An Approach to Interactive Retrieval in Face Image Databases Based on Semantic Attributes

for RSA based on the repertory grid. The algorithm incorporates user relevance judgments as a means to deal with the inherent problems associated with the specification of semantic attributes. The algorithm is implemented and tested on the human face image database and the initial results are encouraging. In essence, we have developed an overall methodology/test bed to facilitate experimentation with different algorithms for RSA. Image Retrieval (IR) problem is concerned with retrieving images that are relevant to users' requests from a large collection of images, referred to as the image database. A taxonomy for and the limitations of the existing approaches for image retrieval are discussed in [1]. Also, to alleviate some of the problems associated with these approaches, a unified framework for retrieval in image databases for a class of application areas is proposed in [1]. The framework provides a taxonomy for image attributes and identifies four generic types of retrieval based on the attribute taxonomy. Semantic attributes play a central role in supporting one of those generic retrieval types, referred to as Retrieval by Semantic Attributes (RSA). Semantic attributes are those attributes the specification of which necessarily involves some subjectivity, imprecision, and/or uncertainty.