It would be fantastic to have this type of feature in RT. I found it when learning resolve.
It takes a photo with a Mcbeth chart and automatically applies color matching/profiling. As an extra feature, it would be great to have the option to also have it apply luminosity matching so that the black and white patches are brought into their published rgb luminosity values.
Having color plus luminosity matching in this way would be a fantastic feature!
Any thoughts from the Devs?
Color correction is already done like this. Iâve contributed a day light shot for my cameras (and need to get back to the tungsten). If your camera doesnât have this in RT, you can certainly contribute the shots!
Well, yes and no⌠Yes, making DCPâs is done by photographing and matching color charts, and interpolation to other lighting conditions is pretty damn good for a dual illuminant case.
But no, because what Resolve does is doing this âon the flyâ. The way I see it, it matches the colors for a single illuminant case (i.e. wherever you shot the picture of your color chart) as part of itâs image pipeline. I can really see how skipping the process of DNG conversion and DCamProf for the DCP generation, could save some workflows quite a bit of time.
RawTherapee already supports DCP profiles. A DCP profile is what you get from profiling a camera using a color target, as in that video. However, unlike in the video, the DCP profile is created using external software:
Instructions for making your own DCP using the open-source program DCamProf:
http://rawpedia.rawtherapee.com/How_to_create_DCP_color_profiles
RawTherapee also ships high quality DCP profiles for many camera models:
RawTherapee/rtdata/dcpprofiles at dev ¡ Beep6581/RawTherapee ¡ GitHub
These profiles are dual-illuminant, created using daylight and incandescent tungsten light. The targets were photographed by various people and the DCP profiles were mostly created by myself. If youâre using a camera for which we already ship a DCP profile, there is not much to gain by photographing your own target under various light sources, unless those light sources vary significantly from daylight or tungsten.
Lastly, if RT lacks a profile for your camera and if you donât make your own DCP profiles then you can get some from Adobe DNG Converter:
http://rawpedia.rawtherapee.com/How_to_get_LCP_and_DCP_profiles
It would be nice if you could profile a color target directly in RawTherapee, but it would be a lot of work to port a program like DCamProf into RawTherapee for not much real benefit, and then you would have to come up with a graphical user interface, which could be made simple, which means most functionality would be automated and not up to the user to adjust, or it could be advanced which would open up all the possibilities for user-adjustment, but then youâre really just duplicating a program which already does just that - Lumariver Profile Designer.
As for matching white balance and luminance values, you can already do that in RT by picking the white balance off the target and by using the darkest and brightest squares to adjust the exposure (Iâd do that using a curve).
Correct me if iâm wrong, but i feel like doing color and luminance correction on a per image basis is very different than a dual illuminant profile for a sensor?
Seems as though it may be a lot of work. But for the sake of argumentâŚ
If i shoot a mcbeth chart bring it into RT and white balance (on one square) it what do I get? Certainly not the published RGB values for the chart, or anything even close really. It gives some correction to the color matrix, and I imagine that adds some benefit for general photography and color rendition. But for anything more precise it leaves all the work in your handsâŚ
For instance, just the advantage of having white balance at each luminosity patch (6 on McBeth, as opposed to one) would be great.
I really think that color matrix correction and luminosity adjustment on a per image basis, and profiling a camera sensor are not the same thing, I think they serve different functions. And that one is far more useful than the otherâŚ
Its funny, cinema guys demand thisâŚthey can profile their sensors but find this to be a necessary tool. In my mind profiling, the sensor offers far less benefit that this type of tool would.
White-balancing, setting post-demosaic white and black levels (not to be confused with raw white and black levels which a user should not adjust), and profiling a camera for a given light condition are all separate things, and not to be confused.
You do get (near) the reference values if you shoot under a light source handled by the profile. âNearâ because there is more to it than trying to perfectly match a few patch values - the matrix is tweaked, the LUT needs to be relaxed to keep graduated colors smooth, the LookTable is tweaked by the tone reproduction operator, there is gamut compression, etc. All this is done using the input color profile very early in the processing pipeline.
The forward matrix and LUT work on white-balanced data. Other than that, what exactly do you mean by âwhite-balancing gives some correction to the color matrixâ?
Iâm not sure what youâre implying. If youâre implying that having several neutral patches produces a âmore accurateâ white balance, that is not the case. The most neutral patch is to be used for white-balancing, and that is either patch 20/D2 or 22/D4 (sources donât agree, but it doesnât really matter as they are so similar).
You can use any patch to set the white balance in RawTherapee, including the non-neutral patches.
Try color-correcting a raw file without using an input profile and let me know how it goes.
As mentioned above, the DCP makes the colors match their reference values as closely as possible, and then it applies other stuff on top of that.
The other things you mentioned - setting post-demosaic white and black levels and white-balancing - are not the DCPâs job, but you can absolutely do those things in RawTherapee using a photo of a color target.
âWhite-balancing, setting post-demosaic white and black levels (not to be confused with raw white and black levels which a user should not adjust), and profiling a camera for a given light condition are all separate things, and not to be confused.â
Yes, but the tool in Resolve handles all of those complimentary processes in one go. Also, a utility named 3d lut which is very interesting but also expensive.
It minimizes deviation in the color matrix given the known RGB values in the chart. And thus white balances across the luminosity patches, and when luminosity matching is enabled it also solves for minimizing luminosity deviation.
For some workflows, these are complementary processes. Thus the development of tools that handle all of them in one go.
âYou do get (near) the reference values if you shoot under a light source handled by the profile. âNearâ because there is more to it than trying to perfectly match a few patch values - the matrix is tweaked, the LUT needs to be relaxed to keep graduated colors smooth, the LookTable is tweaked by the tone reproduction operator, there is gamut compression, etc. All this is done using the input color profile very early in the processing pipeline.â
Unfortunately, day to day we rarely shoot under these ideal conditions. Thatâs why per image matching would be so useful. Iâm sure thatâs part of the logic for developing these types of toolsâŚ
The forward matrix and LUT work on white-balanced data. Other than that, what exactly do you mean by âwhite-balancing gives some correction to the color matrixâ?
I think you are referring to this statement?? âIf i shoot a mcbeth chart bring it into RT and white balance (on one square) it what do I get? Certainly not the published RGB values for the chart, or anything even close really. It gives some correction to the color matrix, and I imagine that adds some benefit for general photography and color rendition.â
What Iâm saying here is that applying a DCP profile and then white balancing the image does not produce (unfortunately for the type of work I do) sufficient minimizing of deviation from the known RBG values. I still have 5-10 minutes of tweaking curves before I get the type of results that resolve or 3DLut produce with a single button.
"Iâm not sure what youâre implying. If youâre implying that having several neutral patches produces a âmore accurateâ white balance, that is not the case. The most neutral patch is to be used for white-balancing, and that is either patch 20/D2 or 22/D4 (sources donât agree, but it doesnât really matter as they are so similar).
You can use any patch to set the white balance in RawTherapee, including the non-neutral patches."
Yes, for people who shoot texture for CG and VFX having white balance across the luminosity range is very important. A very nice feature of the tools mentioned. If the goal is to minimize deviation from the known RGB values then that is a key part of the process. Itâs not good if my 160RGB patches line up, but the 52RGB patches are off by 20 steps. This is a fairly well-established workflow in texture for CG/VFX.
âTry color-correcting a raw file without using an input profile and let me know how it goes.â
This is just to be snarky or there is a real point here? It works just fine, colors are further from the ground truth of known RGB. More saturated in blues and reds.
"As mentioned above, the DCP makes the colors match their reference values as closely as possible, and then it applies other stuff on top of that.
The other things you mentioned - setting post-demosaic white and black levels and white-balancing - are not the DCPâs job, but you can absolutely do those things in RawTherapee using a photo of a color target."
Again I donât think the DCP/Profile is the issue here. I think itâs more helpful to think of this as a separate image processing tool rather than an extension of sensor profiling. I understand there is a relationship under the hood in terms of correcting the color matrix. But this is a per image based tool to correct and balance the color and luminosity on a per image basis given the known RGB.
These tools are already established and in use. No need to debate if they are valid. These workflows are in place across a number of image-based industries.
3d Lut
Resolve
Is it only me, who thinks, posting links to videos for a commercial software should not happen here?
Donât get me wrong, I like the dicsussion. Iâm just against posting links to videos of commercial software here.
Not against the companies and their communities, but these videos give the impression of marketing and produce placement.
I couldnât have expressed this better.
I donât think there is anything inherently wrong with sharing videos of commercial software, if the intent is to help show us how they work.
However, as Iâve said in other threads, trying to show the videos in order to get a feature added is not as compelling as the poster probably thinks they are. Not only do we have a tradition of going our own route, but also just relplicating features from other software (not only commercial software, but other free software as well) is just a boring task, in general. Further, if youâre trying to get a feature added and canât articulate how that feature would fit in with the application you would like to add it to but think that a video of another application says all that it needs to, youâd be sorely mistaken.
In this particular case, it seems like it be a lot of work to add this feature, but not a lot would be gained for most users. If this is indeed the case, developers may be swayed other ways, such as money, copious amounts of alcohol, or other forms of compensation. If that doesnât work and you really need the feature, the code is open and Iâm sure someone will take your money to implement the feature.
I agree! Maybe showing three video links in one post led me to my comment. I apologize!