The Alpha Channel

In this article, you'll learn about the basic properties and definitions of the Alpha Channel. We will help you to understand and use this powerful graphic design feature for your future art creations.

The Alpha Channel is a term for computer graphics and was introduced by Alvy Ray Smith and Ed Catmull in the late 1970s at the New York Institute of Technology Computer Graphics Lab.

The alpha channel concept was fully established in a paper in 1984 by Thomas Porter and Tom Duff.

In the alpha channel, transparency information is stored in addition to the actual image data in various graphic format options (e.g., PNG, PSD, TGA, DDS, or TIFF). If a digital image or video footage contains an alpha channel, it signifies that additional transparency information is contained within the file besides the color components.

The alpha channel contains no color information, only transparency information. The channel is always visualized as a grayscale image. The color is typically represented in the RGB values. An alpha channel usually has the same depth as the color channel of an image. For an 8-bit image, for example, an alpha channel comprises 256 levels. The maximum number of possible transparency gradations depends on the number of bits used for the alpha channel.

The image files that contain the alpha channel are generally composed of the following components:

R = Red Channel
G = Green Channel
B = Blue Channel
A = Alpha Channel

The RGB channels are numerically represented via the HEX Number, which is typically displayed in the Color Picker Interface of your graphic software.

In the example below, you can see the same image with:

1 – A white background.
2 – A checkerboard background
3 –The grayscale image of the alpha channel. Lighter areas are more opaque; the darker areas are more transparent.

The Alpha Value can be measured as a Real Value, a Percentage, or an Integer:

• The full transparency is defined as 0.0 ; 0% or 0, visualized in the alpha channel image as pure black.
• The full opacity is defined as 1.0, 100%, or 255, visualized in the alpha channel image as pure white.

Most graphic design applications offer an adjustable opacity of the alpha value.

If you, for example, reduce the opacity of the foreground elements to alpha = 40%, the color of the underlying layer will become visible at 60% opacity in the areas where both layers overlap. The resulting color-blending effect will create the viewer's illusion of an entirely new color. In our case, the color blending effect of the black source color with an opacity of 40% combined with the white background image creates the illusion of a gray rectangle.

Two black rectangular shapes on a white background. The rectangle on the left has the value of alpha = 100 %. The one on the right has the value of alpha = 40 %.
The blending effect between the black rectangle and the white background layer creates the illusion of gray.

Alpha Channels can be received and exported in three different ways:

1. As an external or separate alpha channel
2. As an integrated (premultiplied) alpha
3. As a straight alpha

1. External or separate alpha

When saving a file in the form of an external alpha channel, the transparency information is saved as a separate file. It is a grayscale file that must be interpreted as a transparency channel when opened. To display an image with an external alpha channel, both the image file and the alpha channel file must be imported, and the correct transparency of the file is established using the rendering application. This process is also called alpha channel interpretation.

2. Integrated (premultiplied) alpha

With the integrated alpha channel (premultiplied alpha), the transparency information is stored in a separate channel, as with the direct alpha channel. The main difference in the direct alpha channel is that the image's color values are blended with a background color. How much of a pixel's original color is used depends on its opacity. Semi-transparent pixels, for example, are saved with 50% black and 50% background color. This integration of the alpha channel into the color channels is not done instead of, but in addition to, storing the alpha channel. This enables the correct display of the image even in applications that do not support an alpha channel.

3. Straight or integrated alpha

In files that contain the alpha channel as a so-called direct alpha channel (straight alpha), the transparency information is stored in a separate channel, in addition to the color channels. A pixel within the file is stored not only with three values (R, G, B) but with four values: (R, G, B, α). For example, a dark blue pixel displayed at full opacity would be stored by the values (0, 0, 170, 255) at a color depth of 8 bits. The direct alpha channel does not change the color channel information of the image.

The binary alpha channel

A binary alpha channel is a minimal alpha channel that uses only 1 bit and can only indicate whether an image pixel is either fully transparent or opaque.

Comparable to the use of a binary alpha channel, a color of an image can also be defined as transparent. This type of transparency is used in Graphics Interchange Format (GIF), which allows one less hue to be used for rendering. However, it is not an alpha channel in the technical sense since the transparency information is not stored individually for each pixel.

Alpha as a masking and selection tool

Alpha channels can be used as layer masks and selecting tools in various image editing programs. Selections can also be saved in the mask image for later use. In Photoshop, temporary alpha channels are created when a layer is masked. The current selection can also be used to crop an image. This is often done by copying an existing color channel to an alpha channel, which can then be quickly edited using the image editing functions (curves, contrasts, brushes, etc.).

Image cropping via the alpha channel is a masking technique that divides a selected area of a subject in black, white and grayscale pixels. White relates to the selected pixels, black relates to the non-selected regions, and the different grayscale partially relates to the selected and non-selected areas.