With 2D 'white blood' venograms, you can probably find a voxel-wise threshold setting that will delineate vessels fairly clearly. You could do this using the AFNI threshold slider while looking at the images. Then you could create a vein mask by using something like

3dmerge -1thresh VAL -dxyz=1 -1clust 1 50 -1tindex 0 -prefix Veno_masked Veno+orig

where VAL is the threshold you select, and the rest of the command is to eliminate small isolated clusters. Of course, you'll have to experiment with this somewhat.

For 'black blood' venograms, I'm not sure what a proper approach is, off the top of my head. It would probably start with using 3dAutomask to build a brain mask, then using 3dcalc to threshold voxels BELOW a threshold that are inside the brain mask.

With 3D TOF datasets, the problem is harder. In my limited experience with these, they only show vessels when using some sort of maximum intensity projection -- on a voxel-wise basis, the vessels are only a little brighter than the surrounding tissue. I suppose it would be possible to do some sort of local orthogonal projections with 3dcalc, and keep only those voxels that are local maxima along at least one projection. This would require experimentation. Better would be an idea Ziad and I had several years ago, which would be to use the projection algorithms in the volume rendering plugin to determine which voxels are bright in many angles of the MIP, and keep those -- but we never implemented this.