Combinaçao de técnicas de interaçao humano-computador para avaliaçao de sistemas de recuperaçao de imagens por conteúdo

The human interaction is a valuable tool to evaluate the performance of a Content-based Image Retrieval (CBIR) system regarding its efficiency and efficacy, as well as the user satisfaction. The fast growing and the strength achieved in the CBIR research field has brought relevant improvement to the pictorial information retrieval, but it is not supported by techniques or guidelines to help evaluating the usability of such systems. Thus, a challenge of CBIR systems is to provide the users intuitive ways to ask queries and to receive their answers increasing the usefulness of such systems. The goal of this paper is to present an approach to evaluate the human-computer interaction of CBIR systems. To do so, we employed a combination of evaluation techniques considering the final users and the human factors that influence their work, taking also into account the large amount of images and data stored in these systems. As a case study, we used a CBIR system for medical images, which was developed aimed at supporting the decision making when the physician is drawing the diagnosis. Our study aimed at improving the diagnoses, as well as the evolution of the patients' treatment, making both more precise and efficient.