A framework for Mixed Symbolic-based and Feature-based Query by Example Image Retrieval

This paper defines and studies the use of query by example (QBE) in the context of photograph retrieval. The novelty of our approach lies in considering an automatic indexing process of photographs as well as a representation of the features of image regions in a single knowledge representation formalism. Both symbolic and feature based representation are used during the query by example process. More precisely, the QBE process is able to take into account the symbolic descriptors of the images but also the extracted features from image under the form of histograms. The aim of this process is to detect the relative importance of the symbolic elements and of the feature elements according to the user's query by example. We experiment the query by example process on two collections of a total of 1100 photographs. The precision measures we have obtained are as good as a baseline defined as explicit textual queries processing. Keyword: Symbolic Image Indexing, Image Retrieval, Conceptual Graphs.