Page 128 - Computational Colour Science Using MATLAB

TYPICAL CHARACTERIZATION PROCEDURE 115

7.5 Typical characterization procedure

Before making radiometric measurements for characterization the monitor
should be placed in the position where it will be used and then turned on (for
most monitors a deguassing process takes place whenever the monitor is initially
powered up). Sufficient time should be allowed for the monitor to warm up. The
warm-up time required for a monitor to stabilize after initial power-up varies for
different devices but can range from 15 min to 3 h or more (Berns et al., 1993b).
For accurate characterization it is important that the monitor exhibits good
spatial independence and channel independence. Spatial independence can be
assessed by measuring the colour difference between a white patch displayed in
the centre of the screen with a black surround and a physically identical white
patch with a white surround. Berns et al. (1993b) measured spatial independence
using this technique for five different monitors and found CIELAB colour
differences between 2.6 (for the best monitor) and 17.4 (for the worst monitor).
Channel independence can be assessed by computing the colour difference
between full-field white and the prediction of the full-field white obtained by
adding the tristimulus values of the full-field pure red, green and blue conditions.
Berns et al. found that the channel-independence error can be minimized by
reducing the maximum value of luminance that can be displayed. Experience
suggests that for many monitors a maximum display luminance of about 80 cd/m 2
provides a suitable compromise between being able to achieve good character-
ization and being able to display reasonable brightness levels.
For characterization purposes it is recommended that patches be displayed at
the centre of the monitor against a neutral field set at about one-fifth of the
luminance of the maximum brightness in order that the measurements are taken
in typical conditions. A spectroradiometer or spectrocolorimeter should be used
to measure the luminance and chromaticities of each of the calibration patches
[note that Equation (4.10) can be used to recover the tristimulus values from the
luminance and two of the chromaticity coordinates]. Three measurements are
needed to obtain the maximum tristimulus values of each of the guns. The digital
input values [d r d g d ] for these patches for a system with 8 bits per channel
b
should be [255 0 0], [0 255 0] and [0 0 255]. These three measurements should
be used to define the system matrix for Equation (7.5).
Measurements of as few as two neutral patches are then made in order to allow
the parameters of the GOG model to be computed but in practice normally
about five neutral patches are used (Luo, 2003). The tristimulus values of the
neutral samples are measured using a spectroradiometer and then Equation (7.5)
is inverted to predict the linearized normalized DAC values RGB. For each of the
neutral samples and for each channel the normalized DAC values and the
linearized normalized DAC values are then known, and therefore the GOG
parameters may be determined using a multidimensional optimization technique
such as the simplex algorithm.