Click slide for next, or goto previous, first, last slides or back to thumbnail layout.
Click slide for next, or goto previous, or back to thumbnail layout.
This slide shows the process a few levels deeper. In the last image the diameter of the cubes is less than 1% of the original cube size. If the result is not accurate enough, for example not all the holes that we want to get have been carved yet, we can easily continue without having to recalculate the earlier results.
Obtaining the finest level results shown in this slide takes about 35 seconds on an SGI O2 workstation. The running time of this algorithm is clearly exponential in the number of subdivision levels, therefore one extra subdivision after what's displayed in the last image takes about two more minutes.
The data used for this model consists of eight views of a miniature chair, and it was provided by Prof. Patrick Flynn at Washington State University. In the slide you can see the intensity images for each view.