Digital Signal Processing Reference
In-Depth Information
Figure 5.16. (left) Grayscale image. (right) Curvelet-enhanced image.
=
Compute the magnitude of
the color
curvelet
coefficient as
a
j
,,
k
(
L
u
α
j
,,
k
)
2
.
α
j
,,
k
)
2
+
(
α
j
,,
k
)
2
+
(
Compute the modified coefficients
˜
α
j
,,
k
L
j
u
j
α
,
α
˜
,
˜
,,
k
,,
k
=
α
σ
j
,
)
.
j
j
,α
j
,,
k
E
E
(
|
a
j
,,
k
|
;
σ
j
,
)
,α
E
(
|
a
j
,,
k
|
;
σ
j
,
)
(
|
a
j
,,
k
|
;
,,
k
,,
k
L
j
u
j
α
j
,,
k
.
Compute the inverse curvelet transform of each of ˜
α
k
,˜
α
k
, and ˜
,,
,,
Convert from Luv to the original color space.
It may happen that the values in the enhanced color components can be larger than
the authorized dynamic range (in general, 255), and it may be necessary to incorpo-
rate a final step to this method, which is a gain/offset selection applied uniformly to
the three color components, as described by Jobson et al. (1997).
Figure 5.16 shows the results for the enhancement of a grayscale satellite image
of Marseille port. The parameters used were (
3).
Figure 5.17 shows a part of the Saturn image, the histogram equalized image,
the enhanced image based on the (Laplacian) sharpening method, and the curvelet
multiscale edge enhanced image (with parameters (
γ,,ρ,τ
)
=
(0
.
5
,
0
,
0
.
5
,
3)). The
curvelet multiscale edge enhanced image reveals much better the rings and edges of
Saturn, while controlling the noise.
γ,,ρ,τ
)
=
(0
.
5
,
0
,
0
.
5
,
5.6 GUIDED NUMERICAL EXPERIMENTS
5.6.1 Software
The pedagogical numerical experiments of this chapter are conducted using both
WaveLab and CurveLab toolboxes. The WaveLab toolbox was introduced in Sec-
tion 2.9.1 of Chapter 2. CurveLab is a collection of C files, Mex MATLAB code,
and MATLAB routines that implement the second generation curvelet transform.
