Database Reference
In-Depth Information
Fig. 6.9
Examples of screen shots captured from the proposed system, obtaining the correlation
between the query image and the target image. The parameters estimated by the registration
algorithm are: ʸ
0
=
180
ⓦ
,
ʱ
=
1
,
t
x
=
0
,
t
y
=
0. The ECAs of both images are clearly extracted.
This results in a CCF score of 0.85
observed that the system correctly retrieved eight out of ten classes from the top
five best matches (e.g.
R
8) by using the ECA method. In comparison,
Method 1 required the top sixteen best matches (e.g.,
n
(
5
)=
0
.
16) to attain this level
of performance. This high performance of the ECA method is very important for a
retrieval system with a high volume of databases.
Figure
6.9
shows examples of screen shots captured from the system to obtain the
correlation between the query image and the target image. Both were in the same
class. The target image was aligned with the query using the image registration
algorithm. The estimated parameters were: rotation
=
180
ⓦ
, scaling (
(
ʸ
0
)=
ʱ
)=
1,
and translation
. The ECAs of both images were extracted, which
covered the area where edge density and the cross-covariance function were greater
than the predefined threshold values. It can be observed from the result that ECAs
are clearly extracted from both images. This results in a CCF score of 0.85, which
is much higher than that of Method 1, at 0.36.
(
t
x
,
t
y
)=(
0
,
0
)
Search WWH ::
Custom Search