Getting the color right (part 2)

Getting the color right – additional steps after calibrating your monitor (part 2)

This is part 2 of a two part article. In part 1 the focus was on camera and printer calibration, this part deals with different types of light and setting the white balance in various situations. (NOTE: to view without the need to scroll horizontally, click on the button at the top of the page to the right of the RSS button.)

Contents of part 2

• 5 White balance and color constancy
  
  • 5.1 How “right” should the WB be set?
    
  • 5.2 Dealing with difficult situations
  
  
• 6 Color temperature, light spectra and Color Rendering Index
  
  • 6.1 Various light sources and their color temperature
  
  
• 7 Some examples of difficult lighting
  
• 8 Putting it all together
  
• 9 Some more questions addressed and (partly) answered
  
  • 9.1 Do I really have to pay attention to all of this?
    
  • 9.2 Does it make a difference whether I shoot JPG or RAW?
    
  • 9.3 Can all screens be adjusted?
    
  • 9.4 What is the difference between a camera-specific profile and a Lightroom preset?
  
  
• 10 Final remarks

5 White balance and color constancy

No doubt you are familiar with how to change the white balance. It can be done in-camera by choosing one of the presets (daylight, tungsten, cloudy etc) or you can set a custom white balance with the help of a gray card, a white card, an ordinary piece of white paper or some more fancy equipment. What you actually do is change all the colors present in the image by subtracting or adding a certain amount of red, green and blue with the aim to get a image that looks natural - which in this respect means that it is more or less like you (thought you) saw it with your own eyes. Setting the WB thus has nothing to do with color management (getting the monitor and printer rendering the same color) but setting the WB is all about getting the image more or less as you thought you saw it yourself. It is not part of the color management process, but something needed because of the way we see things.

Human vision tries to achieve what is called color constancy - our eyes together with our brain want to see a white paper as white under all kinds of lighting and thus the human vision system makes adjustments when we are out in the dark! In fact we change reality to something we are more at ease with or better used to. Human vision is able to deal with quite a lot of natural light sources and their variety in components - a much wider range then the average settings for WB allow. It does so automatically - during the day we do not really notice that the color of light changes, unless you are out at dawn or dusk. The camera however, is dumb – and also more correct. It only records the luminance per photosite. With the knowledge that in front of each photosite there is a filter of known color it is possible to calculate the color that cannot be recorded itself. When generating a JPG or when converting a raw file in PP, the data of the luminance and the filter employed is combined and with some clever mathematics the correct color can be found. The camera has no trouble doing that, but must do so in a relatively short time. In PP we can take more time and thus the calculation (that in fact always is a approximation) can be more accurate. What the camera cannot always do, is match the result exactly to how we ourselves saw it. The camera is not capable to mimic our "color constancy system" yet and needs to make do with a few discrete settings (daylight, flash, tungsten etc) or a user added setting for the white balance. The camera also has a setting that can (over a certain range of color temperatures and more or less continuous light spectra) guess it about right: the automatic white balance (AWB) setting. Getting the WB right automatically is a rather complicated process that needs to be carried out for every image shot. More information on how that is done, see these links:

http://www.stanford.edu/class/ee368/Project_11/Reports/Cohen_A_New_Color_Balancing_Method.pdf
http://www.acad.bg/rismim/itc/sub/archiv/Paper3_1_2012.pdf
http://www.eee.hku.hk/optima/pub/misc/0800_SSI.pdf

In those papers various possibilities of how a camera can calculate a more or less correct AWB are explained - in rather complex and mathematical terms. Also there are quite a few examples that show that each and every method works well - under particular circumstances and not under other operating conditions. Obviously the human "color constancy system" is not yet fully understood so the mathematical model used to mimic it in-camera, still is somewhat limited.

5.1 How “right” should the WB be set?
Generally we strive to get it right – we want to look the image like we saw the scene at the moment of shooting. For some types of images we may like it somewhat warmer – people usually look “healthier” when the balance is shifted somewhat to the warm side. For other images we may prefer a somewhat cooler look. Setting the WB thus also includes a certain amount of personal taste – it is not exact science.

Image #8:



Image #9:



Image #10:




1. Original image, AWB, as shot, profile: Adobe Standard
2. WB: Lr tungsten, profile: Adobe Standard
3. WB: Lr fluorescent, profile: Adobe Standard
   
4. WB: dedicated, based on gray card, profile: Adobe Standard
5. WB: dedicated, based on gray card, profile: camera-specific
   

From the examples shown above (images #8, #9, #10) that were shot at night, it becomes clear that there are situations that we may not want a WB that is exact. Under industrial lighting it becomes clear that the procedure to render white as white no longer may be acceptable. It can be done, but the results look unnatural. That is because the human capability to sustain color constancy is limited and is not capable to deal with types of light that are not available in nature. Once we get out of our color constancy range we do more or less see colors as they are – under sodium lighting we do see the orange colors and do accept them.

In image #10 the last rendering (part 5 of the image) is the best in terms of WB – the color of the interior of my car and the red sign are rendered as they are – in more or less “dark daylight” that is. However, it in no way represents anything I saw myself at the time I took the shot. In city centers with more “human friendly" lighting, results will be better because of all kinds of brightly lit signs with primary and saturated colors, shop windows etc. If those are not available street lighting can also look very industrial. So, under rather unnatural lighting the standard method to get a white balance may not always be appropriate. It may be better to adjust the image by eye to more or less of what you saw or to what you like.

5.2 Dealing with difficult situations
From the above examples it becomes clear that there are situations were it will be difficult to get a acceptable WB. You may succeed in rendering white as “white”, but the result does not look anything like you remember it. The same is true for disco lighting with lots of monochrome saturated colors that change very frequently. We can no longer adjust sufficiently – so we accept them more or less the way they are. When shooting under those circumstances it may be better not to strive to render white as white, but just accept all the color casts!

6 Color temperature, light spectra and Color Rendering Index

The type of lighting used thus can have a large influence on how we see things. Light can be characterized in various ways. First there is the concept of color temperature – a way of indicating whether the light used is white or has a somewhat blue or yellow color cast. In the article written by MichelvA you can find more information about what color temperature is. A table showing the color temperature of various light sources can also be found here.

6.1 Various light sources and their color temperature
Even when different light sources have a identical color temperature, the way we see a object lit by them can differ. That is because some sources of light have a more or less continuous spectrum (like light from the sun) and others may not. Their color temperature may be equal (more or less being the weighted average of all the components included), but the composition that leads to that color temperature can differ quite a bit. That mix of colors can be qualified with the Color Rendering Index. A CRI of 100 means natural white light, when the CRI is lower that means that some components of the visible light spectrum may be far stronger then in natural light while others may be more or less absent and that the way colors are seen under that lighting are perceived as less natural.

A simple method to get a idea of the continuity of the spectrum of a particular type of light is shown here. In that document also some of the more common light spectra are shown. It is clear that some are continuous with more or less equal amounts of the various components, where others are definitely not. The types of light with a spectrum with “holes” in it or with dominant colors in the spectrum will still have a certain color temperature, but unfortunately that color temperature may not be a good starting point to base our white balance upon.

7 Some examples of difficult lighting.

In image #11 there are 4 photos shown of a band playing in a pub. The images themselves are not particularly interesting but they are chosen because all of them contain some sheets of white paper in the background effected by the colored LED-light. All were shot in A-mode, auto ISO (400-1600) and AWB with a a99 and the CZ 24/2 SSM. It was rather dark, no flash was used.

Image #11:



All images show a dominant color, and none of them looks anything like what would be visible in daylight. That was done on purpose – the lights are used to create a certain atmosphere that does not pretend to be natural at all. If under these circumstances you would try to render white as white by making a white balance reading on something in the image that you know is white, the result will change – but not necessarily for the better. It looks different alright, but most likely not more alike what you saw at the venue yourself. Your inner color constancy system realized that this was out of the range of adjustment – and so did not even try to cope with it, you more or less saw the colors as they were based on their RGB components!

Image #12:



In image #12 the same images are shown, and their white balance now is adjusted based on a white piece of paper in the background. The images look different – but they do not look any more realistic than when shot with AWB, although part 4 of the image looks more human friendly. In fact, the images that had the WB adjusted do look less alike as you remember the colors when shooting. Thus, under these circumstances, it may no longer be appropriate to change the rendering the way we usually do by referencing a white object to get the correct WB! This is not the fault of the camera. The reason is that your in-built color consistency system has gone in the no correction mode because it can no longer cope with the light it is exposed to!

If you leave the camera on AWB it still will try to render the color as best as possible, based on the algorithms build into it. The result may be something you like or not. My experience with this type of lighting is that AWB does quite a good job, better then when trying to adjust the WB based on a known white object. The camera in fact does it right – if only red light is used, a white object should be red! Thus, AWB may be a far better setting under these circumstances then using a custom WB or one of the presets available. Some people will argue that tungsten would have been the right setting to use. Maybe so - but the WB's reported by Lightroom in this case were 5700/+18, 2900/-7, 2650/+3 and 5200/+19 (temperature/tint) respectively for part 1, 2, 3 and 4 in image #11. Some thus more or less were akin to daylight with a green cast (part 1 and 4), the other two were regarded as close to tungsten (part 2 and 3). That may well have something to do with the monocolor LED-lighting - if one color dominates and most other colors in the spectrum are absent (the color of the light not being a mix of various colors), the concept of color temperature as we normally use it, may no longer be valid.

The above example of course is a situation where the lighting was chosen to be very unnatural trying to get a “sense arousing atmosphere”. Concert venues of rock-bands and nightclub interiors usually fall in that category. In theaters when attending a ballet venue, a classical concert or a play things may be different. Then light with a more continuous spectrum may be used, where a simple adjustment of the color temperature to tungsten may suffice to get good results. Whatever the case, there seem to be situations that AWB will get you a better starting point then any other setting.

There is yet anther option left if you want to remove a dominating color cast. With most raw converters and image editing programs you will have the option to make a curves adjustment. That option is mostly used to change contrast by introducing a S-curve or to emphasize the midtones. If however the software you use is also capable to individually adjust the curves for the red, green and blue channels, you may succeed in getting a rendering that you like quite a bit more. Just try it and see whether you like what you get.

8 Putting it all together

In this article and the article about color management from MichelvA three closely linked subjects were discussed – color management, profiling and WB setting together take care of the process of getting the color “right”. After that you can change whatever you want, but you then start out with a image that as closely resembles what you saw yourself.

It all starts with a correctly calibrated monitor. What is shown on the monitor is the material you work with – adjustments made there should be reflected in the print or on other (correctly calibrated) monitors. To arrive at that point your monitor needs to be calibrated – showing each color as indicated by the relative amounts of RGB and the color that results of that within the colorspace used. The various colorspaces differ somewhat. Most monitors will not be able to show more then 98% of AdobeRGB, in JPG’s only the smaller sRGB color space is available and in print you will even have less colors and intensities available, as can be shown when setting Lightroom or Photoshop into the proof mode. There is nothing you can do about that – the various color rendering methods each come with their technical limitations that you will have to accept.

In the color reproduction process there are three stages where you can make some form of adjustment. The data from the file the camera made that is supplied to the editing program is changed somewhat by that editing program to overcome shortcomings in the recording made by the camera. To a certain extend the data are rendered as the software developer preferred, which, although roughly correct, may not be exactly how you would do it yourself. You can change that by making a camera specific profile that tells the editing program to interpret the data in a certain way. That is the first step in the processing chain and adjustment may be very useful under certain lighting conditions and when you use lenses from different manufacturers. You can also change the way the printer handles the data coming from the image file - the final step in the processing chain. That is the printer profiling part – a step usually much less necessary unless you are using a printer/paper combination for which no printer profile is available. The second step in getting the color right is adjusting the white balance. Adjusting the white balance can be done by eye on a calibrated monitor or by using some sort of reference (white card, gray card etc). Under some circumstances (lighting with a low CRI value) you may not be able to do that – the best results under those conditions may be obtained by setting the camera to AWB and changing that setting somewhat during post-processing if you like to.

9 Some more questions addressed and (partly) answered

There are a lot more things that can be said about profiling, when to do it, whether it is worth the effort and how it relates to various other subjects. A few of them are addressed below.

9.1 Do I really have to pay attention to all of this?
No. In most situations the results you get will be quite adequate, as long as your monitor is set up about correct. Monitor calibration can be very useful. It can make post-processing much more reliable, repeatable and satisfactory. Correct monitor calibration thus is a essential step to obtain satisfying results in the long term. It is not costly and doing it right once can save you a lot of frustration and time.

Camera profiling needs only be done when results are not satisfactory after you have correctly calibrated your monitor. You can look at it the way you use various spices when cooking - they can add just that little bit extra to whatever you are cooking, but combining the right basic ingredients (correctly calibrating the monitor) is far more important to get a tasty meal. For most types of shooting it will not be necessary (unless you are a precision fetishist), but for awkward lighting situations it may well be essential.

http://www.youtube.com/watch?v=IrEtS3BnNkU

As printer profiling is concerned - I would leave it to the printing professionals

9.2 Does it make a difference whether I shoot JPG or RAW?
It does. The main reason is that JPG files are already "cooked" - the camera has interpreted the raw data and constructed a image file based on that. During that process the build-in converter carried out a specific type of conversion (including changes in contrast, white balance, color rendering etc as can be specified through the menu settings on the camera) and also has compressed the resulting data to get a smaller file-size (essentially a process where some information gets lost forever). The problem with JPG files thus is that you start with less data that also already have been changed in unknown ways...You can profile your monitor for that - but it will be only "correct" if you use the same JPG settings in-camera every time and also the in-camera conversion is carried out the same way. That will not be the case. Personally I would not bother trying to optimize the complete processing chain when shooting JPG-only. A quick and dirty calibration of the monitor based on the suggestions in the links listed above is more then adequate. When starting out with JPG's you have already lost so much in quality and control that introducing more precision later seems gilding the lily.

9.3 Can all screens be adjusted?
Yes, but some more so then others. You cannot expect to get the same results with a cheap HDMI capable screen for home use as with a professional and very costly monitor designed for the graphic industry. When put next to each other you will immediately notice that the costly screen shows far more subtle colors and thus makes processing easier. That is not to say that cheap screens can not be used - careful calibration can improve the rendering quite a bit. So they are quite suitable for post processing at home. In most cases their color rendering can be improved remarkably when you are able to remove color casts and get the luminance, contrast and saturation to more acceptable levels for image editing. Remember that most consumer type monitors are delivered with settings suitable for general office work and a nice dramatic rendering of images - not for critical viewing of graphic objects where correct color rendering is essential.

Laptop screens also can be calibrated. Results may vary though - some screens come with a kind of glass plate over them that gives very brilliant colors when viewed in broad daylight. Quite useful when used for general computing, but less so when using them for image editing. They simply show the images much more saturated and vibrant then they actually are and when printed. When calibrating such a monitor improvements can be made, but most times the monitor cannot be tamed completely. If you want to use a laptop i suggest hooking up a separate larger screen when post-processing and then only calibrate that additional monitor. CRT screens can be adjusted without problems. Checking the calibration at shorter intervals may be necessary though, because when aging, they will loose contrast and luminosity and may also develop color casts over time. CRT screens now are more or less a thing of the past, but if you still use them calibrating can improve their color rendering quite a bit.

9.4 What is the difference between a camera-specific profile and a Lightroom preset?
Sometimes when people hear about a camera-specific profile their reaction is: "I do not need a profile, I have made a preset that does it all for me". The results may indeed come very close - by sheer luck. The way you arrived at that result however differs enormously.

It is important to realize that the software always uses a camera-specific profile. That profile comes with the software and is selected based on the EXIF data. For some cameras you may even have a choice of several profiles that give slightly different renderings. It can be regarded as a look-up table that tells the software to change the incoming information somewhat and in which direction. For example it can be deemed necessary to change colors that contain 50% red by reducing the green and blue a little, or reduce the green a little and increase the blue somewhat. In theory such a look-up table could be made for each and every color that the camera can deliver! That information is stored in the camera-specific profile. Not necessarily for all possible colors, but for quite a number of them. Colors in between then can be adjusted by interpolation. If you adjust the colors you see on screen with the various sliders available, something quite different happens. A curves adjustment will influence the brightness of pixels based on their initial brightness - some will be rendered somewhat darker, others somewhat lighter depending on the adjustments you made. You can do that on a composite basis, or use different curves for the R,G and B channels. Whatever you do, all the pixels will be involved - based on their initial brightness and also on their color if you do it on a per channel basis. The same happens when you play around with the hue/saturation/luminance sliders. There is a effect on all the pixels - proportional to the amount of a specific color in that particular pixel. You can put all those changes, together with the software vendor supplied camera-specific profile, in a preset for future use. That can work quite well - for images that need the same type of adjustment that is.

Making a adjustment through a proprietary camera-specific profile is different. It acts again on each pixel, but does so individually on the basis of the original color in that pixel. That means some pixels get some more green, others some more blue, a little less red, a little less blue and some more green while others remain untouched. The adjustment for each and every shade of color can (in theory) be different. It can be regarded as a kind of individual "3D-adjustment", where the normal changes you apply during editing can be seen as "2D".

Of course that all can subsequently be stored in a preset also, but the important difference is the way colors are adjusted. The sliders available make all over adjustments, a camera-specific profile caters for individual per pixel adjustment. That individual per pixel adjustment is much better suitable for improving color rendering under awkward lighting and when the camera is unable to record some specific color/luminance combinations correctly. Remember camera calibration is the second step to get a correct representation of the image on the monitor. After that you may adjust whatever you like with the help of your editing program and the result need not look natural in any way. If it had to, B&W photography no longer would exist.

10 Final remarks

Calibrating your monitor and eventually optimizing other steps in the processing chain can result in better images. Better meaning that they more represent what you saw when you shot them when rendered on screen and in print. You can go quite far in that, but usually calibrating the monitor will be sufficient. Other steps may only be necessary under special conditions. Camera profiling can be quite useful when shooting with lenses from different manufacturers. Calibrating your monitor will not automatically get you better images. You still need to be familiar with your processing software to obtain optimal results. If you are quite happy shooting JPG's and the prints based on them - please continue by all means. Remember the old days when we shot film: some went to the trouble of developing film, enlarging and developing prints themselves to get the best results possible, others just handed over their shot rolls of film to a local store and were also enjoying the results they got that way. That principle has not changed! Photography can be enjoyed in various ways, choose the one that you like best.

Go to part 1




Edited by romke - 11 December 2013 at 11:58

It's great to see this info all bundled in a comprehensive way.

PS. the a900 is still going strong.