Ontology-Based Image Annotation by Leveraging Social Context

Abstract

Manual image annotation is an extensive and a cumbersome task, yet extremely important for image management and retrieval. The purpose of the authors’ system is to semi-automatically generate ontology-based annotations for a social network by leveraging the annotations provided by the most active user (i.e., the central actor). Context of an image is of central importance in their approach towards semantic semi-automatic annotation. For context of an image, the authors consider several factors like geo-reference, time and relationship among actors in social networks and instead of using image-processing techniques to manipulate and interpret the image, their system leverages the context, which is automatically available along with the image and have also extended Social Network Analysis techniques by considering the granularity of relationships among actors under consideration. The authors use a semantic web framework to represent the social network and to deal with the diversity of relationships. OntoCAIM ontology is developed which not only encompasses Social Network Analysis functionality but also defines mechanism to annotate the images with an underlying ontology.