Digital Signal Processing Reference
In-Depth Information
Original Barbara
MCA Barbara
Barbara Cartoon
Barbara Texture
Figure 8.10. Decomposing the “Barbara” image. See Table 8.2 for the experimental details.
Barbara
We also applied the MCA separation, Algorithm 31, to the “Barbara” image. We
used the curvelet transform for the cartoon part and a local DCT transform with
a smooth sine window of size 32
32 for the locally oscillating texture. Again, we
used Algorithm 32 to approximate the TV regularization. The parameter
×
was set
to 2. Figure 8.10 displays the “Barbara” image, the recovered cartoon component,
and the reconstructed texture component.
γ
8.6.3.2 Applications
The ability to separate the image into different contents, as we show, has many
potential applications. We sketch here two simple examples to illustrate the impor-
tance of a successful separation.
Edge Detection
Edge detection is a crucial processing step in many computer vision applications.
When the texture is highly contrasted, most of the detected edges are due the
