Stereoscopic volume rendering provides powerful depth information, but if f
akes a long time to render two-eye images, Previous algorithms based an rep
rojection methods project the result of one view of a stereo pair into the
at her instead of rendering a new one completely. Because of inaccurate map
ping between the two images, the quality of the reprojected image is not sa
tisfactory. This paper presents a new algorithm to preserve the accuracy of
both images with very little increase in computation time. The efficiency
of the new algorithm conies fro the rise of ray templates and object-order
processing. This algorithm makes two different templates for each eye and r
enders two images simultaneously, tracing the volume only once in object or
der We also extend the algorithm to support image-space supersampling by us
ing more my templates. Experimental results show that the image qualify of
the new algorithm is not only comparable with that of ray casting but also
the rendering speed is near that of the interactive shear-warp algorithm em
ploying object-order processing and spatial data coherency. Copyright (C) 1
999 John Wiley & Sons, Ltd.