Color balance question


(Luis) #1

I have processed the same image with Canon DPP and RT.

The WB camera configuration was set to manual: 5400K. It was very warm to the scene and I had to change the temperature: 5000K in DPP and 4650K in Rawtherapee for get the same colors.

Why this big diference between DPP and RT?

Edited:
Now you can compare side by side both photos.

_MG_2047_DPP_RT


(Morgan Hardwood) #2

Using the same input profile in both?


(Alberto) #3

I hope someone more knowledgeable will chime in, but from what I understand (very little, I admit :slight_smile: there are many algorithms for “colour balancing” to a given illuminant, and that might produce slightly different results. could it just be that DPP and RT implement different algorithms?


(John) #4

The canon software will probably have used an ICC profile for the camera. RT doesn’t by default. You may be able to find it in the DPP software via a search. An alternative is to use an Adobe one or one from somewhere else. When these are used there will be some differences. Adobe usually offer several for each camera. In some ways they are mimicking part of the process the camera uses to generate jpg’s. Canon’s raw software will probably be doing more in that line as well.

If you were shooting Nikon you would very probably notice even more difference.

John


(Luis) #5

Hi Morgan,

the input camera profile in RT was autoselected.


(Shreedhar Inamdar) #6

On my monitor, the colours in these two photos do not look the same. The DPP photo is much warmer (as it should be). What is your basis (independent of how they look to one’s eye) to say that the colours are the same in both?


(Stefan) #7

IIRC, the WB values used in different Raw converters are not necessarily equivalent, so 5000K in RT is probably not the same as selecting 5000K in any other converter.


(Luis) #8

Well, the “same colors” is an imposible thing using two diferents raw developers, however I think they are prety close.

But, really do you think the second photo (DPP) is “much wamer”? I have read this post on my LG monitor, my HP Laptop, my Samsung tablet and my mobile phone, and I can’t find this warmer tonality… Maybe my eyes are different :roll_eyes:


(Alberto) #9

I agree with you. One of the two pictures has more contrast (don’t remember which), but the colors are reasonably close here as well (my monitor should be calibrated – meaning that I did calibrate it, but I might have done it wrong :slight_smile:


(Shreedhar Inamdar) #10

Have a look at this:

I could not get both of them fully but this does show you that the bottom pic much much warmer (I think)


(Luis) #11

shreedhar, please, compare the whole image, the face has more lighting than the legs…

_MG_2047_DPP_RT


(Shreedhar Inamdar) #12

Yep. I agree with your side by side shots that they almost look the same!

Now I can only see little difference in the color of the skin near the left forearm (dpp is redder) and in the background near left shoulder (RT is cooler).

But these differences are not as much as I thought before.


(John) #13

If I look at Aleph’s post the grey backgrounds are obviously different. Darker on RT. The lady is “hotter” on the RT shot, probably more natural and the hair tint is more obvious. Put all together and the processing is entirely different.

When I set a camera profile in RT I have to navigate to it. Not noticed any auto selection.

As I mentioned Canon’s software will very probably be using one of the curves that it uses to map raw into the camera jpg’s. Most cameras have several of them - settings such as landscape, portrait, high key and even diy on on some cameras. RT wont have done that and adjusting colour temperatures to fix the difference wont be the correct way to obtain the same results. It’s the equivalent of using one of RT’s curves panels.

Personally I always set camera white balance to auto and stick with it through out unless I can see good reasons to change it manually during processing.

According to DisplayCAL using either monki display or photo my max de on that shot is well under 1. If it had a bright white in it that could reach 1.5.

John


#14

The colours on Canon’s preview images depend on the “style” selected. These styles are available as icc profiles in DPP’s installation directory (Windows), with cryptic names, but correct descriptions.

Most of the time, my camera is set to standard style, the neutral style should be the most useful when looking at the camera’s RGB histogram. It is, of course, still far from RT’s neutral profile.

Switching styles in DPP (Mac/Win) is an obvious way for comparison. Which style was used for this screenshot (sorry, if I have missed this)?


(Luis) #15

The DPP style was “Standard”, contrast 1, Sharp 4, Saturation 0 (however on the second tab saturation was set at 196)

On RT I did a simple custom “S” curve with only two nodes in Standard mode.

@Jacal, can you share the Canon icc profiles?


#16

I’m not sure, it is free software, but profiles - like everything else - are licensed with Canon, all rights reserved.

Edit: The path on Windows is
C:\Program Files\Canon\Digital Photo Professional 4\DPP4Lib\icc


(Morgan Hardwood) #17
  1. @agriggio is right, though I would say the most likely cause for the discrepancy is that different ICC profiles are being used, using different colormatrices and created based on targets shot in daylight where the exact temperature of the light was not the same.
  2. Don’t focus on her skin or on the contrast in the photo. All you’re interested in is the hue of the background which you’re probably trying to get neutral.
  3. Don’t rely on your eyes, use numbers. Open both images in a color-aware editor (e.g. RawTherapee) and sample the colors of the same area of the background, the a* and b* values should match and have a difference of about zero.
  4. @Ajohn wrote, “The canon software will probably have used an ICC profile for the camera. RT doesn’t by default.” You don’t know what camera was used, so you don’t know that.
  5. @Ajohn wrote, “An alternative is to use an Adobe one or one from somewhere else.” That would not help.
  6. @Aleph a good starting point if you want to match white balance values in both programs is to use the same ICC profile in both programs. If there is still a discrepancy then that indicates that it lies elsewhere. Having said that, I don’t know with certainty whether raw white levels would have an affect - I would guess they do, and both programs are likely to be using different raw white levels.
  7. The most important point is that doing this is pointless. It is irrelevant if the temperature and tint values don’t match across programs as long as setting the WB based on a neutral patch does in fact lead that patch to become neutral (works fine in RT).

More on the subject:


(John) #18

@Morgan_Hardwood

I do shoot Canon and have used their software. As far as I’m aware they effectively use an icc files on all of their cameras that will shoot raw.

The Adobe profiles do differ but I’ve not found that to be a problem. They did tend to bring in more highlight than Canon did but that was some time ago and may have changed due to the number of stops wars in jpg’s.

Photivo comes with a lot of camera profiles. That can be a good source at times if it has one specifically for the camera that has been used.

As you mention though the solution is to use the same profile in both applications.

The other problem can be that the manufacturer’s software may play with colouration. That in some ways put me right off shooting Nikon.

John


(Luis) #19

Then, how can I get an icc profile for my Canon EOS 5D?

I went to Windows:

Archivos de programa/Canon/Digital Photo Professional/icc

but there you can find 37 icc files with cryptic names… :thinking:


(Morgan Hardwood) #20

@Aleph RawTherapee ships an ICC profile for the Canon EOS 5D, it also ships a DCP profile for it. Look in the iccprofiles\input and dccprofiles folders. If DPP ships 37 profiles for your camera model alone, these profiles likely include some effects baked into them. Unless you meant 37 profiles in total, not just for your camera.

Regardless, this exercise seems like a waste of time. The input profile’s job is to give accurate colors. It does. A different reading of WB temp and tint is not a problem.