Information Technology Reference
In-Depth Information
(a)
Dendrogram for CBR on marginal distri-
bution for
column
(b)
Dendrogram for CBR on marginal distri-
bution for
row
(c)
Dendrogram for CBR on marginal distribu-
tion for
column and row together
(d)
Dendrogram for CBR on marginal distri-
bution for
diagonal
Fig. 34.
Dendrogram for CBR on marginal distribution using the similarity measure city block
distance
In conclusion, the marginal distribution using the Euclidean Distance enables us to
separate rotated images, but we did not get images having the same settings in the
same group. We got similar results for the marginal distribution using the Chebyshev
Distance (compare with Fig. 33) and the city block Distance (compare with Fig. 34).
In conclusion we have to say, that the marginal distribution is a nice method to
summarize the image content in x- and y-direction as well as over the diagonal, but it
does not help us in properly grouping images for watershed CBR segmentation.
5.3 Image Description by Similarity between the Regional Minima of Two
Images
Obviously the position of the regional minima has an influence on the segmentation
result. Thus, we study the starting-points of the Watershed Transform (regional
Search WWH ::
Custom Search