ORF-NT: An object-based image retrieval framework using neighborhood trees

Abstract

This study proposes an object-based image retrieval framework, called, ORF-NT, which trains a discriminative feature set for each object class and introduces a neighborhood tree for object labelling. For this purpose, initially, a large variety of features are extracted from the regions of the presegmented images. These features are, then, fed to a training module to select the 'important' features, suppressing relatively less important ones for each class. ORF-NT (Object-based Image Retrieval Framework using Neighborhood Trees) defines a neighborhood tree for identifying the whole object from oversegmented regions. The neighborhood tree consists of the nodes corresponding to the neighboring regions as its children and merges the regions through a search algorithm. Experiments are performed on Corel database using MPEG-7 features in order to observe the power and the weakness of ORF-NT. The training phase, is tested by using Fuzzy ARTMAP [1], Euclidean distance and Adaboost algorithms [2]. It is observed that Fuzzy ARTMAP yields better retrieval rates than Euclidean distance and Adaboost algorithms. © Springer-Verlag Berlin Heidelberg 2005.