DXO 2 & Calibrated Sigma 18-35mm

[xpatUSA]

I don't think there is a direct relationship between CIELab and Color Temperature adjustment.

A further find confirms your thought, Dave. A CIELUV chromaticity diagram with both Color Temperature and CIELAB a* & b* lines (dashed) on it!:



It is quite clear that there is no direct linear relationship between Color Temperature and CIELAB a* - QED. So why Manfred wrote that there is remains moot ...

Can't see one for the constant-K tint lines either ...

[Manfred M]

So why Manfred wrote that there is remains moot ...

The difference between theory and practice, Ted.

I get slightly different results between when I do the adjustments with raw data and an image file. I don't know if the underlying algorithms are different or if rounding or other issues creep in.

Again, the statement I made was that white balance / colour temperature is "bake-in" during raw conversion.

[xpatUSA]

The difference between theory and practice, Ted.

If "theory" is published curves of temperature and tint, and "practice" is what happens when you waggle the sliders in Adobe's ACR, I know which I would trust. And, somehow my grays always look . . . . gray!

I could go into my editor and play with those same sliders - and probably will. I would expect the 'white point' in an image to follow the curve and the intersecting tint lines on a CIELUV diagram, not the axes in the CIELAB, would not you?

I get slightly different results between when I do the adjustments with raw data and an image file. I don't know if the underlying algorithms are different or if rounding or other issues creep in.

Again, the statement I made was that white balance / colour temperature is "baked-in" during raw conversion.

Yes Manfred, quite so, as indeed you wrote in your post #21.

However, what bothered me was in your post #34:

"We have three sliders that affect the colour temperature / white balance that are labelled exposure, temperature and tint. If you understand your colour models, these are nothing more than the L, A and B channels of the LAB colour space."

My post #41 shows clearly that there is no such relationship, so I still don't understand what you mean by it, sorry.

Maybe Adobe dumbed it down somewhere for the punters and it passed into color balance lore, somehow?

[Manfred M]

Yes Manfred, true, as you wrote in your post #21.

However, what bothered me was in your post #34:

"We have three sliders that affect the colour temperature / white balance that are labelled exposure, temperature and tint. If you understand your colour models, these are nothing more than the L, A and B channels of the LAB colour space."

My post #41 shows clearly that there is no such relationship, so I still don't understand what you mean by it, sorry.

Maybe Adobe dumbed it down somewhere for the punters and it passed into color balance lore, somehow?

The LAB colour space has three channels:

L - which is Lightness

A - Green - Magenta

B - Blue - Yellow

That suggests to me that this is the underlying colour model in play in the raw converter.

[xpatUSA]

The LAB colour space has three channels:

L - which is Lightness

A - Green - Magenta

B - Blue - Yellow

Agreed, although most of us call the channels L*, a* and b*.

That suggests to me that this is the underlying colour model in play in the raw converter.

I very much doubt that, sorry Manfred. What is meant by "in play"? Are there any credible references to back the suggestion?

Off to play with my temp and tint sliders ... talk to you later.

[dje]

Ted here is a plot of the Plankian Locus in the L*a*b* color space which I did in Excel. I hope I've got the sums right! I've never seen it plotted this way before. The plot is from 1500 to 10000 with 10000 at the bottom of the plot.




In my previous post, i gave some details of the dng/dcp white balance and color calibration process. It's been a while since I looked at the DNG spec but I just revisited it and need to mention a couple of points of clarification on what I said earlier in relation to dng processing in Adobe software:-

after white balance,the Forward matrices are used to convert the camera RGB values to the XYZ space (the Profile Connection Space). These values are then converted to linear RGB Prophoto space which is the working color space if you like.

The LUT I mentioned is based on HSV rather than L*a*b* as I previously indicated. If there is an LUT to be applied, the RGB values are converted to HSV, the LUT is applied and then the values are converted back to linear Prophoto RGB.

It's covered in Sect 6 of the DNG spec document 1.4.0.0.

If L*a*b* editing is set up in Photoshop, I assume the working color space is changed from linear Prophoto to L*a*b* for this function but I don't really know the details of this process.

Dave

[xpatUSA]

Ted here is a plot of the Plankian Locus in the L*a*b* color space which I did in Excel. I hope I've got the sums right! I've never seen it plotted this way before. The plot is from 1500 to 10000 with 10000 at the bottom of the plot.

Thanks for that, Dave. Looks similar enough to my quick-and-dirty effort in post #40 to suggest that we are in agreement.

In my previous post, i gave some details of the dng/dcp white balance and color calibration process. It's been a while since I looked at the DNG spec but I just revisited it and need to mention a couple of points of clarification on what I said earlier in relation to dng processing in Adobe software:-

after white balance,the Forward matrices are used to convert the camera RGB values to the XYZ space (the Profile Connection Space). These values are then converted to linear RGB Prophoto space which is the working color space if you like.

The LUT I mentioned is based on HSV rather than L*a*b* as I previously indicated. If there is an LUT to be applied, the RGB values are converted to HSV, the LUT is applied and then the values are converted back to linear Prophoto RGB.

It's covered in Sect 6 of the DNG spec document 1.4.0.0.

If L*a*b* editing is set up in Photoshop, I assume the working color space is changed from linear Prophoto to L*a*b* for this function but I don't really know the details of this process.

Dave

Thanks for the clarification re: DNG. I've got the spec on my HD somewhere but, like yourself, don't read it very often. My own preference for intermediate work is 16-bit TIFF, ProPhoto, 1.8 gamma.