x264 --no-dct-decimate option

[malikcis]

Hi,

Do you know if the coefficient thresholding algorithm in x264 that is deactivated with --no-dct-decimate option is patented? Can I implement this algorithm in a design without license issues?

thanks,

[LoRd_MuldeR]

I think that is an encoder-side only algorithm/optimizations, so it should not be covered by any H.264 patents. On the other hand, if you implement a H.264 encoder and are going to use it in a commercial product, you will have to pay patenting fees to the MPEG-LA anyway - not for the "dct decimation" algorithm, but for using H.264 in general. As for x264, you can take any code/algorithm from x264 and re-use it in your own application, as long as your application's license is compatible with the GPL. Otherwise you need to buy a 'commercial license' for x264. AFAIK, even with the x264 license, you still have to pay for the H.264 patents - to the MPEG-LA.

But to be sure, I would suggest to simply ask at:
http://x264licensing.com/contact

[Dark Shikari]

I think that is an encoder-side only algorithm/optimizations, so it should not be covered by any H.264 patents.This is completely untrue, as there are literally hundreds if not thousands of patents on encoder-side algorithms.

However, this is both a very simple algorithm, and taken directly from a JVT document -- which are all required to explicitly mention whether or not they will be filing patents on the material contained within -- so I highly doubt it is patented.

[Marin85]

Sorry for the naive question, but discrete cosine transform is something that is well known to signal processing for more than a century. Whether the signal is an image, sound, video etc., does not seem to matter much. It is closely related to FFT and so any optimal algorithms for DCT come from optimal algorithms for FFT calculations, the latter are probably known for at least two decades, if not more, i.e. even before the whole idea of MPEG. So, how could this be patented?...

[LoRd_MuldeR]

It's not that unusual that apparently "trivial" things get patented. You just need to be the first one to claim a patent for that ;)

Actually it seems that in the US it's relatively easy to get a patent, especially with respect to software. Whether that patent will withstand a lawsuit, that's a different question.

But until someone has a good reason as well as the required resources to file a lawsuit...

[Dark Shikari]

Sorry for the naive question, but discrete cosine transform is something that is well known to signal processing for more than a century. Whether the signal is an image, sound, video etc., does not seem to matter much. It is closely related to FFT and so any optimal algorithms for DCT come from optimal algorithms for FFT calculations, the latter are probably known for at least two decades, if not more, i.e. even before the whole idea of MPEG. So, how could this be patented?...Just because something is related to an old idea, does not mean it cannot be a new (and thus patentable) idea.

You can't patent the DCT, but you can patent a novel DCT approximation.

[malikcis]

Thank for all answers, I think my original question is sufficiently answered.