This article was written by Joe Williams who, in his own words, "... did it so I wouldn't need to rediscover it all six months from now. Perhaps I can send it to a newbie someday to help get him over the mountain that took me two weeks to climb." I'm pleased to make it available here.
Opinions expressed are those of the author and do not necessarily represent the views of www.rogercavanagh.com. Actually, I pretty much agree with everything Joe says - I just couldn't resist putting that in.
For more information on Colour Management, see the links on this page.
Introduction
A Colour Management System (CMS) is essential for serious colour photography. If you ignore it, you may get acceptable results, but understanding CMS is essential for top quality photographic images and prints.
Colour is device dependent, so a colour image data file will produce slightly different colours depending on the device used to capture or display it. If you scan a photo into a computer and then print it, the resulting photo is unlikely to match the original, at least not exactly. Each device works in a different colour space, so an image file will not produce exactly the same colour on every device.
One way to compensate for the device-dependent nature of computer image files is to use ICC profiles. An ICC profile defines the colour space in which a particular device works. Every computer file that represents a colour image exists in some colour space, whether or not it is tagged with an embedded ICC profile. By calibrating a device against known colour standards, a device profile can be produced that describes that device’s colour reproduction.
Colour images are represented in computer files as numbers. In an RGB file, there are three numbers per pixel to represent the values of red, green, and blue in the RGB additive colour model. The more bits used for each number, the more variations of colour can be represented. In an 8-bit JPEG file, for example, there are three bytes per pixel – one for each colour: red, green, blue - so a total of 2^24 colour shades can be represented. That equals 16,777,216 shades of colour. Typically a digital camera will record 12 bits per pixel, so it is theoretically able to represent 2^36, or 68.7 billion colours. Strictly speaking, apart from the Sigma SD9, each pixel in a digicam captures only one colour with a matrix arrangement that allows interpolation of full colour at each site from the surrounding pixels.
Two common input devices are digital cameras and scanners. An image generated by either is only fully defined, if tagged with an embedded ICC profile. Likewise, output devices such as monitors and printers can have ICC profiles to describe how they reproduce colour.
Some common colour standards are sRGB, an attempt by Microsoft and HP to standardize across devices, Adobe RGB (1988), and Apple RGB. There are many more. In the Windows world, untagged files are usually sRGB. An application can be CMS aware or not. If it is not CMS aware, it probably works in the sRGB colour space. If it is CMS aware and opens a non-tagged file, it probably assumes the file is sRGB. Because of the market dominance of Microsoft and Adobe, sRGB and Adobe RGB are very popular colour spaces. Also, because Macintosh computers are widely used in the desktop publishing industry, Apple RGB and ColorMatch RGB are popular.
Digital Cameras
A particular digital camera generates an image file and may or may not tag it with an ICC profile to describe the colour space of the file. Some cameras, such as the Canon 10D, give the user the option of using either sRGB or Adobe RGB.
The table below describes two particular cameras for illustration. The Canon Powershot G3 works strictly in the sRGB colour space. The Canon EOS 10D has a user option for sRGB or Adobe RGB (1998), user selectable by menu.
Notice that any image files generated by these two cameras will either be tagged as sRGB or not tagged at all. Nevertheless, the 10D can produce JPEG files in two colour spaces even though the camera designers did not see fit to tag Adobe RGB JPEG files (apparently, the fault of JEITA who created the EXIF 2.2 specification). The fact that an image file is not tagged does not mean it has no colour space. It simply means the file has not yet been tagged with the appropriate profile. Raw files tend to be proprietary among various camera manufacturers. Canon raw files are in the sYCC colour space, which apparently cannot be used as a working space like sRGB or Adobe RGB. Conversion programs always convert them to another colour space, such as Adobe RGB or ProPhoto RGB.
CMS Applications
Some high-end applications are “CMS-aware”, meaning they know how to use colour profile information tagged onto an image file. An example is Photoshop. Starting with version 6.0, Photoshop is capable of simultaneously working with multiple images, each tagged in a different colour space. Also, a Photoshop user can select (as an application setting) a preferred “working space.”
Photoshop is giving the user four choices:
1 Just forget about colour management because exact colours are not important. This might be a sensible case for web graphics, or if you just don’t care about exact colours
2 If you know the working space of the image (such as with a Canon Adobe JPEG image), then tag the file with the correct profile as selected from a drop-down menu.
3 If you know the working space of the image (such as with a Canon Adobe JPEG image), then tag the file with the correct profile as selected from a drop-down menu, and CONVERT the image to the working space. (Adobe RGB in this case.)
4 If you know the working space is actually Adobe RGB (the current working space), simply tag the image to so indicate.
Assign vs. Convert
There is a subtle point here that must be understood. There is a difference between ASSIGNING a profile to an image, and CONVERTING an image to a different profile. You assign a profile to an image when you already know its working space, and simply need to add an ICC profile to tag along with the image data. None of the actual image data is changed in any way. However, when you convert an image from one working space to another, the actual image data is modified, and then an ICC profile is added to tag along with the image data.
The preceding paragraph may seem a little counter-intuitive because a converted image will look the same on the Photoshop screen before and after a conversion. If, however, you were displaying an sRGB image and assigned a different profile without also converting, then the image would look different before and after the assignment, even though the image file data did not change in any way. Assigning does not change the image data, but converting does change it.
In this case the user wants to open a camera image already tagged as sRGB, but the user has also told Photoshop that Adobe RGB is the preferred working space. Photoshop is quite happy to do any of the following:
1 Open it as an sRGB image anyway, because Photoshop can open multiple images all in their own tagged working spaces.
2 Convert the image to Adobe RGB, tag it as Adobe RGB, and then open it.
3 Just open it without any worry as to how it will look on the screen. Again, this is most useful for web images where exact colours don’t matter. (Very few web users have calibrated monitors anyway.)
Monitors and Printers
A colour-managed application can also deal with ICC profiles for output devices such as monitors and printers. Without colour management, no two monitors will display identical colours. When you email Grandma a JPEG photo, you have no idea what she is seeing on her screen. Ever been in an electronics store and watched the same football game on twenty different TV sets?
Printers are the same. Without ICC profiles and colour-managed applications, what you see is, likely, not what you get. In a fully colour-managed environment you can take a photo, view it on your screen, and print it out with all the colours matching at every step of the process except in situations where the colour gamut of, say, the printer and paper combination doesn’t quite match the gamut of the working space. In these cases, colours that cannot be exactly matched will be modified to fit into the printer space. Using the gamut warning in Photoshop proofing will tell you where colours cannot be matched.
Raw Vs. JPEG
The idea of a raw file is to preserve all of the sensor data captured by the camera. No conversion processing is done inside the camera; so all processing is deferred until the raw file is downloaded to a PC. This is in great contrast to a JPEG where all the conversion processing is done inside the camera, including things such as white balance and sharpening. This process is irreversible, so if the camera white balance, for example, was set wrong, it may be tricky to adjust in Photoshop, whereas, in most raw converters, changing the white balance is trivial.
Worse yet, the 12-bits per pixel of sensor data is truncated to 8-bits for the JPEG conversion. Also, JPEG uses lossy compression to reduce file size, which means it discards some additional colour information as a trade-off for smaller files. Consequently, JPEG may produce some artifacts, such as halos around dark objects contrasted against a bright background. The artifacts can be worsened later during post-processing such as sharpening.
Now don’t get me wrong, JPEG is a fine format and is ideal for the consumer who does not want to “develop” his own film and is used by some professionals who are satisfied that JPEG suits their requirements. But as photographers have always done, they get the absolute best results by having control over every step in the photography process, and do not allow the camera to do processing that they prefer to defer until later. A rough comparison is Polaroid vs. regular film, or taking the film to the local drugstore instead of processing it yourself.
Working with Digital Camera Raw Images
Why even bother with digital camera raw images when the camera will output a JPEG image, complete with its ICC colour profile embedded? To answer the question, let's look at the steps used to convert a raw image to a print. Digital photographers like to call the particular procedure they use a “workflow.”
A raw image from a digital camera could be a 16-bit TIFF file, or it could be a proprietary compressed format that is lossless, preserving all the sensor data. Since most digital cameras record 12 bits of data per colour per pixel, the raw file preserves all 12 bits. Once a raw file is converted to TIFF, it is quite large. For example, a six-megapixel Canon 10D image that is 3072 x 2048 pixels requires about six MB for raw format, and 36 MB for a 16-bit TIFF file.
Prior to Version 7, Photoshop would not even read a camera raw image, and even now it requires an plug-in. So we find that digital photographers often use a number of separate applications to download, process, manipulate, convert, and print images. All of these applications must be CMS-aware for the workflow to produce the desired result.
Here is a simplified workflow at a high level:
- Download the raw images from the camera to the PC. Some programs rename images in the process for better organization of a large number of files. An example of a stand-alone program that does downloads and renames raw files all in one step is Downloader Pro, sold by Breeze Systems. [Editor's note: see the Downloader Pro review.]
- Perform the image processing that the camera would otherwise do internally for JPEG files. Keep in mind that you are using all 12 bits per colour per pixel during this processing, and that the processing can be done over and over until you get it just right. This processing can include white/gray balance, correcting for any colour cast, exposure compensation, levels and curves adjustments, focus/sharpening, and conversion from raw to TIFF format.
- Save the 16-bit TIFF file to preserve all of the colour data that was originally captured by the camera’s sensor. One of the better and widely used programs for this step is Capture One DSLR Pro or Capture One DSLR Limited Edition, developed by Phase One. Another popular program is BreezeBrowser, also sold by Breeze Systems. [Editor's note: for more on working with BreezeBrowser and Capture One, see My D30 Workflow and My C1LE Workflow.]
- Open the image in Photoshop and make any final tweaks. Capture One is powerful enough that often Photoshop may not even be required. Photoshop can also be used to resize, resample, and convert to 8-bit JPEG for web images.
- Stand-alone printing programs are available that take away the work of reformatting for printing that Photoshop requires. A popular CMS-aware printing program is Qimage sold by Digital Domain.
- Finally, programs are available to keep a database of thumbnails on line for images that have been archived to CD ROM or DVD ROM. Most serious photographers save their raw files, and some save the TIFFs also. This takes a lot of storage, and archiving to a medium like DVD is essential. A good example of such a program is ThumbsPlus sold by Cerious Software.
3 August 2009, originally published 2004