.

3 Likes

Image processing is complex, we just don’t try to hide that complexity.

If you’re happy with the look you’re getting, then keep going! Nothing wrong with that.

1 Like

Is my suggestion to the OP an inferior way to do what he wants?

I don’t think so, as long as you’re using the module at the correct place in the pipeline. Color Zones uses the LAB part of the pipeline, so if you push the module too much, you may end up with harsh transitions or banding.

I still use Color Zones a lot, but mostly for smaller adjustments, so I don’t experience the downsides of working in LAB.

Thanks for the explanation. I thought I read somewhere in the documentation that the default pipeline was chosen to be optimal and normally shouldn’t be fooled around with. It sounds like what you are saying is that color zones may not be in the optimal place in the pipeline. If that is correct then shouldn’t it be in the optimal part of the pipeline? Or are you saying that by default it is in the optimal place, but as a user I could screw up and move it to a non-optimal place?

For most cases, the modules are in the optimal place in the pipe by default.

WmBrant, did you find a good way to do your B&W conversions? 3.4.0 is out now so you may want to update.

1 Like

I have re-read the discussion, as bw processing is something that is of great interest also for me, and simply I wanted to point out that the new dt manual has an excellent tutorial on monochrome processing:

https://www.darktable.org/usermanual/en/guides-tutorials/monochrome/

It is so easy to bypass the manual altogether and go straight for the one-to-one interaction and immediate feedback that the forum provides, but the work that has been done on the manual is truly excellent – kudos to @elstoc, @Matt_Maguire and @paperdigits (*) – and I have made a point to re-read all of it. An advantage over a forum discussion is that the explanations are brief and to the point, and if you are in a rush to understand how things can be done, this should be the first thing to check.

I know, it’s obvious!, but as I’ve been the first one in the past to skip the official docs and go ask for help I thought it was good to remind others about this.

In the manual is also discussed workflow and I have just discovered (but maybe it was there since v2.6 and I never noticed?) the option to set a monochrome tag on the “selected image(s)” Lighttable module. What is definitely new in 3.4 however is a feature to automatically set this tag on import – dt now recognizes images show with a black/white profile set in camera – which is great! (I used to have a bw tag that I manually added, I have now replaced this with the internal dartkable|mode|monochrome to account for all the old images already present in my database).

(*) I have just checked on github the first 3 contributors in terms of commits, but of course my thanks are extended to everybody else that has contributed, like @anon41087856 that has spent so much energy to explain things here on the forum and via youtube.

@aadm thank you for the link, the new manual is really great! :slight_smile:

What I’m really struggling with when doing digital monochrome is the simulation of filters, I have yet to find a good way to do this (e.g. a yellow, orange, red or green filter). My main issue is that I don’t see any way to specify a colour bandpass filter in DT. I guess this has to be done with a LUT?

Additionally, I have the nagging suspicion that to get good and consistent results one has to understand the camera, sensor and RGB filter spectral response quite well to really apply the LUTs accordingly. I guess one has to do something like @ggbutcher with his SSFS measurements

In my monochrome dabblings, I’ve been getting decent results with a gray tool that allows adjustment of the contribution of the three channels to the gray tone. rawproc’s gray tool looks like this:

rawproc-gray1

Those numbers are used to multiply each corresponding R, G, and B value to come up with the perceptually pleasing gray tone that pixel represents.

In my film shooting, I used to use red filters to turn blue skies black and make cumulus clouds punch out of the image. If you think of what a red filter is doing, the corresponding settings in the gray tool make sense:

rawproc-gray2

to make images look like this:

Red is easy; for other filter colors, look up their 0-255 RGB values in Google and divide each number by 256.

2 Likes

I believe the new color calibration module is the way.

The monochrome workflow & autodetection has been done as it is not only you loving b&w images :slight_smile: In 2.6 only true monochromes were detected, now it’s based on preview content, the sensor and whether you have been using the passthrough demosaicer.

You may not have noticed yet, also the presets have been modified, you can now define a preset only to be used on monochrome or color – whatever you want.

This together gets you pretty far right now …

2 Likes

Maybe I’m completely on the wrong path, but I always thought that this way of converting the image just mixes the contribution of the channels with the gray target independent of the actual hue, in essence this is just an inner product between the pixel value and the contribution vector?

What I was trying to achieve is to model the transmission behavior of a filter and this is of course wavelength dependent and quite non linear.

For example the (quite handy snap on) filters from adox come with the following wavelength response curves:

I’m pretty sure that all this is way overkill for most people who want to get good b/w images, I’m just curious if a full simulation is actually possible with the current modules

Thank you for pointing out that new section in the 3.4 manual.

The last section titled perceptual approach is a bit puzzling to me, but maybe I just don’t understand something. It suggests that the color balance module is the best one to use to convert to monochrome. In particular it says this at the end:

Other modules such as color zones can also be used to remove color saturation from the image, but these don’t offer any real advantage over the simplicity of the color balance module’s saturation sliders.

The color zones module allows one to have lots of easy control over how different colors are converted to shades of gray. I have found that starting with the preset black & white film gets you a monochrome image and then you can easily move points around on the lightness view of the color graph to adjust different colors. If you don’t like the starting point of the black & white film preset which already has nodes moved around you can double click anywhere on the color graph and all the nodes will go to the middle line, but still monochrome. You can then move nodes to adjust.

Playing with the color balance module I can move the output saturation slider to the far left to 0% to get a monochrome image. Then after doing that I have not found a way to target several colors to get different shades of gray like I can with color zones (or the Lightroom HSL/Color module where I can adjust 8 colors to get different shades of gray: red, orange, yellow, green, aqua, blue, purple, magenta). This is probably my ignorance about how to use this module though. I have read the manual and been playing with it, but I can’t find how to do what I want.

The manual is speaking strictly about converting the image to monochrome by de-saturating it, not about other adjustments you can make using other modules. Color balance is preferred because it works in linear RGB, while I believe color zones uses LAB.

1 Like

You can perhaps use more than one instance of color balance module and make the de-saturation on the uppermost one in the pixel pipeline. Then, you can play with the various color effects on the lower module(s).

Mica @paperdigits has already replied wrt to this.

What I wanted to add is that the manual also discuss the “physical approach” to do monochrome conversion:

[…] Different types of monochrome photographic film have different levels of sensitivity to various wavelengths of light, and this can be simulated by giving the three color channels different weightings when combining them together into a single monochrome channel. The color calibration module allows the three channels to be mixed into a grey channel by varying amounts, and it includes a number of presets that are designed to emulate the characteristics of some well-known types of photographic film.

So when reading the manual I knew I wanted to understand the new color calibration module to do that kind of channel-mixer approach that others (@soupy @ggbutcher and @s7habo) have discussed here. Also a few film simulation presets are already available in color calibration.

Okay, I see.

I have workflow: scene-referred set. As I was just now hovering over some of the modules that are included I see that color zones, local-contrast, defringe, vignetting, etc. have display-referred output, but other included modules have scene-referred output.

I suppose the control and flexibility of color zones for making monochrome is worth going out of scene-referred for me.

You can still use color-zones with a scene-referred workflow, as it comes quite late in the display-referred portion of the pipeline. The other option is to use color balance (possibly with parametric masking).

2 Likes