Image Processing Reference
corrected RGB intensities and is referred to by the symbol Y'. Note the prime mark that
indicates it is a nonlinear property.
Charles Poynton describes at great length the historical confusion between luminance
and luma. Refer to page 595 of his 2003 book Digital Video and HDTV Algorithms and
Interfaces (published by Morgan Kaufmann) for further elucidation on this. Mr. Poynton
also suggests that the use of the term luma is usually appropriate for discussing this
quantity in digital video and computer graphics systems.
This is the primary form in which video content is represented. Because the human eye
responds to these colors individually, it is a particularly appropriate way to represent images.
The eye is most sensitive to the green channel, less so to the red, and least of all to the blue.
Viewers suffering from color blindness may respond to these colors in a different way.
One possible compression approach, then, is to store the green channel at full reso-
lution, remove some detail in the red, and remove even more in the blue.
The popular compression schemes developed by the MPEG standards body don't
operate on the RGB value but on the component luma and chroma values, which are read-
ily available in the video-production environment.
Color Difference Channel
Separating the RGB values into a brightness (luma) value and two color components
requires a bit of math.
The RGB values are processed using a matrix calculation to derive a single luma
value that presents the most accurate monochrome signal. The matrix values are selected
so that displaying just the luma on a black and white monitor will present a pleasing and
appropriately lit image.
Take the blue value and subtract the luma from it to generate a blue-difference chan-
nel. This is also done with the red to make a red-difference channel. The result is a pair of
Along with knowing a lot about how the human visual system works, the engineers
who developed this technique also knew that reducing the resolution of these two color-
difference signals would have little perceived effect on the reconstructed image when
viewed under normal conditions.
Table 5-7 describes some common terminology that is related to this topic. Note that
the Y' value always describes luma and not luminance.
An Experiment to Try at Home
Prove the importance of luma over chroma experimentally like this. First load an image
into Photoshop and go to the channels palette. Apply varying degrees of Gaussian blur to