Input vs. Output Referred Images

A lot of Photography related sources recommend that images be encoded in a large gamut colorspace such as ProPhoto, but don't explain the implications of doing this in in the process of subsequently displaying such photo's.

Input referred:

Images are encoded in a way that represents their unchanged or originally captured values, in an encoding space that is larger enough to store their gamut without clipping. The gamut of the encoding space tells you nothing about the gamut of the images.

Typical: L*a*b*, Raw, ProPhoto, RIMM etc.

Output referred:

Images are modified (i.e. rendered) to fit within the gamut of a specific real world output device (such as display or printer). This means that typically the encoding space is a good representation of the gamut of the images.

Typical sRGB, AdobeRGB, printer profile, display profile, etc.

Before displaying images that are Input Referred, they need to be rendered to a smaller gamut. This may be done manually by adjusting the images carefully to fit within the smaller gamut, or in some automatic fashion such as by hard clipping them to the smaller gamut, or by setting up a specific gamut mapping for each image or set of images that occupy a similar gamut.

Note that setting up a gamut mapping from the very large Input referred encoding gamut to the smaller output device gamut will almost certainly result in a disappointing loss of saturation, because the images deliberately do not occupy the large encoding gamut, and so get unnecessarily squashed down to allow for colors that they do not actually contain.

See also Scenarios.html#LP3.