Color Cast Problems with DxO Optics

[crisscross]

Re-visiting DxO after a couple of months, I notice it has a similar colour-cast (to magenta) as Lightroom on an untouched nef. However with DxO I can find no info on colour space/management at all. Anyone know anything about it? Academic since I would only use it for something I was going to finish in NX2 and regard its 'prepare' window only as a preview, but curious.

[Slipstick]

Hi

I have never seen a colour cast in any of my NEF images in DxO. If I were to correct a colour cast I would look under the 'Color' tab. There are numerous options there, but as I said, I never need to use them.

You could try the white and black points, colour temperature, HSL, or the custom tint. Under the same tab it could be that your color rendering needs some attention. Again, I cannot help, but the correction seems to be under this tab somewhere!

I find it unusual that you get the same cast with Lightroom. This would suggest the problem is in the camera settings. Do you get the same cast if you process directly as a jpg, instead of raw?

I hope you get it sorted.

Regards

Tony

[atvinnys]

just curious: have you checked your white balance? if that one is not nailed correctly, then everything will kinda shift with all the adjustment done on colors...just a thought...

[crisscross]

I carried out an extensive exercise on Lightroom as at that stage it was the only one of several comparison progs with this cast and I attributed it to ProPhoto colourspace. Maybe DxO uses that too as it also runs as a plug-in to LR & PS and may have some common architecture at some stage. It was certainly not cured by equating the colour temperature in LR.

Tony - you can only see it by opening the DxO (or LR) window and placing an NX2 (or GraphicConverter) window at same scale half over the first showing the same area. In the case of DxO it is only a PREVIEW, which can change quite a bit in the converted tif, haven't tried that and it is not a full test if one doesn't know what colour space DxO is running in so comparison prog can also be set to it.

Its not a problem as I would never use DxO as a beginning to end process and can't see anything in LR that cannot be done more thoroughly in combination of Nikon View or DxO followed by Capture NX2.

[atvinnys]

In the case of DxO it is only a PREVIEW,

the I icon actually shows you what's not applied...usually sharpening and some others which can only be seen at 75% and more...

ProPhoto colourspace. Maybe DxO uses that too

I don't think Dxo handles ProPhoto. I am pretty sure it's sRGB or AdobeRGB.

[crisscross]

When talking about 'White balance', I tend to interpret it as synonymous with 'colour temperature'. So I have been back over the image at Adobe RGB - sRGB again, groan post 9 this time in NX2, LR and DxO to see what happened using white balance as a reference. There are conveniently small strips of 'blown' white on the paintwork, which must be WHITE? (grey in Nikonspeak) (100 100 100). In NX2 they are, or maybe 99.47 ish, but you can marquee an area rather than taking a single pixel. In LR the effect of taking a sample is quite drastic and not for the good. In DxO there doesn't appear to be a control as such, but selecting the pipette, you automatically get a before/after split screen, now that is thoughtful, and it goes straight to a colour temp and tint setting; presumably WB corrected.

When previously editing the pic for use, I concluded the best colour temp was 5400. Not quite sure why one would want to 'tint' - in LR it appears to be a magenta-green scale in addition to the yellow-blue of colour temp. In NX2 I would do this using the colour balance tool which (a) has both those scales plus a cyan-red slider, brightness & contrast all in the one tool kit (b) can be applied to selected areas only.

However, setting the colour temp to 5400 and tint to zero in all 3 apps, LR retains the magenta cast (which it had set using WB tint at 10). DxO comes round to the expected colour matching NX2 (etc) -

So yes, DxO more probably is in Adobe RGB; no options, but fine by me. ProPhoto not terribly helpful.

I notice also that LR has exactly the same tools sets for a NEF/RAW image and a tif...so is it actually making any use of the manufacturer's RAW 'back pocket'?

[Colin Southern]

There are conveniently small strips of 'blown' white on the paintwork, which must be WHITE? (grey in Nikonspeak) (100 100 100). In NX2 they are, or maybe 99.47 ish, but you can marquee an area rather than taking a single pixel.

Hi Chris,

I'm jumping in here without reading too much of the previous posts, so it's possible that I've got the wrong end of the stick - so my apologies in advance if this is the case!

But ...

... You can't use blown areas of an image for white balancing. The idea isn't to base white balancing on something white; it's to base it on something that's SUPPOSED TO BE WHITE (or spectrally neutral to be more accurate), but isn't. If you click on something that's supposed to be white (or gray) and the program finds that for example the red component is a lot stronger then it assumes that this is because the image has a red cast - and will reduce the red component across the entire image accordingly so that all channels of your sample are equal, thus nulling the cast. Blown whites can't be used because there's no differentiating information there; even if the image has a strong cast, if any of the channels are blown then it's not possible to accurately determine the colour relationship.

For white balancing, a medium to light grey works best - some say that the closer the reference is to white the better, however I've found that in theory it makes white balancing more accurate (less rounding error), in practice programs like ACR seem to imtroduce a small bias when sampling bright areas.

[crisscross]

You are right Colin about what I said, but ...er...not what was in mind and didn't get into print

Thanks for warning not to use blown for WB.

In the image cited, the clean white paint goes seamlessly into blown and end result much the same going slightly off blown area, which is only detectable using software detector.

I was trying to make sure that DxO didn't use ProPhoto colourspace as appeared might be the case and am satisfied it most probably uses Adobe RGB.

What was in my mind also was some way of establishing whether the colour bias I am getting in LR originates in the software analysis of the colour of a particular pixel or from the pixel being displayed incorrectly on screen. I will pursue that when I have fathomed out how to place realistic patches of numerically and/or physically identical colours in each software.

[Colin Southern]

I was trying to make sure that DxO didn't use ProPhoto colourspace as appeared might be the case and am satisfied it most probably uses Adobe RGB.

Hi Chris,

I'm still a bit confused here ...

If if we're talking about input colourspaces then RAW files fed to DxO don't have any colourspace (internally it might use something like pro-photo with linear gamma like ACR, but that's invisible and of no consequence to the end-user).

If we're talking about output colourspaces then I image it could / should be whatever it's told to use - easy to just open up in any good editing program and see what it's set to.

[crisscross]

If if we're talking about input colourspaces then RAW files fed to DxO don't have any colourspace (internally it might use something like pro-photo with linear gamma like ACR, but that's invisible and of no consequence to the end-user).

Not sure this is quite the case with NEF; here is the end of my dialogue with Nikon from when I inadvertently changed the camera to shooting mode Ia=sRGB:

"The image will show up as sRGB in any Exif Data as this was the mode the image was captured in at the time of shooting. You can change the colour space in Capture NX and other software applications but the Exif Data will retain the original information for the captured images. The bottom left hand corner of the image will show you the colour space you are working in at the time of any editing in Capture NX. "

which confirms what I found and didn't much like in real life. BTW did you see my question on whether this file tag can be changed by moving the underscore in the file name back to its usual place?

If we're talking about output colourspaces then I image it could / should be whatever it's told to use - easy to just open up in any good editing program and see what it's set to.

Agreed; however its nice if prog used for editing allows the colourspace to be set for during editing and display on screen and it makes quite a difference. See Nikon quote for NX2. The same applies in Canon DPP, but I can no longer use that as the batchworker/converter module has never (yet) been made to work consistently in mac OS10.5 (mine being one it won't work on).

NB it matters to me as my preferred workflow NEVER converts the image; the web sized jpg is just an extract. When importing a tif from DxO or someone elses image into NX2, I still save it as a nef after editing. This is because the nef format is not only the Nikon form of RAW, but NX2 equivalent of PSD, ie conserving all the edit steps for re-edit later.

[Daniel Salazar]

Hi Chris, what I could realize is that DxO uses either sRGB or Adobe RGB, you can choose, however it also gives you the chance to choose another profile you have.

DxO also gives you the chance, when correcting colour, to choose or change the colour depending on the camera, I suppose that it's rendering the colours of the picture.

What I do not understand is if your questions/problems with the colour workspace are more philosophical in order to learn or if you're having troubles.

Chris, important is to remember that Macs are working with a colour space set by the ColorSync Utility and not exactly set by every application as with the PCs, therefore you should review some apple troubleshoots or reading something like this:

http://images.apple.com/pro/pdf/Color_Mgmt_inTiger.pdf

or contacting the colorsync mailing list

http://lists.apple.com/mailman/listinfo/colorsync-users

[Colin Southern]

Not sure this is quite the case with NEF; here is the end of my dialogue with Nikon from when I inadvertently changed the camera to shooting mode Ia=sRGB:

Hi Chris,

Not sure if we're talking about the same thing here - sRGB and Adobe RGB are only applicable to RGB format files (ie "JPEGS" (if camera generated, exluding cameras that write TIFFs); a RAW capture contains R, G, and B channel information, but it's not an RGB format file, and by definition, can't contain a colourspace. It could pass a metadata tag saying "sRGB" etc, but I don't know what the point would be - if it does I would think it would be more of a bug than anything (for a RAW file).

BTW did you see my question on whether this file tag can be changed by moving the underscore in the file name back to its usual place?

No - I missed it. I wouldn't think that that would work. The Metadata tag is for the software's benefit - the name change is just for us humans

For what it's worth, Canon do the same thing - so I set mine to sRGB to get the nicer looking filenames (before changing them when I convert to DNG) - but it doesn't make any difference to the RAW capture.

[crisscross]

Hi Chris, what I could realize is that DxO uses either sRGB or Adobe RGB, you can choose, however it also gives you the chance to choose another profile you have.

Thanks again Daniel. I know DxO allows choice of output colourspace, but have not found anywhere to choose working colourspace.

Becoming philosophical, except that I usually temper my philosophy with common sense. The difficult bit (as in the rest of life) is squaring philosophy (which underpins language) with own experience

[Colin Southern]

I know DxO allows choice of output colourspace, but have not found anywhere to choose working colourspace.

Hi Chris,

Working colourspace doesn't matter; it never makes it outside of the PC into the real world. For what it's worth, ACR uses a linear-gamma pro-photo space whereas Photoshop (internally) uses CIEXYZ or CIELAB (can't remember which without looking it up)

[crisscross]

Hi Chris,

Working colourspace doesn't matter; it never makes it outside of the PC into the real world.

I think it does matter - I don't see how you can do editing by eye if you are looking at less of the information than is available or a disneyfied version (sRGB) of it. It would also matter if it makes me happier without any technical justification whatsoever, as I aim to enjoy editing

[Colin Southern]

I think it does matter - I don't see how you can do editing by eye if you are looking at less of the information than is available or a disneyfied version (sRGB) of it.

Hi Chris,

What colourspace programs like DxO, ACR, or PS use to represent the image internally bears no corelation to what you see on the screen; What you see is simply whats rendered for the display device. Or put another way, "the map is not the terrain" ... it may be a good representation of the terrain, but one that none-the-less has has limitations that the actual terrain doesn't have.

A prerequisite of conversion and adjustment programs like DxO and ACR is not to lose image quality due to limitations within the programs; if they used something like sRGB or Adobe RGB then this could easily happen (in fact it WOULD happen) - that's why they represent images in enormous spaces like LAB that are capable of describing colours that can't even be physically reproduced, so the limitation isn't in the colourspace - it's in the smaller spaces (input or output) that colours are lost.

Sorry - difficult for me to articulate well - am I making any sense?

[Bill44]

Hi Chris,

What colourspace programs like DxO, ACR, or PS use to represent the image internally bears no corelation to what you see on the screen; What you see is simply whats rendered for the display device. Or put another way, "the map is not the terrain" ... it may be a good representation of the terrain, but one that none-the-less has has limitations that the actual terrain doesn't have.

A prerequisite of conversion and adjustment programs like DxO and ACR is not to lose image quality due to limitations within the programs; if they used something like sRGB or Adobe RGB then this could easily happen (in fact it WOULD happen) - that's why they represent images in enormous spaces like LAB that are capable of describing colours that can't even be physically reproduced, so the limitation isn't in the colourspace - it's in the smaller spaces (input or output) that colours are lost.

Sorry - difficult for me to articulate well - am I making any sense?

You are to me, it's similar to putting highly refined racing fuel into a VW beetle, the beetle cant extract the performance out of it. For beetle read monitor/printer.

[Colin Southern]

You are to me, it's similar to putting highly refined racing fuel into a VW beetle, the beetle cant extract the performance out of it. For beetle read monitor/printer.

Hi Bill,

I've been trying to think of a good analogy, but am failing miserably,

How about this ...

If you're going to tinker with the planets in our galaxy, you can't do it if you have to stuff them into something the size of only our solar system to work on them -- you need to give them the space they need - so rather than stick them in something that's too small (resulting in some of them getting cut off), you stick them in something bigger - much much bigger, like a universe so that the "size of the container" doesn't artifically limit anything. If you're going to be performing surgery best to do it in a room the size of a football stadium than one the size of a phone booth!

I think this profile / space thing is something that trips up a lot of people - you see something on the screen and think that that's what you're really working with, but your not - it's only a "translation" or an analog of the real thing - and you only get to see a small part of the overall capacity of the beast. LAB and sRGB colourspaces are a good example - LAB colour can describe colours that can't be physically reproduced (eg yellow with 100% luminocity), but program like Photoshop have to try and convert all of those into "equivalents" in far smaller spaces like sRGB and Adobe RGB, which just can't be done - so it does the best job it can - and we end up believing that the conversion is 100% faithful whereas in reality it might be 100% faithful to a colour that's in gamut for both colourspaces - or it might be a million miles out ... it just normally doesn't tell you that it's million miles out - so we're left beliving that what we're seeing is really the way it is. "Luckily" weird and wonderful colours in LAB colour are so far out of gamut for all of our printers and displays that they tend to get clipped in a similar fashion, so reds might go more towards orange and blues more towards cyans rather than suddenly popping up with a new colour that the world has never seen before.

[Bill44]

Personally speaking I don't know why so many people get their knickers in a knot about it. Back in the film days the average shooter bought Kodachrome or Fujifilm or Agfa or the equivalent in transparency film. Usually only the Pro's and the very dedicated bothered to buy Pro grade film. Unless they were one of the lucky few who bothered to do their own lab work, they happily sent the film off to the lab and accepted whatever came back. Let's face it, some of the colours induced by those old films would not be acceptable now.

Now, in this digital age, people are getting very concerned about colour space. Lets face it, and I'm being brutally honest here, photographic colours are like fine wine and how many of us could really tell the difference? If your honest you will admit that very few of us could detect the fine shades of colour difference that may be available in the higher end spaces.

A bit of further information. Not only, as Colin has said, are some of these colours not physically reproducible, but even if they were it is beyond the capability of the human eye to see them.

[crisscross]

Consider DxO modules that allow you to simulate pics taken with one camera as pics taken by another or simulating the old colour slide favourite films. There would hardly be any point in doing it if it didn't show.