[vtkusers] SIFT legal issues (PATENTS)

[Bill Lorensen]

David,

> Luis,
>
> As usual, an impressive, informative, and comprehensive response, thanks!
>
> Am I correct in the summary that "Yes, you can submit code of patented
> algorithms to the VTK Journal. It does not become illegal until the
> code is compiled."?

I would not assume that. It's like selling the plans to a patented widget.

>
> So when people say "we used SIFT..." in a journal paper, they are
> using the executable provided by Lowe. If they had compiled their own
> implementation, they would be breaking the law. Correct?
>

I doubt if this is true. I think in general, researchers are clueless
when it comes to IP. For example, GE held (it has expired) the patent
on Marching Cubes. In 2002 MC had the largest number of citations of
any published Siggraph paper. I know that almost none of the
implementations used the GE code. Some may have used the VTK code.

> Is it fair to say that these types of algorithms are still nice to
> have in ITK/VTK format for those users who wish to use it for
> "amusement only"?
I don't think we can include any Patented algorithms in either
toolkit. It is too hard to manage. And it makes our customers
uncomfortable.

In the past there have been exceptions.
In VTK, there used to be a number of Patented algorithms. General
Electric owned these patents. We had explicit approval from GE to
distribute the code which was kept in a separate, optional directory
called Patented. Also, Kitware had some algorithms they expected to
patent. Each of the files that implemented the GE patented algorithms
had these comments (with appropriate description and patent #):

 THIS CLASS IS PATENTED UNDER UNITED STATES PATENT NUMBER 4,710,876
     "System and Method for the Display of Surface Structures Contained
     Within the Interior Region of a Solid Body".
     Application of this software for commercial purposes requires
     a license grant from GE. Contact:

         Carl B. Horton
         Sr. Counsel, Intellectual Property
         3000 N. Grandview Blvd., W-710
         Waukesha, WI  53188
         Phone:  (262) 513-4022
         E-Mail: Carl.Horton at med.ge.com

     for more information.

The Kitware "patented" algorithms contained comments like this:
     THIS CLASS IS PATENT PENDING.

     Application of this software for commercial purposes requires
     a license grant from Kitware. Contact:
         Ken Martin
         Kitware
         28 Corporate Drive Suite 204,
         Clifton Park, NY 12065
         Phone:1-518-371-3971
     for more information.

Once the Marching Cubes patent expired, we removed it from the
patented code. Kitware also decided to move their "Patent Pending"
code into the non-restrictive directories.

In an early version of VTK we had implemented Taubin's mesh smoothing algorithm:
A Signal Processing Approach to Fair Surface Design, by G. Taubin, Siggraph'95.

We discovered that IBM had a patent on the algorithm and removed it from VTK.

For ITK, which was developed under contracts from the National Library
of Medicine, some contractors submitted implementations of code their
institutions had patented. I don't know the wording of their contracts
but I assume they allow some non-commercial use.

My advice is to avoid (boycott) patented algorithms. There is a lot of
good work that can be done without them.

Disclosure: I have 31 "software patents":
http://marchingcubes.org/index.php/Lorensen_Patents
Some are scary: http://www.google.com/patents?id=VyAbAAAAEBAJ&dq=patent:4984157

Bill