Image Processing Reference
In-Depth Information
TABLE 6.15
D E Statistics for Downsampled LUTs of Different Sizes
Reduced
Measurement
Colors
Error
Metric Maximum D E Mean D E
95th Percentile D E
D E a *
27
8.56
3.00
4.92
D E 2000
6.78
1.27
4.56
D E a *
48
7.90
1.77
3.51
D E 2000
4.42
0.96
3.45
D E a *
125
7.53
1.37
3.0
D E 2000
4.08
0.76
3.0
D E a *
216
7.41
1.24
3.08
D E 2000
4.13
0.74
3.05
D E a *
343
7.41
1.20
3.19
D E 2000
4.14
0.70
2.61
D E a *
512
7.37
1.1
3.17
D E 2000
4.11
0.65
2.48
D E a *
4913
7.30
0.99
3.12
D E 2000
4.18
0.57
1.94
TABLE 6.16
D E Statistics for Different LUT Sizes
Downsample LUT Size Mean D E a * Maximum D E a *
95th Percentile D E a *
27
3.11
8.84
4.69
64
2.09
8.94
3.93
125
1.46
8.38
3.07
216
1.18
7.05
3.29
2197
1.03
8.80
3.12
be small. We use the second derivative as a measure of the smoothness of
fit. If no
restriction is made on how smooth the
it must be, then it ends up matching the data
exactly. This will create problems if the data is noisy, since we must not
t to the
noise in the data. On the other hand, if the
it is too smooth, then it might not
approximate the data very well. To
fit, both of these factors must be
considered. In this section, we consider the problem of smooth curve
nd the best
fitting in 1-D
[13] through 4-D cases. As an application, the proposed algorithm can be applied to
digital printer calibrations. In the 1-D case, it can be applied to the gray balance
calibration [14,15], and for the 2-D case, it can be applied to the 2-D printer
calibration [16]. Finally, as a 3-D example, we use the 3-D or 4-D profiling systems.
Section 6.6.2 presents the 1-D through 4-D smoothing algorithm. Section 6.6.3
describes applications to printer calibrations with simulation results.
Search WWH ::




Custom Search