Pixel Depth in Color Images
The next step up in image complexity is color. There are several different formats of color image (fig. 5.2), and the number of colors they can display depends either on how many channels they comprise or whether they are indexed color, i.e. GIF (graphics interchange format).
GIF images have no place in the printing industry and are strictly for on-screen viewing such as on websites. Indexed color can show a maximum of 256 colors, which can come from anywhere in the entire RGB range, rather than having to be shades of the same hue. If the colors are dithered, i.e. pixels of one color are scattered throughout a patch of another color, they visually merge to create the appearance of a color that is not actually present as one of the 256. Because of this ability, they can appear to be very convincing “full-color” continuous-tone images on screen. So convincing, in fact, that you cannot tell just by looking at your monitor whether the image is GIF, JPEG, TIFF, or EPS format, or whether it is seriously compromised in terms of potential print quality.
If you do have to use a GIF image, it needs to be converted into an acceptable format before it can be printed, and even then it is probably not going to be great because of having been constrained to 256 colors. However, occasionally there are ways to get there (see Chapter 7 for some suggestions on how to turn a GIF or lower-quality JPEG image into something that will print as a convincing color image).
In a 24-bit RGB color image (which is the most common format produced by desktop scanners and digital cameras) there are three separate “channels,” each one holding an 8-bit image. (A channel is the part of an image that holds all the information for any one of the component colors.) When they are displayed simultaneously, the color each one contains becomes transparent, thus allowing all their overlapping color combinations to be seen. (They can also be looked at individually, in which case it is usually more useful to view them as grayscale images, especially if you are considering changing an RGB image into grayscale format—see “Useful grayscale options” in Chapter 11.) Thus the three channels together are potentially able to display all the resulting color combinations. This means that each pixel can display a mix combining any one of the 256 shades of red, plus any one of the 256 shades of green, plus any one of the 256 shades of blue—in other words, any one of 16,777,216 colors.
Now consider what the addition of yet another channel does to this number, since in a CMYK image yet another grayscale map is added. Thus, our 16.7 million is multiplied again by 256, giving us a potential range of 4,294,967,200 colors. This is clearly more colors than we can ever distinguish between. Even the most accomplished painter who spends a lot of time considering color might only be able to distinguish 20–30,000 of them. However, even though our computers offer us more colors than we can ever distinguish between, it is not a problem—just think of it as an advantage that we cannot actually use.
Someone once tried to sell me a scanner based on its ability to produce 48-bit CMYK images—lovely, no doubt, but pointless. That is 12 bits per channel. Even in an RGB image, a channel depth of 12 bits would allow each pixel in a single channel to display any one of 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2, or 4,096 colors. Multiply that by itself three times (because it is a three-channel image) and you have nearly 69 billion colors, which is pretty silly. In CMYK it is even sillier: 281 thousand billion.
The important information about any scanner is much more likely to be contained in the answer to the following question: what is its optical scanning resolution? At what resolution can it scan before it starts interpolating the image resolution, i.e. spreading the same information over a progressively larger and larger number of pixels? The optical resolution is what will allow you to pick up detail that will otherwise get left behind, so it is extremely important—much more important than having oodles of colors that you cannot even distinguish between. Decent optical resolution starts at around 1200dpi for “reflective” art, where the light reflects off the object and bounces back to the sensor, and 2400 or (much better) 3600dpi for transparencies, where the light passes through the object to the sensor. The reason for the difference between these is that reflective art such as a print is usually much bigger than a transparency, which can be as small as a 35mm slide. These days, even a fairly cheap scanner is capable of producing an RGB image from reflective art that can be turned into a decent CMYK image.
Given that the range of color in RGB is larger than in that of CMYK, it is perhaps surprising that a CMYK image can hold a greater number of colors than the same image in RGB. This is due to confusing the range of colors within the two color spaces with the mathematical limitations imposed on digital images of both types.
The key factor here is that the number of colors potentially available in either format depends only upon the number of color channels in the image and the bit depth of those channels. If we compare a CMYK image with four 8-bit channels with an RGB image with only three 8-bit channels, clearly the CMYK image must have the potential to display more colors. That potential is a purely physical limitation imposed by the way the software deals with things. It has nothing whatsoever to do with how many colors actually exist in either system—which, perhaps confusing the issue slightly further, is infinite in both cases: but only because in theory there is no limit to the number of shades you can divide a color into. In practice, the range is larger in RGB. RGB contains almost every CMYK color, but CMYK does not come even close to containing every RGB color. If it did, there would be nothing to discuss—and many fewer problems!
5.2 This image has been divided into different bit depths and formats. From left to right: Bitmap (1-bit deep), grayscale (8-bits deep), duotone (16-bits deep, i.e. two 8-bit channels), and CMYK (36-bits deep).
