In an image coding system that derives a luminance signal and two bandwidth-limited color-difference signals, constant luminance prevails if all of the luminance information is encoded into one signal that is supplemented by but totally independent of two color signals carrying only chrominance information, e.g., hue and saturation.

Constant luminance is only achieved when the luminance and chrominance vectors are derived from linear signals. The introduction of nonlinear transform characteristics (usually for better signal-to-noise and control of dynamic range prior to bandwidth reduction) before creating the luminance and chrominance vectors destroys constant luminance. Current video systems do not reconstitute the luminance and chrominance signals in their linear form before further processing and, therefore, depart from constant luminance. Note: When R, G, B information is required to be recovered from the set of luminance and color-difference signals, the values correlated to the original signals are obtained only if the luminance and chrominance signals have been derived from the linear functions of R, G, B or have been transformed back to linear. Constant luminance not only provides a minimum of subjective noise in the display (since the luminance channel does not respond to chrominance noise), but also preserves this noise minimum through chrominance transformations.