Digital Signal Processing Reference
In-Depth Information
% Take inverse curvelet transform
imrdctg2 = real(ifdct_wrapping(Ct,1,J-coarsest));
subplot(223)
imagesc(imrdctg2);axis image;axis off
set(gca,'FontSize',14);
title('(c)');
Figure 5.18(b) depicts both the DCTG2 and UWT coefficients in decreasing or-
der of magnitude. This image contains many highly anisotropic structures. The supe-
riority of DCTG2 over UWT is then obvious. The reconstruction from the 2 percent
DCTG2 largest coefficients is given in Fig. 5.18(c). The difference with the original
image is visually weak.
5.6.3 Denoising Using DCTG2
Suppose that we want to restore a piecewise regular image (away from regular con-
tours) contaminated by additive white Gaussian noise of standard deviation
σ
.A
(b)
(a)
14
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
12
+
10
+
8
6
UWT
DCTG2
4
++
2
6
8
10
12
14
16
18
20
(c)
log
2
(Index of retained UWT coeff)
Figure 5.18. (a) Original “Barbara” image. (b) The DCTG2 and UWT coefficients in
decreasing order of magnitude. (c) Reconstruction from the 2 percent DCTG2 largest
coefficients.
