Image Processing Reference
In-Depth Information
control as compared to 1-D channel-wise or 1-D gray-balance TRCs and hence can
provide more accurate color rendition than 1-D TRCs, but are still less accurate than
3-D LUTs [5].
8.5 ONE-DIMENSIONAL AND TWO-DIMENSIONAL PRINTER
CALIBRATION USING PRINTER MODELS
In this section, we show how to construct the 1-D and 2-D printer calibration TRC
LUTs if an updated printer model is given. The model can be in any of the forms
described in Chapters 7 or 10 or even a simple input
output LUT obtained experi-
-
mentally while characterizing the printer. We
first discuss the algorithms for 1-D
calibration, which involves both 1-D channel-wise linearization and 1-D gray
balance to equivalent neutral targets.
8.5.1 O
NE
-D
IMENSIONAL
C
HANNEL
-W
ISE
(I
NDEPENDENT
)C
ALIBRATION
In channel-independent calibration, each channel is independently linearized. By
linearization, we would like to make the
D
E
a
*
obtained by measuring color patches
from paper to be linear so that the channel-linearized printer can emulate an ideal
printer, which has the characteristic of linearized
D
E
a
*
from
paper is the Euclidean norm between target color and the paper white in the device-
independent color space (L*a*b*). The
D
E
a
*
from paper. The
D
E
2000
metric is not normally used for the
paper-based 1-D channel-wise calibration.
As an example, let us assume that we choose the cyan channel for linearization
and printed N cyan patches (M
¼
Y
¼
K
¼
0) that represent the step wedge with the
cyan digital count of d
0
¼
255, and measured the L*a*b* values
of each cyan patch using a color sensor. Let the corresponding measured L*a*b*
value of the ith patch be (L*a*b*)i.
i
. Note the
0, d
1
, d
2
,...,d
N
1
¼
first patch is the paper white which
means that (L*a*b*)
0
is the L*a*b* of the paper. We now compute the
D
E
a
*
from
paper for each patch as follows:
D
E
i
¼k(
Lab
)
i
(
Lab
)
0
k
i
¼
0, 1,
, N
1
(
8
:
1
)
...
The function f(.) is formed by normalizing the
D
E values to obtain a value of 255 for
the cyan value of 255. It is de
ned as
D
E
i
max
i
D
E
ðÞ
255
d
i
!
fd
ðÞ¼
(
8
:
2
)
To linearize the printer, we need
D
E
i
to be a linear function of di.
i
. To achieve this, we
find the inverse of the transformation given by Equation 8.2, which is the TRC of the
cyan channel. That TRC can be represented by following equation:
c
¼
f
1
Cyan TRC
:
(
C
)
(
8
:
3
)
Search WWH ::
Custom Search