First Previous Next Last Index Text

Slide 21 of 42

Now here, we're sampling everywhere on a grid, with no respect to the original boundary, and at even a coarser resolution than before. But still, the same curves are being traced out as before. This works, because of the original assumptions of material uniformity and isotropic blurring. And, we should have good coverage of these curves for at last two reasons- the boundaries of real objects tend to assume a variety of positions and orientations relative to the sampling grid, and, because real measurement devices necessarily band limit, there is always going to be some blurring of the boundary, so there will be intermediate values to fill out these curves. I should note however, that we have found it sometimes helps to blur the volume a little bit as a pre-process to increase the coherence of these scatterplot curves.