bane/gkms tutorial: Step 3: Distilling the Histogram Volume

As with the other gkms commands, you get usage info for distilling histogram volumes by supplying no arguments, and invoking just "gkms info":
usage: bin/gkms info [-d 1|2] [-m <measr>] <input> <output>
Default values are in {}s
    -d 1|2 : dimension of histo-info to be created
             {1}
  -m measr : how to collapse axes of the histogram volume
             "mean": average (as done in GK's published research)
             "median": value at 50% percentile
             "mode": most common value
             {mean}
     input : histogram volume being analyzed
    output : histo-info to be created

There are not too many parameters here. The dimension of the info file is determined by the dimensionality of the opacity function you want to create. The default ("-d 1") is good if you're working with plain one-dimensional opacity functions ...

gkms info engine-hvol.nrrd engine-info1.nrrd
... or the "-d 2" is for when you're interested in exploring the power of two-dimensional transfer functions ...
gkms info -d 2 engine-hvol.nrrd engine-info2.nrrd

The "info" files generated by this step are arrays of floating point numbers which summarize important aspects of the histogram volume. Specifically, the 1-dimensional "info" files contain essentially g(v) and h(v), and the 2-dimensional "info" file contains essentially h(v,g). See this section of my MS thesis for definitions of these functions. One dimensional opacity functions are simple and familiar, but for many measured datasets the expressive power of two dimensional opacity functions facilitates a much higher level of control over which boundaries are rendered, and how.

There isn't any visual feedback for these programs, so there aren't any pictures to put on this page. Chances are, once you've generated the info files, you can delete the histogram volume, since it isn't used by any other gkms commands.

The "measr" parameter is not one which I've described in any of my published work, but which can be useful for analyzing histogram volumes from some more complicated datasets. ... More information as time permits ...

Valid HTML 3.2!