IsoBAS

This paper introduces a binary accelerating structure, IsoBAS, to support GPU-based full-resolution real-time isosurface rendering for large datasets with dynamically varying isovalues. By looking up this structure during the ray-tracing stage, the algorithm can rapidly locate active cells along rays. The experimental results demonstrate that the proposed algorithm is much faster (up to 28 times) than current state-of-the-art algorithms. Furthermore, it has a much smaller memory footprint, now enabling GPUs to process large datasets on which previous GPU algorithms would fail due to a lack of available memory. To the authors¿ knowledge, this is the fastest dynamic isosurface rendering algorithm ever reported for operation on a single computer. Graphical abstractDisplay Omitted HighlightsA new GPU isosurface rendering algorithm by using a binary spatial-occupancy structure.A multi-GPU implementation is presented which can avoid transferring the depth buffers across the multiple GPUs.Very smaller memory footprint compared to previous GPU isosurface algorithms.We performed performance comparisons with state-of-the-art algorithms.