Digital Signal Processing Reference
In-Depth Information
Fig. 2.24
Segmented images
obtained by the conventional
FCM applying on the KL
transformed color spaces:
(
a
) Mosaic; (
b
) Ball;
(
c
) Akiyo; (
d
)News
sequences
Fig. 2.25
Segmented images
obtained by the OSFCM
method: (
a
) Mosaic; (
b
)Ball
(
c
) Akiyo; (
d
)News
sequences
extracted. Similar to the PCT approach, the OSFCM method does not separately
use the noise-subspace. The color objects with high signal-subspace projects will be
erroneously included in the similar color pixels.
Figure
2.26
shows the results obtained by the SEFCM. We can find out that
most of the noise has been removed compared to the previous obtained results. The
major defeat of the SEFCM method appears on the clothes of Akiyo since both
signal and noise subspace projections do not perform simultaneously. In order to im-
prove the performance, the CEFCM algorithm adopts the signal and noise subspace
