Binary Composition Tree Hardware: A Scalable Solution for Real-Time Volume Rendering

Recent advances in PC graphics hardware like the VolumePro 500-2X and GeForce 3 have resulted in remarkable volume rendering performance. However, the limited amount of video memory found on a single card prohibits its use in large-scale volume graphics applications. The trend has thus been to make use of the combined rendering power and storage capacity of multiple graphics cards to achieve interactive rendering of large volume data. Such a parallel rendering approach typically requires a final image compositing step, which if done in software, can quickly become a bottleneck of the overall rendering performance as more cards are used. In this paper, we present the design of Binary Composition Tree Hardware (BCTH), its integration into a Volume Graphics (VG) cluster prototype equipped with both VolumePro 500-2X and GeForce 3 cards, and test results using large medical volume data sets. Our test results show that BCTH makes cluster rendering a viable approach to the large scale volume rendering problem.