Compression - JPEG2000

Author

JPEG2000

2005-12-08, 3:55 am

hi,

is SPIHT or EZW used in the JPEG2000 standard? how do they work?
anyway, which articles do you reccomend for deeper understanding of
JPEG2000?

do you think JPEG2000 will ever replace JPEG in future? is it really
the still image compression standard of the future?

thanks

Thomas Richter

2005-12-08, 3:55 am

Hi,

> is SPIHT or EZW used in the JPEG2000 standard?

No, and no. JPEG2000 uses the EBCOT.

> how do they work?

There's no one-sentence description of SPIHT or EZW, neither EBCOT. The
basic idea of EZW is to make use of inter-band correlations to order
wavelet coefficients according to significance, then encode them embeddedly
one after another in this order. The EBCOT works quite differently, it only
makes use of inter-band correlations where it splits the wavelet banks
into codeblocks of 64x64 coefficients and then encodes them from MSB to
LSB in a bitplane encoder. On doing this it computes rate and distortion and
is thus able to optimize the overall PSNR.

> anyway, which articles do you reccomend for deeper understanding of
> JPEG2000?

Even though I haven't read it, I would believe that the Taubman/Marcelli
book is the prime source here. Other than that, there's the standard, though
this is not very readable.

> do you think JPEG2000 will ever replace JPEG in future?

Depends in which area. I don't think it is likely that JPEG2000 will
make JPEG-1 obsolete for the typical internet-imaging. However, it seems
to have a good future for medical images (or all other images provided they
are large enough), and for digital movies. JPEG2000 is the image compression
standard for the digital movie initiative.

> is it really
> the still image compression standard of the future?

That question is a bit too broad. Is BMW the car maker of the future?
Depends on the target application.

So long,
Thomas

cypherswipe

2005-12-08, 7:56 am

>>do you think JPEG2000 will ever replace JPEG in future?

Possible, but not likely (at least not for a long while). JP2 has 2
major drawbacks. 1) It is very slow to compress/decompress compared to
other image formats (especially JPG). 2) It isn't completely free. The
basic version (part 1) is free, but parts 2+ are subject to licenses
and fees.

[color=darkred]
the still image compression standard of the future?

No. Not "the" single standard. The standard for photographic images
perhaps, but PNG is much better for graphic type images.

Thomas Richter

2005-12-09, 9:55 pm

cypherswipe wrote:
>
>
> Possible, but not likely (at least not for a long while). JP2 has 2
> major drawbacks. 1) It is very slow to compress/decompress compared to
> other image formats (especially JPG).

Definitely, and no arguing about that.

> 2) It isn't completely free. The
> basic version (part 1) is free, but parts 2+ are subject to licenses
> and fees.

*Cough*. This is a bit confusing. First of all, JPEG2000 part-1 isn't
"patent free", or "FREE" as the GPL defines this if this is your
definition of "freedom" (it isn't mine, but that part aside). It is just
that the companies that contributed to the standard guarantee you that
you can get a licence royality free if you apply for the licence.
Second, part-2 isn't patent free either, and you cannot get licences
without having to pay money *for some parts*. However, part-2 is a
selection of extensions you most likely never need (all the images I've
seen so far are completely within part-1), and most (though not all) of
the extensions you might possibly need are again "royality free". For
example, the most useful parts of part-2 are combined in something
called "JPX baseline", and that again is "royality free", though not
"FREE" in some other senses.

Second, let me recommend on the "freeness" of jpeg-1: Your statement
here isn't quite correct either; the situation is, to some degree,
similar to jpeg2000: JPEG-1 *baseline* is - depending on whom you ask -
patent free (Forgent technologies definitly would not agree, though I
don't really follow their claim on the VLC patent either). However,
similar to JPEG2000, JPEG-1 has extensions: The QM-coder (arithemtic
coder), for example. And these extensions are, quite similar to the
JPEG2000 extensions, *not* patent free, nor royality free. For the QM
coder patents from IBM and Mitsubishi apply, and probably of some other
people.

>
> the still image compression standard of the future?
>
> No. Not "the" single standard. The standard for photographic images
> perhaps, but PNG is much better for graphic type images.

And there is JPEG-LS for lossless compression, which is much faster than
JPEG2000,
and compresses equally or better, and there is JBIG-1 and JBIG-2 for
bi-level (black&white) images, that perform better in this area, and....

All this doesn't diminish the effords or quality of JPEG2000. It is just
that every compression file format has to find its market niche, and
applies well to certain types of images. If you look for "compresses
all", you're wrong with JPEG2000, and PNG, and JPEG-LS. And, even in
this group - the counting theorem forbids this. (-:

So long,
Thomas

cypherswipe

2005-12-15, 7:55 am

The only way you'd ever wind up with a single "perfect" format would be
if it were a compound format that could use one of several compression
methods: a wavelet/JP2k method for photographic images (both lossy and
lossless), and good zlib/PNG method for logos, and a good JBIG/bitone
compression for bitonal images. Ideally it would also include a
layering capability similar to DJVU (LDF and some other formats are
similar, but have much poorer results. It's too bad lizardtech screwed
DJVU over so badly), which would make it suitable for mixed content
images.
On the other hand, such a complex format would no doubt have flaws of
it's own. (Look at all the compatibility issues TIFF has.)

Thomas Richter

2005-12-16, 3:55 am

Hi,

> The only way you'd ever wind up with a single "perfect" format would be
> if it were a compound format that could use one of several compression
> methods: a wavelet/JP2k method for photographic images (both lossy and
> lossless), and good zlib/PNG method for logos, and a good JBIG/bitone
> compression for bitonal images. Ideally it would also include a
> layering capability similar to DJVU (LDF and some other formats are
> similar, but have much poorer results. It's too bad lizardtech screwed
> DJVU over so badly), which would make it suitable for mixed content
> images.

Well, there are of course container formats like PDF that can may make use
of several image formats, depending on what's contained. And there are
things like JPEG2000 part 6 which also allow mixed content - bi-level
graphics is there either "fax" or JBIG-2 compressed, mixed with JPEG2000.

> On the other hand, such a complex format would no doubt have flaws of
> it's own. (Look at all the compatibility issues TIFF has.)

That's exactly the point. The overhead goes up, and so does the complexity.

So long,
Thomas

Sponsored Links