Affiliations: L3I, Université de La Rochelle, France; IFI, MSI team; IRD, UMI 209 UMMISCO; Vietnam National University, Vietnam. {hien_phuong.lai, muriel.visani, jean-marc.ogier}@univ-lr.fr, alainboucher12@gmail.com
Note: [] Address for correspondence: L3I, Université de La Rochelle, France
Abstract: The feature space structuring methods play a very important role in finding information in large image databases. They organize indexed images in order to facilitate, accelerate and improve the results of further retrieval. Clustering, one kind of feature space structuring, may organize the dataset into groups of similar objects without prior knowledge (unsupervised clustering) or with a limited amount of prior knowledge (semi-supervised clustering). In this paper, we present both formal and experimental comparisons of different unsupervised clustering methods for structuring large image databases. We use different image databases of increasing sizes (Wang, PascalVoc2006, Caltech101, Corel30k) to study the scalability of the different approaches. Then, we present a new interactive semi-supervised clustering model, which allows users to provide feedback in order to improve the clustering results according to their wishes. Moreover,we also compare, experimentally, our proposed model with the semi-supervised HMRF-kmeans clustering method.
