There are quite a few different pixe types used within nGI. This chapter explains most of them. The simplest type is monochrome and capable to describe colorless shades of gray, ranging from darkest black to brightest white. There is also an additive color type built from red, green and blue primaries, named rgb. To both the monochrome and the rgb type, nGI adds opacity or alpha, the opposite of transparency. This gives the ability to describe gradually transparent colors. In addition to these four basic color types, there exists a set of color spaces which all have their own color types.
Color types are usually multi-dimensional. For example, an rgb type can be considered as a vector with red, green and blue components. Besides the rgb type, nGI provides a few other types that one could classify among the color types. Grayscale and color can be combined with transparency or alpha.
Color types can be converted into other types. The arrows specify automatic or implicit conversions. Other conversions are narrowing (i.e. information is lost) and need to be explicit.
The color types combine with the arithmetic types and are used in a lot of other types, such as buffers (images, histograms, palettes, profiles, etc.), locators and views.
The arithmetic types consist of various integer types and some floating point types. The integer types are either 8 bit, 16 bit or 32 bit wide, and come in unsigned and signed forms.
An 8 bit unsigned pixel can hold values between 0 and 255 (28 − 1). By convention, 0 is assumed to correspond to the darkest possible black while 255 is assumed to correspond to the brightest possible white. Here is a memory layout of an 8 bit unsigned pixel.
A 16 bit unsigned pixel can hold values between 0 and 65535 (216 − 1). By convention, 0 is assumed to correspond to the darkest possible black while 65535 is assumed to correspond to the brightest possible white. However, since displays are not capable of displaying such a wide brightness range, usually window scaling or leveling techniques are used to display a subset or this range only. If the pixel is 16 bit signed it can hold values between -32768 ( − 215) and 32767 (215 − 1).
A 32 bit unsigned pixel can hold values between 0 and 4294967295 (232 − 1). By convention, 0 is assumed to correspond to the darkest possible black while 4294967295 is assumed to correspond to the brightest possible white. However, since displays are not capable of displaying such a wide brightness range, usually window scaling or leveling techniques are used to display a subset or this range only. If the pixel is 32 bit signed it can hold values between -2147483648 ( − 231) and 2147483647 (231 − 1).
Here is a little sketch that shows the available integer types. Follow the arrows to see how you can convert between integer types without a loss of information. Any conversion that is not following the arrows will lose information.
Two floating point types are commonly used: 32 bit and 64 bit. With both types 0.0 is assumed to correspond to the darkest possible black while 1.0 is assumed to correspond to the brightest possible white. Values below 0.0 are considered darker than the darkest black, and values above 1.0 are considered brighter than the darkest white. This may seem like a contradiction, but it actually works. A 32 bit float type contains, from right to left, a 23 bit mantissa, an 8 bit exponent and a 1 bit sign. A 64 bit float type contains, from right to left, a 53 bit mantissa, a 10 bit exponent and a 1 bit sign.
