Visual Information Management Systems

This paper discusses various questions connected with visual information management, a database tool, specially designed to store and organize a large collection of digitized images, and delete images, and search for images using content-based visual information retrieval. Ideally, a visual information management system (VIMS) provides access to the images both by any text information associated with the image (such as the date of the image or its title), also called metadata, as well as through their content - i.e., trough their visual features that can be computed using computer-vision and image-processing techniques (examples are color histograms, shape measures, and texture measures). Queries by example are the most suitable for VIMS: user provides an entire image as a template and the system finds images that are visually similar. As VIMS computes image features, the main task in processing a similarity query is to compare these features and come up with some measure of how they differ. Once the distance between the template image and each target image is computed, target images canbe ranked in order of their similarity with the template image. Object recognition and scene understanding are still research areas, so systems using these technologies can be built within only narrow domains.