Image Processing Reference
In-Depth Information
linearize each separation independently, and, therefore, do not always yield a neutral
gray balance for C ¼ M ¼ Y.
Since the TRCs (gray-balanced or 1-D channel-wise linearized) contain all the
possible entries and are independent of each other, each separation can be processed
on a real-time basis without interpolation after RIPping. If a color sensing instrument
becomes available, then the construction
generation of the gray-balance TRCs and
processing of each pixel with the updated TRCs can also be used as a means to
improve the stability of neutral colors. Whereas, when real-time processing is applied
with 1-D channel-wise TRCs, we can only stabilize colors along the separations
(not neutral colors, such as C ¼ M ¼ Y). Hence, a system can be engineered
to automatically gray balance the print engine and to automatically linearize
the separations independently to overcome environmental and other disturbances
that cause color variations on paper. When 1-D channel-wise linearizations are used
in conjunction with a gray-balance calibration, proper separation rules have to be
applied while creating and using the TRCs to avoid undesirable loop interactions.
=
8.4 TWO-DIMENSIONAL CALIBRATION
With 1-D gray-balanced TRC LUTs, as described in Section 8.3, we can achieve
color balance to neutral colors by constantly monitoring the colors along the neutral
axis (i.e., the L* axis in the L*a*b* color space) and producing color-balanced
TRCs. This will provide good control along a sensitive color axis. Although, this
approach can make other colors reasonably accurate, it does not control colors
everywhere in the color space like the 3-D color control approaches described in
Chapter 7. Accordingly, one might ask the following question: Why not just do 3-D
color control more frequently to stabilize the color output of engines instead of
simply updating gray-balance TRCs on a more frequent basis? There are
three reasons why this strategy is not preferred: (1) processing of 3-D LUTs for
high-speed RIPping is not cost effective when compared to processing 1-D gray-
balance TRCs, (2) memory requirements are very minor for 8-bit TRC processing
(256 bytes of memory per separation TRC), and (3) the number of levels that can be
corrected in a 3-D LUT is limited along each separation axis to a much smaller
number (e.g., 33 in a 33 3 pro
le LUT) as compared to 256 levels in the 1-D TRCs,
thus limiting the quality levels for rendering neutral colors. A full resolution 3-D
LUT with full lookup can avoid the shortfall in levels, but can be prohibitively large
(a full 3-D LUT size for a 8-bit system would be 3
(256) 3 bytes
49.75 M Bytes
of storage), especially when customized for each media and halftone screen.
Gray balance alone can introduce undesirable hue shifts in the reproductions
of sweeps from white to secondary colors, such as those colors lying along the red
axis (M ¼ Y and C ¼ K ¼
¼
0) as well as those along the blue and green axes. On the
other hand, with a 2-D calibration [4], we can achieve gray balance as well as control
over colors lying on other critical axes in the color space. In Section 8.6.3, the 2-D
calibration approach is discussed. In this case, the TRCs are 2-D and the LUTs
representing each TRC would have 256 2
65,536 entries, quite reasonable as
compared to full resolution 3-D LUTs. Obviously, as we try to control more axes
in the color space, the LUT size increases. The 2-D TRCs offer signi
¼
cant levels to
Search WWH ::




Custom Search