Talk:JPEG 2000

From Wikipedia, the free encyclopedia

Contents

[edit] Say NO to Commercialism

j2k-codec is a commercial package... please remove it from the link section or add some other commercial packages as well...

[edit] Edit

The article mentions that 'editability' is a feature of JPEG 2000, but does not explain. If I knew anything about this, I'd fix it.

[edit] Disadvantages?

Right now the article looks like it was written by some JPEG 2000 promotion group, with the first section being "Superiority of JPEG2000". I chuckled when I saw that. I think this should be renamed "Advantages" and another section with "Disadvantages" created right below. Disadvantages possibly being slower compression/decompression (not so good for digital cameras with limited processing power), patent issus (mentioned at the end of the article), lack of support and very slow adaptation, and that images that are "good enough quality" with no visible artifacts are pretty much same size with JPEG and JPEG2000. At least in my tests JPEG2000 seems to win only with quality that clearly shows some noticeable artifacts; It's just that JPEG then shows more. So it's questionable if the format is all that useful, as images with no visual artifacts are usually desirable (and not that big either). Magnus below at Test-Image-section also said something that confirms my observations: "But I think that JPEG2000 isn't better than JPEG at higger queality, it smooths more de image and deletes some detail that JPEG show it." At some compression ratio for some images, JPEG2000 may indeed look worse for the same file size. Negleting to mention this in the article is just hiding the facts. Also in my experience, lossless JPEG2000 is worse than PNG (compressed with pngcrush) even for photos, not just diagrams as the article suggests.

But I didn't edit the article because my info may be wrong or old. Someone better informed should do it if she seems it as appropriate.

There sure should be a section on disadvantages. I tried out JPEG 2000 a bit. And well, the disadvantages outweigh the advantages. First of all, although it eliminates the "jpeg artifacts", it smooths the image too much. A #8 compression on jpeg 2000 is a much larger files size than a #8 regular jpeg, and if compensated to match file size, the regular jpeg preserves more detail. And anyway, hardly any applications support jpeg 2000, so what's the use? Althepal 03:27, 8 January 2007 (UTC)

I wonder if it would be okay to insert into the article criticism that isn't backed by any notable publication. I too find J2K disappointing. The only thing it might do better than JPEG is extreme compression (say 1/50), and that's not what most people are after. The only general use for it I can think of is lossless photos compression that compresses better than PNG and is somewhat compatible. But that isn't too promising considering JPEG-LS almost always compresses better and always requires much less horsepower. Maybe the situation could be better with better encoders (and maybe they have improved sincen I last tested)?

So what do you reckon, should this be added in? ehudshapira 06:26, 17 June 2007 (UTC)


User ehudshapira wrote: The only thing it might do better than JPEG is extreme compression (say 1/50).

My reply: If you simply feed an image into JPEG 1991 and JPEG 2000 and do extreme compression on it, JPEG 2000 will definitely come out looking better. However, for extreme compression JPEG 1991 is just about as good as JPEG 2000 if you do things as follows in the 1991 compressor. First the compressor reduces the image size -- by a factor of two for each of horizontal and vertical, for instance. Then the smaller image is compressed in the ordinary JPEG 1991 way. Later, the decompressor decompresses it in the ordinary way, and then the image is zoomed back up by the factor of whatever it was reduced by. The consequence of reducing and re-expanding it is an image that suffers from some blurring artefacts. The re-expansion is done by interpolation. Blurring is not as objectionable to the human eye as the texas chain-saw massacre of blocking and ringing that happens in JPEG 1991 when you do extreme compression on it. What JPEG 2000 produces under extreme compression is a blurry image and it does so by reducing the image's resolution, which is the same thing as reducing the size. The 1991 standard doesn't include the feature of reducing and re-expanding for extreme compression. User sean 89.234.85.117 20:36, 5 July 2007 (UTC)

[edit] Test-Image

Perhaps, somebody can work it into the article.80.185.8.102 20:58, 1 January 2006 (UTC)

It's an interesting test image; it has been saved at 20% quality (6.9kb), and the JPEG2000 image has been saved with the same file size.
But I think JPEG2000 isn't better than JPEG at higher quality. It smooths more of the image and deletes some detail that JPEG would show.
I have found another comparison with some levels of compression with the Lura Wave plugin for Irfanview :[1] and these are the results:
Images in EMBED tag (browsers with a JPEG2000 plugin)
Images in IMG tag (browsers with naive JPEG2000 support, as Konqueror)
The frist table are images without colour subsamplig (4:4:4), and the second with a 4:2:2 subsamplig.
Sorry if my English is very poor.
--Magnus Colossus 16:08, 16 February 2006 (UTC)
i demand lena. 60.48.74.92 00:40, 27 July 2006 (UTC)

[edit] JPEG/JPEG 2000 comparision with JPEG images?

I don't understand why the test images provided here try to show the differences between JPEG and JPEG 2000 by showing them next to each other on a JPEG image. This is like comparing CD quality sound and HD quality sound by copying both of them on a Audio CD... --Abdull 21:01, 12 June 2006 (UTC)

I just came here to say the same thing. They took a JPEG 2000, turned up the loss, and then to show it to people on Wikipedia, they re-compressed it as a JPEG? If the final JPEG as saved at really high quality settings, then it's not as bad as the recording example given above, but it is still NOT VALID. It can NOT be a fair comparsson of the artifacts of JPEG-2000 vs JPEG when they're both a JPEG! It needs to be something lossless. I assume PNG is the most widely supported by web browsers?

Fixed. I hope you like the new image. --Shlomi Tal ☜ 18:45, 24 June 2006 (UTC)

[edit] Section on Techical Discussion

I've read the techinal description section, and I think it can be re-written a fair bit. In particular, the description can be organized much better, and the rich feature set offered by J2K can be made more explicit (eg: the progressive nature of the code-stream, compressed domain manipulations, and so on).

I'll probably be able to do this in a little while - but since I'm new to editing Wiki, I thought I'd make a post first and get people's opinions on this before I did anything substantial.

--Brokentooth 22:21, 20 July 2006 (UTC)

[edit] Free software

Regardless, however, of the monetary cost of licensing JPEG2000 patents, it would still be impossible to comply with the Debian Free Software Guidelines (the acid test of software freedom) with freely-licensed patents unless the licenses were redistributable and irrevocable, even if the licensed application is modified. This alone could hamper adoption of JPEG 2000 for web purposes as it would exclude open-source web browsers (most notably Gecko-based browsers) and popular LAMP web applications.
I know very little about free software licensing in general, but surely it could be implemented in a library with free software compatible license (even tho the license itself obviously wouldn't be a free software license) and therefore included as a non free software portion of a free software application. Non free software proponents may not want to do this, but it sounds to me like this is suggesting it's not possible to have JPEG2000 support in freesoftware which I'm pretty sure isn't true. Besides that can't you write the source code even if you can't distribute the binaries? Isn't this how XviD works? Also, VLC seems to have survived despite including support for the patented XviD...? Not to mention GIF was supported while it was patented. I understand why free software proponents dislike (or hate) patents, but I feel the current wording is misleading as it appears to suggest patents must be a deathnell Nil Einne 15:12, 11 January 2007 (UTC)

Yes, Debian can put software in the non-free section when it doesn't meet the DFSG, and has before.


[edit] Legal Issues

I wanted to bring this up because doing a patent search on "wavelet transform" and "wavelet image" seems to bring up many patents, possibly around 500 or so. So the statement about the heavily patented area seems to be true on it's face although there's not a study or a white paper to point to. Just a patent search will bring the facts to light.


this phrase: "JPEG 2000 is not widely supported in present software due to the perceived danger of software patents on the mathematics of the compression method, this area of mathematics being heavily patented in general[citation needed]."

is invalid. Do it based on any research?

"perceived danger of software patents " can't be the reason of not widely support. The work processed to clean up JPEG2000 standart from submarine patents is a lot bigger than for example JPEG or GIF. But JPEG and GIF is more sphread now. So this assumption on reason of "not widely supported " must be removed ASAP, it is just lie.

[edit] "JPEG 2000 is included in most Linux distributions."

What is this supposed to mean? AnonMoos 03:03, 5 October 2007 (UTC)

I think it means that most Linux distributions SUPPORT it as a standard. Inclusivedisjunction (talk) 07:01, 19 January 2008 (UTC)

[edit] machine judgment of quality

Images machine-judged to be of equivalent quality for both compression schemes often look better to humans in JPEG 2000 at low bitrates.

I am removing the above because all it means is "A certain machine test of image quality doesn't work". If humans judge that one image is better quality than another, then a working metric will give them different scores. The whole purpose of such a metric is to predict human judgment of quality. TomViza 19:48, 4 November 2007 (UTC)

[edit] Article cleanup

The tone, partiality, article structure, and difficult readability all need to be addressed. The article is not only extremely technical it is addressing the perceived advantages of JPEG2000. Non-technical users will not understand this article even at the introduction. It needs the more technical bits moved to another section and a gentle introduction, history, and body copy must be written. --KJRehberg (talk) 05:07, 17 January 2008 (UTC)

[edit] What is the perf difference?

The article makes it sound like the encoding and decoding performance is significant, but I see no hard numbers. I looked for them on the web, and couldn't find any either. If someone has some speed and/or memory numbers, it would be nice to add, because this article has the important numbers in many other places. KeithCu (talk) 10:17, 15 March 2008 (UTC)