Visual Database Organization in the Reduced Biplot Domain via the Appending Technique

C. Theoharatos, N.A. Laskaris, G. Economou, and S. Fotopoulos (Greece)


Database organization, appending technique, Wald Wolfowitz test and minimal spanning tree.


A common problem in image libraries is the structural organization into semantically relevant classes, for efficient navigation and retrieval practices. This paper proposes a novel method for image database organization directly in a reduced low-dimensional space. The management of a generic image database is pursued using a well-suited data analysis methodology that is based on simple geometrical concepts. In the core of our proposal lies the appending technique, an efficient geometrical scheme that inserts any given query image directly in a reduced MDS-derived space. The main novelty of our method is that it is embedding-based and reduces the computational complexity of the proximity data from the original high-dimensional space.

Important Links:

Go Back