Using Graphics Hardware for Multiple Datasets Visualization

Abstract

We have applied three graphics hardware-based approaches to support concurrent visualization of multiple sets of volumetric scalar data. They include volume rendering, clipping and isosurface extraction methods, which exploit 3D textures and advanced per pixel operations. These methods are expected to give better interactive frame rates for multiple datasets visualization (MDV) compared to the software-based methods. The rendering time in each case increases nonlinearly with the increasing the number (N) of the datasets being visualized. We can identify three regimes, which can be characterized by different time-N slope value. The first regime with small slope value continues up to about 5 datasets, then the second regime with medium slope value continues up to about 25 datasets, and finally the third regime with much larger slope value continues up to 35 datasets. With volume shading enabled, the rendering time increases on average whereas the transition and maximum N values decrease. We propose the dynamic-resolution approach for increasing the maximum N and frame rates for above MDV techniques.