Channels and bit depth

 

Data is stored in computers using discrete units known as bits.  A bit is the smallest unit of information a computer can store.  A bit has a single on/off value.  For purposes of this discussion, it takes several bits grouped together, called a byte, to store a single letter, such as the letter B, or a single digit number, such as 3.  A byte typically consists of 8 bits.

Images are stored in computers using pixels (short for picture element).  A pixel is the smallest, complete unit of information about a single point in a digital image.  In an image stored in the RGB color model, one pixel stores the hue, saturation and lightness values about a single point.  Unlike letters and numbers, it can take 8 or more bits to store a single pixel.  The number of bits needed to store a pixel depends on how many channels the image has and the image's bit depth.  An image's color model determines how many channels it has.

---

Color Model

The common color models in digital photography are Grayscale, RGB, CMYK and Lab.  A grayscale image has one channel.  A RGB image has three channels: one for Red, one for Green and one for Blue.  A CMYK image has four channels: one for Cyan, one for Magenta, one for Yellow and one for Black.  The Lab color space has three channels: a Lightness channel to represent luminance, a channel named a that represents green/red and a channel called b that represents blue/yellow.

 

Channels

A channel, also known as color channel, is the information that defines the color and tonal information for just one of the colors in a color model.  When the information from all channels is blended together, that is how we get our image.  Figures 1A through 1D show the channels for each of the four color models.

 

Figure 1A.  A grayscale image has only one channel

Figure 1B.  A RGB image has three channels

Figure 1C.  A CMYK image has four channels

Figure 1D.  A Lab image has three channels

Adobe Photoshop Channels

In Photoshop, when we click on the Channels panel, we will see something similar to Figure 1 above.  For a color image, the top 'layer' is the composite channel.  The subsequent layers represent the individual, or component, color channels.  The composite channel represents all of the individual color channels blended together.  It does not actually physically exist in the image file.  Rather, it is a logical rendering of all the individual color channels.  The individual color channels are physically stored in the image file.

If we deactivate all channels except one of the individual channels, we will see a black and white version of the image in the Photoshop document window.  We can think of this as a mask that tells Photoshop how much of the channel's information to use when displaying the image.  When the various channels are blended together, we get our color image, as represented by the composite channel.

---

Bit Depth

Channel bit depth is the number of bits used to store  the information for a single channel for a single pixel.  A pixel is stored using either 8 bits or 16 bits per channel.  An image that uses 8 bits per channel to store its information is considered to have a bit depth of 8.  An image that uses 10, 12, 14 or 16 bits per channel is considered to have 10, 12, 14 or 16 bit depth respectfully.  However, since a computer uses bits in increments of 8, to store a 10, 12, 14 or 16 bit depth image, the computer will use 16 bits per channel.  So a 12 bit image will use 16 bits of storage but it will not contain 16 bits of information.  The other 4 bits will not be used.  Figure 2 is a graphic representation of a single RGB pixel being stored at 8 and 16 bits per channel.

In simplest terms, the more bits per channel the more information contained within the image.  More information means more possible colors and tones within a color model's color space.  How much more?  A single bit can only have one of two possible values.  Therefore, 8 bits used in combination can have up to 256 possible values because 2 to the power of 8 is 256.  Since a grayscale image has only one channel, a single grayscale pixel can store 1 of a possible 256 values.  However, a RGB image has three channels.  This means a single RGB pixel can store 1 of 16,777,216 possible values because a RGB pixel has three channels and each channel can have 256 possible values for a total of 256 x 256 x 256 = 16,777,216 possible values.  Table 1 below shows the maximum possible values a color model can have for a specific bit depth.
 

 

Color Model	Bits Per Channel	Bits Per Pixel	Possible Values
Grayscale	8	1 channel * 8 = 8	28 = 256
RGB	8	3 channels * 8 = 24	224 = 16,777,216
CMYK	8	4 channels * 8 = 32	232 = 4.3 billion
Grayscale	16	1 channel * 16 = 16	216 = 65,536
RGB	16	3 channels * 16 = 48	248 = 281.5 trillion
CMYK	16	4 channels * 16 = 64	264 = 18,446.7 trillion

Table 1.  Comparison of Color Model and Bit Depth

 

More is Not Necessarily Better

Referring to Table 1, one might assume the best way to store an image is in the CMYK color space with a 16 bit depth because this would give us 18,446,744,073,709,600,000 possible values.  However, we need to keep the following in mind.  First, the CMYK color model has a narrower color gamut than RGB.  So while there are a few colors in the CMYK color space that are not in the RGB color space, the RGB color space contains a broader range of colors than CMYK.  Second, computer monitors are not able to display that many different colors and monitors can usually reproduce more colors than an inkjet printer can.  Third, more bits may mean more data but it also means more storage.  An image with 16 bits per channel will be twice the size as its 8 bit counterpart.  Fourth, even though one of the major new features of Photoshop CS is the ability to work with 16 bit images, not all of Photoshop's commands will work with 16 bit images.  The image will needed to be resampled to 8 bit to use these commands.

---

Recommendation

Having said all of that, what do I do?  I use the RGB color model even in my black and white images.  Why?  Because a single 8 bit RGB pixel can contain 1 of  16,777,216 possible values whereas a 8 bit grayscale pixel can contain only 1 of 256.  But why does that matter in a black and white image?  Because those 16.8 million possible values include tone, not just color.  This is one of the reasons I developed the Tone Management System for creating black and white images from a color image: to retain as much information as possible.  You can learn more about it on the Tone Management System page.

What about bit depth?  I try to capture and work in 16 bit as much as possible.  Many digital editing techniques stretch, compress, reshape and do all kinds of things to an image's tonal range.  The image can begin to lose its smooth gradations from one tone to another and banding, also known as posterization, can appear.  Banding will appear as gaps in a histogram.  By working in 16 bit, we give ourselves more information to work with and reduce the risk of banding.