A compact feature representation ad feature re-weighting in content-based image retrieval

摘要

We describe a compact feature representation based on the elements of Colour Co-occurrence Matrices (CCM) in HSV space. The feature vector consists of all diagonal elements and one representative value for all non-diagonal elements of the CCM. To reduce semantic gap we use Relevance Feedback, and describe how the integration and synergy of existing feature re-weighting techniques can lead to better retrieval accuracy. We experimented with three image databases of sizes 2000, 2020 and 8365, and having number of semantic categories 10, 12 and 98 respectively. The results demonstrated the superiority of our feature representation and feature re-weighting method in terms of retrieval accuracy and online computation time.