Automatic image description based on textual data

In the last two decades, images are quite produced in increasing amounts in several application domains. In medicine, for instance, a large number of images of various imaging modalities (e.g. computer tomography, magnetic resonance, nuclear imaging, etc.) are produced daily to support clinical decision-making. Thereby, a fully functional Image Management System becomes a requirement to the end-users. In spite of current researches, the practice has proved that the problem of image management is highly related to image representation. This paper contribution is twofold in facilitating the representation of images and the extraction of its content and context descriptors. In fact, we introduce an expressiveness and extendable XML-based meta-model able to capture the metadata and content-based of images. We also propose an information extraction approach to provide automatic description of image content using related metadata. It automatically generates XML instances, which mark up metadata and salient objects matched by extraction patterns. In this paper, we illustrate our proposal by using the medical domain of lungs x-rays and we show our first experimental results.