Newbie: how should I approach this colormatching ..?

[ChristopherPerez]

Got it.

I now remember looking at the DNG Converter list of .dcp files and thinking there was a lot missing, compared to what I ended up getting out of the RAW Converter. That’s something for me to go back and confirm.

[ChristopherPerez]

I was able to confirm that the same Camera Profile directories are in the DNG Converter and RAW Converter. Using Wine I downloaded and installed the DNG Converter and found the Camera Profiles in ~/.wine/drive_c/ProgramData/Adobe/CameraRaw/CameraProfiles

I see that there are Adobe standard .dcp files in one directory and camera specific directories under “…/Camera”, where Adobe attempts to match the specific “look” of each camera model supported by the RentWare.

Point RawTherapee to these directories (or move them somewhere more convenient) and you’re good to go.

[ggbutcher (Glenn Butcher)]

Downloaded your raw and sent it through rawproc, my hack raw processor that lets me specifically control every step in raw conversion. The only way I could get the hue of the camera holder’s jacket was to turn off the color conversion from camera space to display space (I don’t use an intervening working profile). So, methinks RawTherapee isn’t doing any color conversion, working or display, and you’re looking at the unconverted camera data.

I’ll let someone with more RT experience sort that out…

[PD1 (PD)]

I may have got a better match with darktable

Journaliststuderende .ARW.xmp (11.7 KB)

[Acebone (Esben Maaløe)]

@PD1 - Yeah, it much better in your darktable version. To my eye the RT version is too yellow. But RT can do it, because @Alpacalypse 's RT version is very good too - in fact I really think he nailed it!

@ChristopherPerez - I copied all the profiles to another directory, and pointed RT to that. Did you have more profiles in the RAW directories than in the converter’s, or is it the same?

Can’t wait to check out peoples pp3s and learning to do it myself

[ChristopherPerez]

I had the same number of directories and files in the RAW and DNG Converters.

I’m assuming they’re exactly the same, though I’ve not verified the details of each .dcp. I may spot-check that tomorrow.

[xpatUSA (Ted Cousins)]

@Acebone

I’d still like to know how I should approach it without a profile though, in order to learn about working with colors directly, so if anyone has tips on that, the question is still open!

I’m thinking that HALDClut tech. ( a special image form of color look-up tables) might do a perfect job … if only I knew how to produce an HALDClut from an existing image.

Maybe: How to create HaldCLUTs from in-camera processing styles

Hmmm … there is a difference between the images - most likely lens correction in the one but not the other.

P.S. procedure in the link didn’t work for the posted images, probably because of that difference.

P.P.S. So I took the jpeg and edited it to look like the raw conv. and tried again, but no:

There must be another way - the joy being that the same HaldCLUT could used on all other raw conversions in that series of shots, I imagine.

[xpatUSA (Ted Cousins)]

@xpatUSA

There must be another way

Got it! In the GIMP plug-in G’MIC under ‘Color’ there is >Transfer Colors. I selected the ‘PAC’ option, et voila:

Darker but the colors match.

It came with a bonus which RT can use:

[Acebone (Esben Maaløe)]

That bonus … is it a LUT or what is it?

EDIT: oh, that was a stupid question - still, that is what being a newbie is all about

Man, I am learning from this thread <3

[xpatUSA (Ted Cousins)]

@Acebone

That bonus … is it a LUT or what is it?

EDIT: oh, that was a stupid question - still, that is what being a newbie is all about

Definitely not stupid!

See: Film Simulation - RawPedia

While you were typing, I put it into my RawTherapee HaldCLUT folder, opened your raw, went to ‘Film Simulation’, applied the HaldCLUT, then applied Auto in the left Tab, et voila:

Not perfect but I did everything in a rush - for example the HaldCLUT PNG can be edited to increase global saturation, brightness and stuff.

[Acebone (Esben Maaløe)]

Here’s my own attempt. Just colorwork, no profiles or luts of any kind. Well, that’s not entirely true, I used the lens profile for my lens, but that doesn’t alter any colors.

This thread has been an awesome inspiration, and I learned a lot, including valuable info on profiles and luts – nothing wrong with using those, I just wanted to learn about the coloring part of the workflow as well.

Journaliststuderende interviewer Bitten foran Frederiksberg Byret.ARW.pp3 (13.4 KB)

They’re far from identical, but I focused on the color of the jacket, and I got pretty close, even though I didn’t nail it a 100%

The “workhorse” in this solution is the channel mixer. It can actually target specific colors, if you remember to subtract other colors when you boost one. From a video I saw on youtube, I learned that the trick is: If you alter one color in one channel, you have to subtract/add the same amount to the other colors in that channel. However, I ended up with this strange result.

As you can see, the blue is in the blue channel (bottom) is boosted from 100 to 108, which is counteracted by subtracting 4 from the red and the green channel respectively. Net change: 0

But the red is pretty strange: I boost red by 200, and subtract 100 and 3 from the others - giving a net change of 97 … It works, but I am not sure why

[priort (Todd Prior)]

If you play with the adobe DCP files that might also be an option…If you can get the Adobe profile editor to run… it makes it very easy to make color specific adjustments to the profile…

Creating Camera Profiles with the Adobe DNG Profile Editor

[ChristopherPerez]

Looking at the RentWare Camera Profiles for the ZV10, it seems that “Standard” comes close to matching the jpg. There are differences, however. If you bump the exposure slightly and nudge the tints I think you’ll be able to get a very close match.

[xpatUSA (Ted Cousins)]

@Acebone

As you can see, the blue is in the blue channel (bottom) is boosted from 100 to 108, which is counteracted by subtracting 4 from the red and the green channel respectively. Net change: 0

But the red is pretty strange: I boost red by 200, and subtract 100 and 3 from the others - giving a net change of 97 … It works, but I am not sure why

The three-channel color mixer is mathematically the same as a 3x3 matrix. Matrix algebra is no fun at all so I talked to ChatGPT about it and ended up talking about “row normalization”. It seems that, when messing with the sliders, it is beneficial to keep the sum of each channel’s outputs (each “row”) constant - but sum of the squares also gets a mention:

@Me

Is it possible that, in a three-channel digital color-mixer, row-normalization is beneficial?

@ChatGPT

Yes, row-normalization can be beneficial in a three-channel digital color mixer, which typically deals with RGB (Red, Green, Blue) color values. In digital color mixing and image processing, normalizing the rows of a matrix that represents RGB values can have several advantages, depending on the context and the goal of the normalization:

1. Color Normalization: Normalizing RGB values can be used to adjust the intensity and balance of colors. For instance, by normalizing the rows of a matrix where each row represents the RGB values of a pixel or a set of pixels, you ensure that the color’s intensity is scaled relative to its maximum possible value. This can be particularly useful in situations where color data varies significantly in intensity and you want to achieve a uniform intensity level across an image or set of images.
2. Improving Visual Consistency: Row-normalization can help maintain consistency in how colors are displayed across different devices or media by normalizing the distribution of color intensities. This can make images more adaptable to different lighting conditions or display settings.
3. Preventing Channel Overflow: By normalizing RGB values, you prevent any single color channel from dominating or exceeding the permissible range (typically 0 to 255 in 8-bit color channels). This can help in avoiding color distortion due to channel overflow.
4. Enhancing Contrast: In some cases, normalizing color values can enhance contrast by maximizing the use of the available color range, which can make images appear more vivid and details more distinct.

The specific method of normalization can vary depending on the desired outcome. For example, normalizing such that the sum of the squares of the RGB values in each row equals 1 (L2 normalization) maintains the direction of the RGB vector but normalizes its magnitude, which can be useful for maintaining color ratios while adjusting intensity. Alternatively, ensuring that the sum of RGB values in each row equals a constant (like 255 for full intensity) can help in maintaining overall luminance while adjusting the relative contributions of each color channel.

[xpatUSA (Ted Cousins)]

@xpatUSA

@Acebone

As you can see, the blue is in the blue channel (bottom) is boosted from 100 to 108, which is counteracted by subtracting 4 from the red and the green channel respectively. Net change: 0

But the red is pretty strange: I boost red by 200, and subtract 100 and 3 from the others - giving a net change of 97 … It works, but I am not sure why

The three-channel color mixer is mathematically the same as a 3x3 matrix.

See here for a good reference:

explanation of channel mixer working

A year ago I started using Darktable and, as many, digested all YouTube tutorials. One module stayed an enigma, the channel mixer. The DT manual on it is short and lacks background info, the tutorials vary in extent and approach: some intuitive, others mathematical. I frequently read blogs with statements as ‘I still have problems, don’t fully understand’. Some tutorials needed a second one as the first wasn’t sufficiently clear. Apparently a module hard to explain and grasp. So I scraped all p…

[xpatUSA (Ted Cousins)]

@xpatUSA

See here for a good reference:

A year ago I started using Darktable and, as m

After reading the link I decided that:

a) keeping the output totals the same for each output channel keeps the channel balance the same.

b) changing the ratios of a channel changes the color but the hue stays about the same if the channel total is not changes. So in RT, I started with the ARW’s red(out) = 100, 0, 0 and changed that to 220, -60, -60 but left the other two channels alone. Voila:

oops - big file but I hope you get the idea.

[priort (Todd Prior)]

You might like this little summary

The Channel Mixer.pdf (68.9 KB)

[xpatUSA (Ted Cousins)]

Not bad, thanks.

Hope it helps the OP …

[Thomas_Do (Thomas)]

My try in darktable. Left original JPEG, right darktable.

Journaliststuderende interviewer Bitten foran Frederiksberg Byret.ARW.xmp (9.5 KB)

[xpatUSA (Ted Cousins)]

@xpatUSA

b) changing the ratios of a channel changes the color but the hue stays about the same if the channel total is not changed.

Um, not true: was playing today with keeping the channel totals to 100 and the color-checker’s red Hue did change, in this case from 355 deg to 344 deg.

In other words, this:

went to this:

Pardon the misleading prior statement quoted above.

Good old human color vision: to me, although the patch looks “redder” above it is in fact a little bit “bluer” …