Partial Order Structure Based Image Retrieval

Abstract

Image retrieval plays an important role in the growing computer vision applications. The computation of the unrelated images in large scale image retrieval task seriously reduces the retrieval efficiency. In this paper, a new Partial Order Structure (POS) based image retrieval method is proposed. Partial order structure diagram is an effective visualization tool in Formal Concept Analysis (FCA) theory, including object partial order structure diagram and attribute partial order structure diagram. There are two contributions in this paper. First, we design an association rule according to the object partial order structure (OPOS) method to measure the correlation between the query image and the database, and then improve the database to be retrieved. Second, we perform a query expansion according to the attribute partial order structure (APOS) method to improve the generalization ability of the query information. Experimental results on two databases verify the effectiveness of the proposed algorithm.