Exploiting the Spatial Distribution of Interest Points Using Hierarchical Clustering for Improved Scene Retrieval

Sebastian Palacio Bustamante

Mastersthesis, TU Kaiserslautern - Computer Science, 8/2013.


Today's methods for doing content-based image retrieval mainly rely on the information provided by local interest points without taking into account their underlying spatial relationships. These links convey important information to determine what is represented within an image. This thesis demonstrates how the spatial information can be effectively conveyed using a hierarchical agglomerative clustering (AHC) based on the spatial proximity of interest points. Spatial relationships are hence ensured by clustering locally adjacent interest points together, albeit not forcing any particular arrangement upon them. This approach provides robustness to affine and scale transformations. Moreover it can be used independently of any feature description algorithm. The AHC allows the improvement of performance in a CBIR system while keeping processing and memory usage almost unaffected. During experiments, a performance increase of up to 12,47% with respect to a reference system using bag of features was achieved. The AHC also proved to be a pyramid of grid partitions over the surface of an image as well as a hierarchical random clustering of interest points.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence