A bit off-topic: a prof. at the Uni once told me that David Lowe (who invented SIFT) never got any money from the patent (which I think belongs jointly to him and UBC) because commercial implementations would make a few changes to the algo and just use it, without citation or license.
If any HN'ers know more about this, I would be glad to hear it
While it does seem to be regarded to be easy to avoid infringing on the SIFT patent by changing the algorithm in a minor way (e.g. using SURF features but doing SIFT-style search), I know there are companies that have licensed SIFT from Lowe, and I would expect he's been paid for those licenses.
IP is protecting its usage (real world applications, ideally, with assessed damages), rather than the implementation. The analogy is: bulb is patented, but the patent shouldn't cover the bulb factory as well. You can pay for the damage, or shut down the factory, but you cannot force to dis-integrate the factory.
You can use patent algorithms if you have proper license, but it shouldn't forbidden various implementations from ground up. (however, you can form the argument that the implementation is dedicated to violate the patent rights, thus, should be responsible as well).
As always, software patent is a bch. Without a court trial, no one knows the exact answer.