O. Kao, T. Bretschneider, and G.R. Joubert (Germany)
Dynamic Feature Extraction, Parallel Image Retrieval,Cluster-Based Image Database
Image retrieval with dynamically extracted features com pares user-defined regions of interest with all sections of the pictures stored in an image database. Thereby, objects regardless of the actual pictorial environment are of inter est. This paper discusses an application of Gabor-Wavelet Networks as a solution of the object-based search. In con trast to existing template matching methods deviations re garding scaling and rotation can be partly compensated. However, the improved retrieval quality and flexibility re quires large computational resources, which are satisfied by the developed cluster-based architecture. The retrieval quality and the obtained speedup are examined by a num ber of experiments.
Important Links:
Go Back
