Q: Same in cam AWB settings look different?

Bas_Hamstra · October 19, 2022, 7:25pm

Hello friends, I am trying to get the hang of RT and I get more and more comfortable, and the results are good. What I do not understand is: when I use in-camera AWB and compare RT with DPP (Canon 6D this time) the colors are different: RT is slightly warmer and “redder”. I analysed this further and I am pretty sure it is a WB difference. But how can that be when both use the same in camera AWB?

Next I dialed into both RT and DPP a fixed color temp of 5300. Again results are different: RT is warmer and “redder”. And yes, I CAN equalize the difference, by warming up the WB in DPP and increase a bit of tint (increase magenta).

My question is: how can there be a WB difference when I dial in the exact same color temp? I just want to have “control” and therefore understand this difference. It’s not like one of the 2 is wrong, in fact I often like RT’s defaults a bit more.

Thanks for any insights…

Bas Hamstra

priort · October 19, 2022, 8:16pm

I suspect the color profiles are not the same and you would have to be sure you could disable all processing to compare between the two I would think…

Edit
This came from a poster on the Canon forum and basically sums it up…

Based on my experience with other (than DPP4) post-processing software there are 3 different kinds of profiles the pertain to raw processing. They are referred to as Input Profile, Working Profile, and Output Profile.

Output Profile is the ICC Profile that gets embedded into the metadata (if desired) for the resulting image file.

Working Profile pertains to the color space used during editing.

As I understand it Input Profile pertains to the camera used to shoot the picture, which I assume to relate to the sensor since NOT much else matters in the case of raw files.

DPP4 allows some choice for Working Profile. Those choices appear to be sRGB (default I think), Adobe RGB, Wide Gamut RGB, Apple RGB, and Color Match RGB. I don’t see any way to specify or choose an Output Profile. What I get in the resulting image files is an ICC profile named “sRGB v1.31 (Canon)”. However, there is NO mention anywhere about an Input Profile.

In the case of Rawtherapee (example of other post-processing software) it is possible to select DNG Color Profiles (i.e., .dcp files) which are produced and distributed by Adobe. It is believed that these files serve the same purpose for the various Adobe products that support post-processing of raw files. Apparently, these files do NOT conform to the ICC standards for color profiles but, I believe, they contain the equivalent information plus some. These DCP Profiles are camera specific and there may be several for each camera that appear to relate to the various picture styles that the camera offers and is presumed to pertain to what gets used by the camera to produce camera developed jpg files in those picture styles.

Insofar as these DCP Profiles are produced by Adobe rather than Canon it certainly is fair to question how they go about creating such files for all of the vendors whose cameras are supported by their software. One would be inclined to think that all of the various camera manufacturers assist with this effort. However, I can provide no evidence to that effect.

In summary, it does seem to me that Canon does NOT need to produce individual profile files that conform to the ICC standard (i.e., ICC Profiles) in order provide the equivalent information to its’ own software. However, it would seem that Canon could/should be willing to inform its’ customers about how this is done. That might be all I’m asking for!

priort · October 19, 2022, 9:39pm

Just as an example…

Default art settings… uses Automatch tone curve… RT is similar…

Load neutral without this and you get this…

Initial image uses automatch color profile as well but if you turn that off

If you switch to dcp as the color profile for color management then you get multiple canon profiles to choose from

There are 3 components that can impact the look within each of these profiles and you could also disable the adobe tone curve and apply again the RT one but using the various curve models which all impact the colors quite significantly… so the point is the color that you get in one processing software will not match up unless you can compare apples to apples which is hard esp with the proprietary ones that aren’t always so transparent about what they are doing…

EDIT:

My point is you can see all these in ART/RT but not with canon so the result will be unique and not necessarily a WB interpretation issue…

So I guess what happens if you disable the tone curve in RT and use the DCP for color and play with the 3 parameters available also matching the profile if you use one in camera ie landscape or vivid or what ever… then does RT look more like the Canon software??

Thanatomanic · October 20, 2022, 7:26am

Hi @Bas_Hamstra, there are multiple steps to convert the raw photon counts of your camera’s sensor to an RGB value on your screen. Two main steps are:

White balancing of the separate R, G and B pixels of the sensor
Matrix multiplication of the RGB-triplet with some calibration matrix.

The first step is connected to the “White Balance” module in RT. Choosing “Camera” should assume the same settings as you entered in your camera and should therefore give (nearly) identical colors regardless of the software that processes your image. However, the second step can vary a lot. Canon’s software might use a completely different calibration matrix than RT does. This can subtly change colors, e.g. blues becoming greener, reds becoming yellower. This is probably the effect you observe.

I do not know Canon’s own software and whether you have the option to either circumvent or adapt the calibration matrix. Even then, Canon might add some ‘magic sauce’ in the mix, so that you will never be able to get an exact replica in RT.

the temperature and tint that you see in RT bear no

jorismak · October 20, 2022, 1:23pm

White balance is doing by multiplying the values in the R, G and B channels.

So what’s really happening, is that there are 3 values of ‘multipliers’.

But you are entering a temperature and tint. So, there is a conversion done somewhere to go from temperature and tint, to multipliers.

From what I’ve seen, every software uses a different formula for that or a different algorithm behind it… or even something completely more different (like Darktable’s color-calibration module).

I’ve never, never seen the ‘temperature’ value matching up between two different pieces of software.

And that’s even before we’re talking about the camera matrix / camera profile in use as @priort said, which can affect the ‘tone’ of the image you see at a certain white balance setting.

You have a point in that ‘as shot’ should result in something very close, but even then there are differences in the profiles and the colour-model used and other algorithms at play. Could maybe even be a noise difference that gives the appearance of ‘something warmer’ for example.

Bas_Hamstra · October 20, 2022, 1:43pm

Thanks, it is at least good to know that this is not unexpected behaviour (same temp value not matching up between different converters). I have accepted also that colors never will match up totally, when different color profiles are used (Canon vs ACR).

I can nail down a very close look between the 2 with color temp = 5300 but somehow the tint in RT set to 1.1 in stead of 1.0. If I remember correctly for my Oly I had to use 1.150. It seems as long as I dial in these slight adjustments everything is fine.

Thanks for taking the time to answer,

Bas