Image indexing with caterpillar decomposition

Given a large image database and query image, efficiently finding all similar database images is an important problem facing researchers in computer vision. This paper proposes an approach for indexing image databases, where images can be represented as tree structures. The algorithm is based on graph-theorical concept named caterpillar decomposition which has been used in the literature for graph embedding. The experimental evaluation of the proposed approach and its comparison with the previous indexing algorithms demonstrates its overall efficacy for image retrieval.