JPEG Committee On The Ball, Seeks Prior Art

Sangui5 writes: "It seems as if the JPEG Committee has noticed the recent patent fuss, and is working on the prior art angle. Good to know that even though there's a new standard, the committee is standing by their previous work."

Not reading the article, eh?

[cduffy]

According to the article, JPEG 2000 has had extensive work done to obtain royalty-free licensing. In general, it is thus implied that the JPEG committee believes JPEG 2000 to be unaffected by the patent claims which allegedly restrict the existing JPEG standard.

Re:Jpeg

[mfos.org]

Not at all, the algorithims are from seperate families. GIF and PNG are lossless compression, while JPEG uses discreet cosine transform (lossy) I believe. I could be wrong however. As for having an open alternative to JPEG, I think JPEG2000 is supposed to fit the bill.

Re:Jpeg

[Anonymous Coward]

GIF lossless? Hardly. The only reason it ever succeded as a image format at all was for it's use in the internet. It's just plain smaller (still is) than most image formats, and takes very little CPU power to decode it in a timely fashion. I mean, we're not talking photographic quality here: 255 colors and an alpha channel, with the ability to have some animation quality. JPEG and PNG, on the other hand, are horses of a different color all together (no pun intended.)

PNG is lossless, yes. But, for small monotone colors (the kind used here on /. for example), it actually beats JPEG on size most of the time. (We all know that compression speaking, zlib is pretty darn good at shrinking lot of repeating patterns). To get good quality out of JPEG (I mean photographic quality here), you need to substantially increase the file size (depending on encoding methods). I have found that most JPEG compressor engines run defualt at around 70% quality, and can usually squeeze some pretty good results out of that 70%. Iv'e personally found that I can almost always distinguish .jpgs from the origional tiffs (24bpp tiffs here) up untill 95% (which still nets pretty darn goood compression). The plain truth is, that if you need good quality, 24-32bpp PNGs are pretty hard to beat (for internet use). If you need better quality and compression at the same time, you better just zip that sucker up and do with it what you will.

Re:The role of standards bodies

[dvdeug]

At least the first group Forget (intentionally misspelled) contacted wasn't the developers of The Gimp or something.

Why would they? When you're doing something like this, open source people don't have cash to pony up, and help keep people dependent on the technology in question.

Re:Jpeg

[stuuf]

JPEG cuts the image into 8x8 pixel blocks, then compresses them. JPEG2000 uses streaming wavelet compression that can decompress and increase resolution as the file is transferred. The current patent doesn't affect gif/png, but it might make someone think they can announce a patent on png just like Forgent did with jpg.

Re:Isn't JPEG just a FFT?

[morcheeba]

It's actually a 2-dimensional DCT (discrete cosine transform), some quantization applied inequally (the low-frequency components are better represented; this is the lossy part), and then entropy-coded (Huffman or arithmetic, aka zip-like lossless compression) in a cool zig-zag fashion. Here's a quick, decent summary. [rasip.fer.hr]

The patent process needs good prior art databases

[ukryule]

What the JPEG Committee is proposing is a general 'prior art database' which should be generally useful for any image related dispute. This is a 'good thing'(tm) whether anyone is currently trying to claim JPEG prior-art or not. The current patent process desperately needs quick and easy ways to search for prior art.

To see why, consider the standard process for creating a patent in a large company:

1. You write up an overview of the patent, and submit it. Presumably you know your field, so the first 'prior art filter' is you - have you heard of anything similar?
2. You hand it over to your companies patent agent. (S)he will probably be assigned to a particular field (e.g. 'audio/video/image processing'), so understands the area, but is not going to be an expert.
3. The patent agent reads through your explanation and does a prior art search - and returns to you a selection of things that may be relevant.
4. You explain how your invention is novel compared to these. If you convince him, then the wheels are set in motion, and your company (eventually) submits a patent application.
5. The Patent Office reads it and searches for prior art. If they find none, your patent is granted, while if they find something, then it is up to you/your company to dispute their findings.

So, in steps 3 & 5 you have legal experts who understand the area, but are probably not technically expert in the exact field of the patent who have a responsibility to search for prior art. They are also under time pressure, as they have loads of proposals to deal with. So what they do is pull out a few relevant keywords from the proposal and search on them in some prior-art database.

The most obvious (and easy) database is the existing patents DB. Now, I'm sure they have other databases they use, but whenever I've been through the process, nearly all the potential prior art which has been returned to me via the patent agent has been previously published patents. So if an idea hasn't been patented before, then it's got a good chance of getting accepted as a new patent.

So if the JPEG group build an extensive, easily searchable catalogue of prior art (with times, keywords, etc.), then it will make the patent agents life a lot easier, thus increasing the quality of patents.

Re:Jpeg

[wheany]

GIF lossless? Hardly.[...]255 colors and an alpha channel

Gifs are lossless. Just because you can't use more than 256 colors, doesn't mean the format is lossy. And Gifs don't have an alpha channel, per se, you can just define one of the 256 colors as being completely transparent or not.

Re:Good!

[ukryule]

The idea is just silly. Makes me want to go patent the Redbook standard and sue the RIAA.

Funny you should mention that ... the CD Specification [bizland.com] is stuffed full of patents (mainly from Philips & Sony) which are due to expire in a year or so. I doubt you'll notice the couple of cent drop in price of a CD due to not having to pay the patent holders when they do expire though :-)

The only recent time Philips has got upset with the RIAA was when they weren't sticking to the Redbook standard [slashdot.org]!

Re:It'll Expire Next Year

[tlambert]

Actually, it'll expire 17 years from date of issue, since it is grandfathered as a submerged patent filing. In other words, it's governed by the old rules because it was filed under the old rules.

Patents files on or after June 8 1995 are 20 years from date of filing; before that, patents were from date of issue, not of filing, and their term was 7 or 14 years, and grew to 17. One of the reasons for the change to a 20 year term was the move to date of filing as the baseline date.

Either way, it's too damn long a period for this industry.

-- Terry

Re:Prior Art?

[TitaniumFox]

Actually, she was quite a cutie in her first Playboy photo shoot. No silicone. Miss Feb, 1990.

Check her out here [homestead.com]. (Nope, not nude.)

Nope

[spacefrog]

Although it does smell a bit like Rambus, the situations aren't really similar at all.

The big difference here is that Rambus was a member of the standards body in question (JEDEC). The agreement they signed to become a member of this standards body obligated them to disclose patents. They didn't and thus violated a contract.

As far as I can tell, Forgent is not a member of the JPEG organization, nor did they ever propose to the JPEG body that they adopt their IP as a standard.

The two situations may look similar on the surface, but that is where the similarities end.

Re:Jpeg

[BJH]

Time for you to go back to school and learn the difference between lossless and lossy compression, methinks.

GIF only appears lossy because what you're compressing with it requires more colours than it can provide. The GIF format itself uses only lossless compression.

Re:Does JPEG 2000 have an open license?

[neibwe]

A discission I found from Google states that:
...JBIG and JPEG 2000 both mandate use of one or more patented techniques
which are owned by companies that take part in JPEG.[1]

And mentions some nice licensing guidelines:

I would welcome a standard wavelets codec, and an associated standard

format, but I'm not very interested unless they are either free of
patents or include a free-of-charge unlimited license.

--Nick Lamb njl98rSWAPWITHATSIGNecs.soton.ac.uk: When will Gimp support JPEG2000 [berkeley.edu]

But earlier on Slashdot in Slashback: Alternatives, Ads, Apple [slashdot.org] A discussion regarding the licensing scheme to Jpeg2000 pointed to it (the license) being open.

...because those companies who currently claim patents on part 1 of JPEG2000 have also agreed to license their patents to

the general public without royalty

--yerricde, User #125198: JPEG2000 is royalty free [slashdot.org]

Can anyone verify that Jpeg2000 has an unencumbering license?

I doubt that Forgent have read through the patent

[Anonymous Coward]

If you read through the patent claim forgent has you can see it is RLE encoding based on LZW and not approximation like JPEG has. Forgents patent is non lossy where JPEG is, actually it has more claim to PNG than JPEG in that case. Which is not very good either if you ask me. I also had a look through the database at the patent office and there are hundreds of similar claims to all kinds of compression algorithms. Some are so fuzzy you could applicate them to anything. Somehow I feel this is only the beginning. The patents really needs a change in the laws in order to clarify the patents and get rid of these ambushes. Fortunately in EU you don't have to care about patents as long as it is for free use.

Sorta like Unisys

[cout]

Perhaps a closer analogy is Unisys. For many years, GIF files were the thing to use. They were popular on Compuserve, then on local BBSes, and along with JPEG, became the image file format of choice on the web. After all of that, Unisys decided to take advantage over their LZW patent, and require a small royalty for any applications that used GIFs.

It wasn't too much later that slashdot came around and posted a link to http://burnallgifs.org/. I wonder how long it will be until they post a link to http://burnalljpegs.org/.

Re:Uh-oh: time running out?

[schon]

If Phillips no longer holds a patent, what (monetary) incentive do they have in trying to force record companies to adhere to the CD standard instead of corrupting it with their wacky, hopeless "protection" schemes?

Whether Phillips holds a patent or not is irrelevant.

The reason Phillips (rightly) got upset is because the offending companies were still using the CD-logo, which is trademarked by Phillips.

Trademarks don't expire, as long as the company that holds them continues to pay for and police them.

Trademarks vs. Patents

[Idarubicin]

Actually, there is a principle similar to that which you describe. Go Googling for 'laches' or 'law of laches' and you'll find the relevant material.

If a patent holder is aware of infringing activity and doesn't do anything about it for a period of time (six years in the United States) then the infringer is not liable for damages.

However, unlike a trademark, a patent does not lapse without enforcement. As soon as the patent holder does get around to notifying the infringing party, then they can start claiming damages from that point on.

In other words, they can't sue every instance of 'infringement' that took place over the last fifteen years--they have forfeited that right. They may, however, demand royalties for further uses of JPEG compression. Assuming, of course, that their patent does cover the method in question, and that it holds up in court, and no prior art is found, and so forth...

IANAL, YMMV.

Re:Something I don't get.

[Anonymous Coward]

Patent law has something called laches.

Basically, if an infringing company has publicly used a technology for (+-) 6 years and the infringed party does not give notice of infringement, then they are restricted from receiving royalties that accrue prior to the notice of infringement. This is not a hard rule as the infringed may show good reason why they did not give notice in this time frame:

http://www.converium.com/web/converium/converium .n sf/articles/5731FF9F4372B6ED85256B43006EA07D?OpenD ocument

There is also a current ruling that sided against undue stalling in the process of patent prosecution (the process of filing and obtaining patent and possible related divisional and/or continuation patents):

http://www.kenyon.com/symbol.htm

There are also other antitrust implications related to set standards and patents, but these implications do not preclude one from holding a patent on a "standardized" technology.

Re:Nope

[markmoss]

As far as I can tell, Forgent is not a member of the JPEG organization

The last time Forgent's patent (actually Concurrent Labs) was discussed, one poster said that he had been involved with JPEG, and Concurrent Labs was a member in 1992-95 (IIRC). This patent was granted several years before CL joined JPEG. All the members, including CL, signed agreements to reveal all patents and applications related to the standards under discussion. CL never brought up this patent. This means one of three things:

1) CL was in breach of their contract with the JPEG organization.

2) CL reviewed this patent vs. JPEG's compression methods and decided it did not cover JPEG, so it didn't have to be brought up.

3) The left hand didn't know what the right hand was doing - that is, their still-picture people on the JPEG project didn't even know about the video compression patent.

When Forgent bought CL, they bought up their liabilities along with their assets. So they had better be arguing #3, because #2 is an admission that their suit is groundless as far as anything in the JPEG standard before 1995 goes, while with #1 JPEG can sue to be "made whole" by requiring Forgent to license it's patent(s) for free for JPEG applications. And I doubt that CL was ever big enough to make #3 very believable...

Re:Besides the obligitory "Forgent-ery" joke...

[DotComVictim]

You are 100% correct. They obtain an intermediate image between (delta between two different frames after motion compensation). Then, the DCT is applied to this delta, quantized, and the coefficients are encoded.

The steps applied to the delta are the core of the JPEG compression. However, they are not mentioned in the list of claims in the patent! Further, the patent itself points out prior art on the use of DCT quantization.

Basically, there is no way for this claim to stand, especially when it affects far too many people with deep pockets (possibly more important than the technical points).