Noob question here but what is the benefit if moving from the ‘Local Contrast’ display referred module to the ‘Contrast and Texture’ scene referred module.
I really liked the simplicity of the former and have struggled to achieve the same results with the latter (my results are more ‘wishy washy’ with the latter).
I’m keen to master the latter if being scene referred is a significant difference.
Thanks.
Hi Dave,
The first paragraph of the manual explains it well:
TL;DR: scene referred modules have more latitude in shifting pixel values before they start showing artifacts, because they work before the tone-mapper (e.g., sigmoid or. AgX). That said, there is no reason not to use the display referred modules if they work well for you and let you achieve the desired result.
Note that that display-referred modules can coexist happily with the scene-referred workflow and modules: they go after the tone mapper and work in the display-referred part of the pipeline.
Not a noob question. I personally use both the new module and the old local contrast module. I find them different in the look they achieve. I will say that the new contrast and texture module has fantastic sliders for shadows and highlights adjustments. I use it to brighten or darken shadows and to restore highlight details. That is not to say that I don’t then also use the shadow and highlights module which is display referred to further help out.
I am not a darktable user who feels the need to avoid display referred modules if they achieve the look I want. Just being a scene referred module doesn’t automatically guarantee a superior look in my view. Others will disagree with that I respect their alternative viewpoints as DT is all about choice.
For me the main point is (in my view) with display referred a pure black is defined as 0% and a pure white at 100% so any adjustments must respect those boundaries. With scene referred what looks pure white or black on screen may conceal hidden details that can be recovered before the tone mapper defines blacks as 0% and white as 100%. Most of the time our adjustments are working with in between values so display referred is not the evil monster some users will say it is. But I definitely want to recover highlight details before the tone mapper.
Good luck with DT.
I’m glad you are checking out the new module, and definitely not a noob question! The two key areas they differ in is with how features are extracted and how the modifications are applied.
-
Feature extraction in the old local contrast is either using a Laplacian pyramid or bilateral blur. The new module uses a pyramid of exposure invariant guided filters. These behave differently and I will leave it to you which you visually prefer over the other. The bilateral filter is known to have small artifacts at strong edges the other two should be safer.
-
Method of modification, this aspect will have more impact. The old display referred module works in LAB and adjusts the L component for local contrast while keeping AB constant. Working on the already tone mapped image means that highlights and bright parts of the image already are compressed and might not have much details to edit to begin with.
The new method works with the exposure of the image and changes the local exposure. The biggest advantage with this approach is that highlights and bright areas have their original texture and details which can be modified and altered before the tone mapping which is a huge advantage when for example working with clouds.
A bonus with working with the new module is that it also includes the low pass dimension of contrast editing as quite competent highlights and shadow sliders. You will have negligible differences in your mid-tones as those are usually quite untouched by the tone mapper, you might even be better of with the display referred local contrast module if you want to specifically target mid-tones and keep darks and brights untouched.
Hope that helps!
Remember that the module is still in beta even though it is in master and a at least one of the improvements I’m working on will sightly alter the result.
Thanks Jakob, and other responders. The merits of the new module seem clear. I’ll continue to experiment until my knowledge is no longer the weak link ![]()
I suspect that is why I see the two modules as having different looks and don’t try to see one as the replacement for the other. An excellent new addition to the Darktable arsenal of tools.