|
|
|
| Authored by: Anonymous on Wednesday, March 27 2013 @ 11:54 AM EDT |
The way I see it, motion-predictive video encoding was
pretty mature by 1999, so Nokia couldn't get a very broad
patent. First off, the claim construction would have to make
"motion model" read on the very mature "motion vector"
methods, which could be quite a stretch. Then the encoding
algorithm has to use a very specific method of decision
making (compute and compare a compund cost function) in a
very specific part of the VP8 encoding decision tree:
deciding whether to use a motion vector from an adjacent
macroblock in a "simple copy" mode or to use the same
adjacent-macroblock's motion vector in an "adjusted copy"
mode (i.e. NEW_MV with a non-zero mv_best).
So, suppose instead of computing Nokia's specific compound
cost function, my choice algorithm is (1) of the three MV
options in the spec, choose the MV source with the lowest
unmodified prediction error, (2) if that choice happens to
be mv_best, use NEW_MV mode and apply the adjustment.
I'm not an expert, and this is surely not the smartest
workaround, but you can get the picture. This patent covers
a very specific efficiency tweak that is probably useful in
getting minimum bitrate on some classes of test video, but
it should be a snap to code around it and still have a very
good encoder.
If that's the case, it's not essential and it's not a
blocker.
It might be worthwhile looking at some open-source encoding
programs and see whether the actual implementations they've
chosen are anywhere near infringement.[ Reply to This | Parent | # ]
|
|
|
|
|
