We have shown that semi-automatic generation of opacity functions is possible for datasets where the regions of interest are boundaries between materials of relatively constant data value. The histogram volume structure presented here captures information about the boundaries present in the volume and facilitates a high-level interface to opacity function creation. The user controls which portions of the boundary are to be made opaque, without having to know the data values that occur in the boundary.
Given that boundaries in the volume are always manifested by a curve of a particular shape in the histogram volume, it makes sense to apply computer vision object recognition techniques to the histogram volume. We are investigating the feasibility of using the Hough transform to detect the curves in the histogram volume and measure their intensity . Also, it may be possible to adapt the methods to non-scalar data, such as comes from multi-echo MRI. Finally, as mentioned before, we are interested in performing perceptual studies to validate the claim that direct volume rendering can, unlike isosurface rendering, accurately convey surface quality or measurement uncertainty to the viewer.