Progressive content-based retrieval from distributed image/video databases

We describe the architecture and initial implementation of a content based retrieval mechanism from heterogeneous image archives. In particular, we propose an architecture to produce local representation of the images stored in heterogeneous archives and a progressive framework that reorganizes the image into a hierarchical representation both in the pixel domain and in the feature space. Consequently, search operations can be performed in a hierarchical fashion and thus significantly reduce the total amount of data that need to be processed. Dramatic speedup has been achieved for many search operations such as template matching, texture feature extraction, and histogram extraction.