Skip to main content
. Author manuscript; available in PMC: 2011 Sep 10.
Published in final edited form as: IEEE Trans Vis Comput Graph. 2010 Sep-Oct;16(5):763–776. doi: 10.1109/TVCG.2009.107

TABLE 2.

Comparison between OFB and Octree-Based Surface Paintings on the GPU

Construct (ms) Paint (fps) Memory (MB)
Octree (8/9/10) OFB (2562/5122/10242) Octree OFB Octree OFB
Bunny 4934 / 15732 / 56445 3.0 / 4.5 / 7.0 25/12/3 360 / 343 / 309 4.25 / 16.1 / 52.4 6.75 / 27 / 108
Horse 4002 / 9818 / 31557 3.6 / 5.2 / 8.0 32/18/5 426 / 416 / 401 1.5 / 5.8 / 24.3 5.62 / 22.5 / 90

The GPU octree implementation is publicly available at http://lefebvre.sylvain.free.fr/octreetex/. The table shows the performance (including painting and rendering) and memory requirements for octrees of depths 8, 9, and 10, and respective OFBs with sampling grids of sizes 2562, 5122 and 10242. The first row gives the times for OFB and octree construction. The memory requirements of the octree implementation include only the color data stored at the leave nodes of the data structure. The additional memory required by the index structures could not be measured with the given implementation.