Color spaces, from Wikimedia Commons; reused under a Creative Commons license. Image by Cpesacreta. What is clear here is that none of the common color spaces or reproduction methods can display the entire visible spectrum.
Starting from the start, a colour space defines all of the possible tones and hues that a single pixel may take. The values may take either a RGB value or CMYK value; the set of three numbers represents the amount of each colour present from a scale of 0-255, which gives 256 possible tonal values for each colour (red, green, blue or cyan, magenta, yellow and black). Combining every possible permutation of these gives you 16.7 million possible different colours for RGB. Note that you don’t get 4.3 billion colours for CMYK, because the K (black) value controls the overall brightness and density rather than contributing another possible hue to the mix.
The reason why we have different colour spaces is because they define the limit of reproducible tones for a particular reproduction method – be that screen, web, print, or TV transmission. The most commonly used colour spaces are either a type of sRGB or newsprint CMYK; both of these actually offer very limited reproduction potential. Many of you will notice that most images you see on the web, or in print, are lacking in depth and tonal subtlety; this is simply because the desired tone simply cannot be reproduced, and as a result lands up defaulting to the nearest possible colour value – which is obviously not going to accurately represent the original image. When a colour goes outside the possible range, this is known as gamut clipping. Some of the better proofing software can be configured to display a warning when this occurs; Camera Raw shows a small warning triangle in the upper-right near the histogram.
For photographers, the important colour spaces you need to be aware of for digital display are Adobe RGB, sRGB and ProPhoto RGB. For print, it’s whatever variant of CMYK your printer uses. Let’s start with the former. Adobe RGB is the most common wide-gamut colour space available; almost every camera today – even compacts – has the option to output in this colour space by default – use it. sRGB is a limited, mostly web-safe gamut that varies slightly depending on the standard; one camera maker’s sRGB won’t be the same as another, and the sRGB displayed online will be different yet again – possibly depending on the browser, or your system settings, or any one of many other factors. Avoid this wherever possible. The final RGB – ProPhoto – provides the widest of all gamuts; however, most monitors aren’t capable of displaying the majority of the possible colours, and few web browsers support it. I’m going to leave covering CMYK until the section dealing with print.
Although the tonal limitations of each colour space depend very much on the specific colour space themselves, it’s safe to say that in general, you’ll see the difference between Adobe RGB and sRGB most prominently in the blues and greens. There’s a particular shade of sky blue that seems to be nearly impossible to reproduce in sRGB, for instance. CMYK has similar restrictions to sRGB, but biases towards cyan instead of green and blue; overall, it lacks the vibrance and saturation that RGB can deliver – unsurprising given that the constituent colours are not red, green and blue! CMYK is used only for print proofing and is never found in camera or monitor; this is because the native components of these devices are formed of RGB photosites or pixels.
The observant of you will have noted that 256 tonal values represent 8 bits of information per colour channel. So why do we bother with 14 bit raw files, and working in 16 bits? Simple: although our output may always only be 8-bit, the amount of information we have going in affects the amount of work we can do to a file before we start to see posterisation (separation of areas into distinct blocks of colour with no tonal variation or smooth transition between adjacent zones). If we have 256 input values, do some contrast adjustment (effectively, ‘stretching’ the histogram) – we might now make some of the levels cover adjacent levels, resulting in only say 150 truly distinct tonal values for a particular channel. Most of the time, software will cover these ‘steps’ reasonably well; however, the reality is that you will see some posterisation. Working in a 16-bit colour space – with 65,536 tonal values per channel – avoids this problem mainly because we can’t actually achieve this many distinct tonal values through most reproduction methods; everything is effectively down sampled before output. Make sure you have your editing software convert any files in other colour spaces to the working colour space, too.
In-camera, the best option you have for maintaining accurate colour is to shoot RAW, Adobe RGB and whatever the highest bit space your camera offers; for the current batch of Nikons that would be 14-bit NEF, Adobe RGB, lossless compressed or uncompressed. (Lossless compressed only discards information in portions of the tonal register that aren’t being used or are adjacent duplicates, not any of the actual image data). Any time you’re shooting JPEG, you’re limited to 8 bits – and every file is compressed; avoid shooting JPEG unless you absolutely have no choice (A non-compressed JPEG is effectively a TIFF or bitmap). After being used to the tonal elasticity of of manipulating a good RAW file, you’ll be surprised at just how fast a JPEG will clip or posterise when manipulated – it’s one of the reasons that I almost always avoid buying a camera until there’s full RAW support for it through my usual workflow (ACR>PS). And it’s also important to note that RGB channel histograms and overexposure warnings are important: once a channel clips, it’s gone for good, especially with JPEG files. Although most raw converters will allow for some interpolation of surrounding tonal information to recover and reconstruct some highlight data, it won’t be that accurate.
All of this care during capture would be useless if not maintained during the postproduction process – that’s the importance of your screen workflow. Firstly, you need to have a good monitor that’s capable of displaying a wide gamut; the best of today’s bunch (Eizo, some NEC, Apple Cinema Display) are capable of covering almost all of the Adobe RGB gamut; if you’re serious, this is the kind of monitor you want. Secondly, it needs to be calibrated – i.e. ensuring what you see on screen accurately represents the actual data. The Monitor Calibration Utility for Apple (under System Preferences, Displays) is actually pretty good at this if you do all of the steps properly – a handy tip is to have an image which you’re familiar with open in another window while you run the wizard to ensure that the end output looks accurate to you. For Windows users, you’re recommended to invest in a Spyder or something similar – this generates a profile that the monitor then uses create its display output.
The next step along the chain is output: what are you going to use the image for? If it’s print, save as a 16 bit uncompressed TIFF; this will give the printer as much information as possible to work with when performing the RGB to CMYK conversion. I don’t actually recommend performing this conversion yourself unless you have the exact colour profile your printer is going to use, otherwise you might land up with some strange hue shifts. If it’s for screen or web use, then a jpeg is fine – most viewers are not going to have the right equipment to view the full gamut anyway; thus it’s better that you run a test proof under as close to actual viewing conditions as possible. I wouldn’t advocate going to sRGB unless you know that’s the only possible output; your best choices these days are Safari and Firefox – both are available for Windows and Mac. Similarly, ensure that your web hosting service preserves as much of the colour information as possible; the only one I know of that doesn’t convert things to sRGB is Flickr. Facebook et al are absolutely horrible – not only do they compress the hell out of the image, but they also shift everything into a very restricted web-safe sRGB that makes things appear both tonally blocky yet ‘coarse’ at the same time, due to the compression. Do NOT use Facebook to display images unless you have no choice, don’t care, or didn’t have the tonal information to begin with (smartphones, for instance) – everything just looks bad.
Printing is a whole article unto itself, but I’m going to touch on it briefly here: the main disconnect between the print workflow and the capture workflow is colour space; screen viewing involves an additive method where R, G and B are mixed together to make the desired colour; print uses C, M, Y and K inks subtractively to create colour. The reason for this is simple: pixels are backlit, prints aren’t. You’re dealing with reflected light off the print medium as opposed to transmitted light. Although this conversion process keeps improving, the mapping isn’t perfect and there still remain portions of the Adobe RGB gamut that can’t easily be reproduced in print. Part of this is due to the subtractive method; part of it is due to the fact that ink drops are either there or they aren’t – to create the illusion of tonal variation, printers use very, very small drops and dithering or half toning to leave white space between the dots. Note that print DPI is not the same as screen DPI – one pixel may be represented by anywhere up to 12 dots of ink! By far the best option for printing is to ensure your RGB output file is as accurate and full of information as you can manage; then find yourself a competent printer. I’m not going to get into self-printing – suffice to say that I did try, but between the wasted test prints, the clogged heads, the cleaning cycles…it simply wasn’t economically feasible for me to maintain my own printer.
There’s one final thing you have to take into consideration when an output image is being viewed: the effect of ambient light. It’s less important for devices that supply their own light – LCD panels, for instance – but it still matters because ambient brightness might overwhelm the panel and make colours appear dull or washed out. It’s far more critical for print viewing; the colour temperature of ambient light will affect the perceived white point, as well as the light reflected off the print itself. This causes toning or shading of the print; a good print master will adjust the yellow and blue components of an image for the intended display location; for instance, under tungsten light you have to remove some yellow component and/or increase the blues slightly as the light itself will provide that; the opposite is true for LED lighting. This is why all critical print proofing should be done in daylight – or under a daylight (5500-6000K) source. There are special daylight spectrum fluorescent tubes available for this purpose.
You’ll notice that I haven’t said anything about black and white workflow – it still matters, though less so. Assuming your printer can create a pure black and white without any hue shifts, then the important part is to check your grayscale space – this works in a similar manner to colour spaces, but controls the gamma of the image rather than the actual range of possible tonal values. This is critical to ensure that the resulting output image has the right density.
I’m going to finish with a final note on my own workflow. I run a 15″ MacBook Pro, calibrated with the Monitor Calibration Utility. I shoot 14-bit lossless compressed RAW in Adobe RGB on my D800E, M9 and OM-D; the RX100 is JPEG-only for the moment (Update: now supported by ACR/ LR as at October 2012). Files are opened in Adobe Camera Raw for initial adjustments (even JPEGs) before conversion to the working file format in Photoshop; I keep everything in 16 bit Adobe RGB until output. My web output is 8-bit JPEG; everything else is 16-bit lossless compressed (LZW) TIFF, or Photoshop (depending on the use or client). I will do some CMYK conversions for clients if they can supply the working CMYK space; otherwise, if it’s print, I leave the conversion to my printer – he knows the output capabilities of his equipment far better than I do, and I’ve yet to be disappointed with any of his output. It’s important to note that although I’ll select the colour space that retains as much of the original tonal information as possible, there’s also no point in bloating files if the information simply wasn’t there to begin with in the first place; I’m not going to save a conversion from a 12-bit RAW file as a 16-bit TIFF because there simply isn’t that much information after manipulation; let alone a JPEG. These will be saved as 8-bit compressed TIFF files instead.
Although the colour management process can be daunting, it’s important to invest time in understanding it and get it right – you’ll find afterwards that your images look a lot more consistent, regardless of the display medium. MT
Visit our Teaching Store to up your photographic game – including Photoshop Workflow DVDs and customized Email School of Photography; or go mobile with the Photography Compendium for iPad. You can also get your gear from B&H and Amazon. Prices are the same as normal, however a small portion of your purchase value is referred back to me. Thanks!
Images and content copyright Ming Thein | mingthein.com 2012 onwards. All rights reserved