1996 | ReviewPaper | Buchkapitel
Quantitative measurement of the performance of raster-to-vector conversion algorithms
verfasst von : Osamu Hori, David S. Doermann
Erschienen in: Graphics Recognition Methods and Applications
Verlag: Springer Berlin Heidelberg
Enthalten in: Professional Book Archive
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
This paper presents a methodology for measuring the performance of application-specific raster-to-vector conversion algorithms. In designing and building image analysis systems, comparison of several algorithms is often required. Unfortunately, many methods of comparison do not give quantitative performance measurements, but rather qualitative, and often subjective, evaluations. Our key observation is that there is a need for domain, or task-dependent evaluation of the output. By specifying the input data in the same parameter space as the intended output of the system, we are able to evaluate the quality of the output and how well it conforms to the intended representation. We provide a set of basic metrics, but we emphasize that in general, such metrics may be task-specific. In this paper, the performance of three approaches to raster-to-vector conversion — thinning, medial line finding, and line fitting — are compared using this methodology.