It is tempting to think of an image as something that corresponds with a single display window, or to a single file such as a JPEG file, but really a GIMP image is a rather complicated structure, containing a stack of layers plus several other types of objects: a selection mask, a set of channels, a set of paths, an "undo" history, etc. In this section we are going to take a detailed look at all of the components of an image, and the things you can do with them.

The most basic property of an image is its mode. There are three possible modes: RGB, grayscale, and indexed. RGB stands for Red-Green-Blue, and indicates that each point in the image is represented by a “red” level, a “green” level, and a “blue” level. Because every humanly distinguishable color can be represented as a combination of red, green, and blue, RGB images are full-color. Each color channel has 256 possible intensity levels. More details in Color Models

In a grayscale image, each point is represented by a brightness value, ranging from 0 (black) to 255 (white), with intermediate values representing different levels of gray.

Essentially the difference between a grayscale image and an RGB image is the number of “color channels”: a grayscale image has one; an RGB image has three. An RGB image can be thought of as three superimposed grayscale images, one colored red, one green, and one blue.

Actually, both RGB and grayscale images have one additional color channel, called the alpha channel, representing opacity. When the alpha value at a given location in a given layer is zero, the layer is completely transparent, and the color at that location is determined by what lies underneath. When alpha is maximal, the layer is opaque, and the color is determined by the color of the layer. Intermediate alpha values correspond to varying degrees of translucency: the color at the location is a proportional mixture of color from the layer and color from underneath.

In GIMP, every color channel, including the alpha channel, has a range of possible values from 0 to 255; in computing terminology, a depth of 8 bits. Some digital cameras can produce image files with a depth of 16 bits per color channel. GIMP cannot load such a file without losing resolution. In most cases the effects are too subtle to be detected by the human eye, but in some cases, mainly where there are large areas with slowly varying color gradients, the difference may be perceptible.

The third type, indexed images, is a bit more complicated to understand. In an indexed image, only a limited set of discrete colors are used, usually 256 or less. These colors form the “colormap” of the image, and each point in the image is assigned a color from the colormap. Indexed images have the advantage that they can be represented inside a computer in a way which consumes relatively little memory, and back in the dark ages (say, ten years ago), they were very commonly used. As time goes on, they are used less and less, but they are still important enough to be worth supporting in GIMP. (Also, there are a few important kinds of image manipulation that are easier to implement with indexed images than with continuous-color RGB images.)

Some very commonly used types of files (including GIF and PNG) produce indexed images when they are opened in GIMP. Many of GIMP's tools don't work very well on indexed images–and many filters don't work at all–because of the limited number of colors available. Because of this, it is usually best to convert an image to RGB mode before working on it. If necessary, you can convert it back to indexed mode when you are ready to save it

GIMP makes it easy to convert from one image type to another, using the Mode command in the Image menu. Some types of conversions, of course (RGB to grayscale or indexed, for example) lose information that cannot be regained by converting back in the other direction.