3dAutomask first finds the 3dCliplevel value. It then clips the data (like your 3dcalc command). It proceeds with some extra steps:
- keeping only the largest above-threshold cluster of nonzero voxels in this mask
- filling in any small holes in that mask
- eroding off any loosely connected regions in the mask
- redilating any closely connected regions that were lost in the erosion
You should look at the two masks, but you will probably prefer the 3dAutomask generated dataset. If you think it is a little too small, you can dilate it outwards with the
-dilate option.
Zorkon the Masked