The difficulty is is that every mesh is different, so automated tools may not give you the results you want. At least with reconstructme meshes are generally of the same sort of size and quality, so an automated approach can potentially work.
In Meshlab you can save a number of filter choices together as a script, under Filters > Show current filter script. What I do is select the largest continuous mesh (or multiple meshes, as required), with a very small selection just on the edge, with the continuous (rather than individual) vertices selection tool. I then have a script that inverts the selection, deletes the selection (usually all the stray bits of the mesh), then runs clean up choices of Remove duplicated faces, Remove duplicated vertex, Remove unreferenced vertex, Remove zero area faces, Select non-Manifold edges, Delete selection, Select non-manifold vertices, Delete selection.
Depending on how well this does - it may crash Meshlab, or delete too much of the mesh! - I then run a filter to select overly-large vertices using Select faces with edges longer than... which needs to be tuned to the mesh to select the right size. There is the option of using Z-brush painting to clean up any remaining areas that are part of the mesh that are still not great, or just selecting areas to delete.
Lastly, I run a Surface Reconstruction: Poisson on the mesh. Again, the values that can be used depend on the final result you are after. Generally, set the Octree depth between 6 and 11 (usually won't go much higher than this without crashing), and the Solver divide to Octree depth -1. This *should* give you a manifold mesh. I tend to import these into Netfabb Studio Basic (free) to check, rotate, scale and cut a flat bottom on the mesh, ready for printing.
Tony Buser's video shows an additional couple of tidying up steps you can do with other programs.
Meshlab has no undo, so saving regularly helps! I found following these videos by (I think) the creator of Meshlab very useful:
http://www.youtube.com/user/MrPMeshLabTutorials?feature=watchHope that helps!