The Myth of Universal Colour

February 6, 2012 • Tim Parkin

While we were working on the Big Camera comparison, one of the things that became quite clear was that the different sensor devices we looked at were producing images whose colour was quite different. More importantly, when we tried to fix the colour from one to look like another, it proved impossible.

This rung a few bells with me from a couple of years ago when I was looking at whether it was possible to simulate Fuji Velvia 50 by creating some form of Photoshop action or icc profile. It quickly became obvious that although we can approach some of the colour changes that Velvia introduces (for instance blue shadows, tendency to move colours toward their primaries), there were certain colour changes that were impossible to fix. Trying to change one part of the colour range would inevitably affect another part. Eventually I gave up with this, concluding there was either something magic going on or my Photoshop skills weren’t up to it.

You need to be logged in to see this part of the post
In Digital, Needs Paid Subscription, Science of Photography

14 Responses

  1. AlexeyD

    A great article (still requiring a lot of digesting for me to do ;) . One thing I would like to comment upon is applicability of film ICC profiles. Whilst there are some limitations to the usage of this approach in Photoshop, the most successful way of approaching this that I know about so far was implemented in Raw Photo Processor (RPP). They are build from real films and the whole process (with examples from various photographers) is described here http://translate.google.com/translate?hl=en&sl=ru&tl=en&u=http://raw-rpp.livejournal.com/35303.html (please note that this may be NSFW due to nature of some of the photos).

    On February 8, 2012 at 2:51 pm • Reply
  2. AlexeyD

    Also re: as the Nikon D3X and Sony A900 both have the same sensor…

    They are not really the same sensor – none of the Sony sensor used by Nikon is technically the same. The ADC and processing is Nikon’s, the filters and microlenses are Nikon’s. Effectively it is a Nikon sensor build on Sony’s technology.

    On February 8, 2012 at 3:24 pm • Reply
    • Tim Parkin

      Hi Alexy – I was referring to the sensor in terms of the ‘chip’ excluding the CFA/Anti-Alias and ADC, the raw silicon as it were. I agree that these details are different (as I understand has been confirmed elsewhere). These are such crucial aspects for the colour (as you mention in your next comment).

      On February 8, 2012 at 6:00 pm • Reply
  3. AlexeyD

    Having now absorbed all the article details, I am not sure if metamerism index alone is enough to define the camera/sensor colour fidelity (for the lack of the better term). Quite of few of the older generation cameras used a more dense CFAs with better colour separation in RGB channels and not tried to reproduce the peculiarities of CIE standard colorimetric observer (i.e. where red channel has a spike in blue area). This resulted in a better colour especially considering the Bayer demosaicing. In pursuit of the increasing sensor sensitivity, camera manufacturers tend to lower the CFA density so the filter are no longer R,G and B exactly and colour separation has weakened as a result of that. This is typically compensated by camera profiles to some degree but I don’t think they can compensate it completely. It is interesting to see this trend being somewhat reversed recently as camera manufacturers starting to regain that colour fidelity – Sony’s cameras and Olympus pens do have a very good colour reproduction.

    The reason I find the metameric index alone a bit confusing is that for example the Nikon D2x, Kodak SLR/n are two older examples of the cameras that produced very good colours out of the camera with later Nikons being not that good in that department (D3, D300s etc) yet their metameric indexes are roughly the same on DxO area (I have not found one for Kodak). It would be interesting to see how the newer cameras now being introduced will do in this area.

    On February 8, 2012 at 5:07 pm • Reply
    • Tim Parkin

      Quite agree (and I think I mention that in the article) although the metameric index is a good indication of how susceptible to colour problems the camera might be. For instance I don’t think the Canon 5D2 has great colour but it doesn’t do too badly. This may be because it only has an issue with a single colour pair in the ISO test but that colour pair is quite bad. Another camera might achieve a similar score by having all of the colour pairs having a slight problem.

      Also, if the colour pair is one that is predominant in your subject matter (such as the infra red metamerism in synthetic dyes or the chlorophyll metamerism) then just one bad colour can mar fashion or landscape performance.

      On February 8, 2012 at 6:04 pm • Reply
      • AlexeyD

        Thanks Tim, regarding Canon 5D and 5D2 – I heard that from a few people using those two cameras in succession. It would be very interesting to see what the upcoming Canon 5D3, Nikon D800 and Fuji X1 have to offer colour wise.

        On February 8, 2012 at 7:33 pm • Reply
  4. Robin Sinton

    Don’t forget that the colour that you see will be to a large extent determined by the manufacturers profiles in whatever raw conversion software that you use. If you need as accurate colour rendition as is possible you can make your own profiles using the scripts which can be found here.
    http://www.rags-int-inc.com/PhotoTechStuff/ColorCalibration/
    However, as many people have said, Velvia isn’r accurate but it doesn’t matter. It’s the picture that counts.

    On February 9, 2012 at 4:14 pm • Reply
    • Tim Parkin

      I agree that the picture is the thing that counts but there is nice inaccurate and nasty inaccurate and making pictures with nasty inaccurate isn’t pleasant – and whatever profile you use, you can’t fix metameric inaccuracy… So my personal recommendation is to look at the colour of pictures that you like and find out which cameras or film they came from – Personally, for DSLRs and high end compacts, I like the results from Sony and Nikon and Olympus isn’t too bad..

      On February 9, 2012 at 5:16 pm • Reply
      • Robin Sinton

        Good point Tim, but doesn’t it also depend on the colour rendition of the respective lenses? There are a lot of variables creeping in here. Don’t forget that metamerism also exists in film. I took a product shot many years ago of a green carpet that turned out brown as the colour neatly fitted in between the gaps in the dye spectra. In practical terms how much metamerism do you see in a landscape. I prepared a portfolio consisting of 10 prints of which three were from scanned film. I asked my professional lab to tell me which ones were which and they couldn’t, so I’m happy. How many people who look at photographs are worried about the exact accuracy of the colour rendering. (commercial photography excepted)? Back to, “Its the picture that counts”.

        On February 9, 2012 at 5:34 pm • Reply
        • Tim Parkin

          I think the contribution from metamerism from the glass on the lens is minimal although some lenses do impart an overall colour cast (canon yellow is a case in point).

          As for what metamerism we see in the landscape. Nearly all colours we see have metamers and so a camera that is cannot see metamers as the same will have problems. e.g. two red berries on different trees may be metamers, i.e. they look the same colour to the eye. but a digital sensor may see one as a cool red and another as a warm red).

          I would be wary of drawing conclusions about colour from a single sampling of ‘professional’ printers. In terms of ‘colour’ differences in digital and film, both myself and Dav Thomas correctly assigned Nikon/Canon to many different pictures in the Wildlife Photographer of the Year competition and another colleague was able to assign digital/film to a very high accuracy for a well known european photographer. These differences are not necessarily subtle “only visible side by side” effects.

          For instance, take a look at this page

          http://static.timparkin.co.uk/static/tmp/cameratest-2/800px.html

          and look at the differences between the different cameras (if you fancy it, try manipulating one of the images to look like another, e.g. try making the Phase P45+ look like the Sony).

          So yes, it’s the picture that counts but the colour is an integral part of that..

          On February 9, 2012 at 7:06 pm • Reply
  5. Robin Sinton

    I think that the point that I am trying to make is quite simple. All of the points that have been made are valid, but, does it matter? Unless you are holding a print in front of you to do a direct comparison with the scene, does the exact colour of a red berry matter (unless it’s so far out that it looks blue). The light will be different in any case. When we look at a photograph I would suggest that the accuracy of the colour rendition is not a high priority in deciding whether or not we like it. Apart from anything else we don;t know the exact colours of the scene and the eye-brain combination will interpret it differently depending on the viewing conditions. I think that as photographers we have to beware of falling into the trap of making the impossible assumption that we can quantify and reproduce the scene exactly. If we go down this route we run the danger of forgetting the aesthetics and getting bogged down in the technicalities. If you took half a dozen photographs of a scene taken with different cameras at the same time and shuffled them around. Could you honestly say, “That’s a Canon, that’s a Nikon etc”. Yes, you could say that one is different to that one, or, I like that one better, but at the end of the day it’s going to be a subjective decision. There will always be times when we don’t necessarily want to produce a technically correct colour for aesthetic effect. Pre dawn light is an example. Back to “It’s the picture that counts”.

    On February 9, 2012 at 9:42 pm • Reply
  6. Michael SA

    Thanks Tim – I think this is a very interesting and useful article (even if it is a bit geeky!!). I think that for landscape photography, a camera system that produces ‘pleasing’ colours (this is very, very subjective of course) is really important both for the end result and for the enjoyment of making images. An awareness of the fact that different digital cameras can produce slightly different colours, and the metamerism scores, may therefore be of help to some in evaluating which digital camera (if any) to purchase. Perhaps testing different cameras in the field makes sense prior to purchase. There may be a parallel here with choosing different films, however film is obviously more flexible because it can be changed, whereas a digital sensor is fixed – all the more reason to evaluate the ‘colour properties’ of a digital camera prior to purchase – of course only if this matters to the user.

    Michael

    On February 18, 2012 at 9:39 am • Reply
    • Tim Parkin

      Hi Michael – I’m glad you got past the geeky stuff :-) I really wanted to just get that point across more than anything, that different cameras produce fundamentally different colours and so it should be an important part of a purchasers criteria.

      On February 18, 2012 at 9:46 am • Reply
  7. Michael SA

    I agree, and I think that highlighting the metamerism score is useful in this context, especially as it hasn’t been widely covered before. For those who like rich/deep colours, I think DXO’s ‘colour depth’ score is also useful.

    On February 18, 2012 at 9:59 am • Reply

Notify me of followup comments via e-mail. You can also subscribe without commenting.