Mapping Associations - Mapping Scientific Frontiers: The Quest for Knowledge Visualization - page 98

Fig. 3.6 Visualization of 279 images by color histogram

diagrams and visualization displays. Figures 3.7 and 3.8 show the screenshots of
two visualization models of the InfoViz image database by layout and by texture,
respectively. Both layout and texture similarities were computed by the QBIC
system.
The overall structure of the layout-based visualization is different from the color-
based visualization shown in Fig. 3.6 . This is expected due to the self-organizing
nature of the spring-embedder model. On the other hand, visualizations based
on the two schemes share some local structures. Several clusters appear in both
visualizations. The spring embedder algorithm tends to work well with networks of
less than a few hundreds of nodes.
Unlike the layout version, the texture-based visualization has a completely
different visual appearance from the color-based visualization. In part, this is
because the color histogram and color-layout schemes share some commonality in
the way they deal with color.
Now we compare the Pathfinder networks generated by different features
extracted from images. The number of links in each network and the number
of links in common are used as the basis for network comparisons. The degree
of similarity between two networks is determined by the likelihood that a number of
common links are expected given the total number of links in the networks involved.
Alternatively, one may consider use the INDSCAL method outlined later in this