Out-of-core rendering of large volumetric data sets at multiple levels of detail

Abstract

Advances in equipments and techniques for image acquisition have contributed to the availability of massive high-resolution data volumes. Several fields of knowledge have benefited from these technological improvements, such as medicine, geology, biology, fluid dynamics, remote sensing and surveillance, among others. For instance, computed tomography, ultrasonography and magnetic resonance imaging are commonly employed in non-invasive medical diagnosis. More recently, X-ray microtomography imaging techniques have allowed for higher resolution images. The visualization of such large volume data sets using traditional in-core volume rendering has serious limitations, since all data may not fit in the computer's primary memory. To address such a problem, this work presents an architecture for out-of-core volume rendering at multiple levels of detail. Experiments conducted on several data volumes demonstrate the effectiveness of the proposed approach in terms of memory storage and computational time required in the rendering process, signal-to-noise ratio measured at each level of detail for the rendered volumes as well as frame rate during the user's interaction.