Thread Subject: dct co-efficients

Hi,

I would like to know if the co-efficient values in DCT used for an application like JPEG are fixed.
Or these are parameters that can an user can tweak.

Regards,

sbr487 <sharan.basappa@gmail.com> wrote in message <907870040.29899.1259076065233.JavaMail.root@gallium.mathforum.org>...
> Hi,
>
> I would like to know if the co-efficient values in DCT used for an application like JPEG are fixed.
============

Co-efficents of what exactly?

And is the DCT still be used for JPEG? I thought JPEG had moved to wavelets years ago...

On Nov 24, 7:42 am, "Matt " <x...@whatever.com> wrote:
> sbr487 <sharan.basa...@gmail.com> wrote in message <907870040.29899.1259076065233.JavaMail.r...@gallium.mathforum.org>...
> > Hi,
>
> > I would like to know if the co-efficient values in DCT used for an application like JPEG are fixed.
>
> ============
>
> Co-efficents of what exactly?
>
> And is the DCT still be used for JPEG? I thought JPEG had moved to wavelets years ago...

For an in depth look at the commonly used jpg check the spec:

http://www.w3.org/Graphics/JPEG/itu-t81.pdf

As far as wavelets and JPEG are concerned, there is a JPEG2000
standard ( .jp2 files). It seems only to be used frequently by
academics, researchers and others who don't need to interface much to
the rest of the world. The photo industry and web concentrate on the
widely accepted older standard ( .jpg files).

When I hear wavelet advocates make the JPEG claim (and I haven't since
last week), I always have to stop and ponder whether they are
intentionally or ignorantly misleading and what either one must imply
about the rest of their message.

Dale B. Dalrymple

Maybe I was not clear enough.
At a very low level, DCT is mainly mult/addition of spatial samples with co-effs, the question was whether the standard enforces these co-effs. Or it recommends them and one can be off (more/less accurate) depending on his needs.

On your comment about DCT usage today, I think wavelet is probably used in newer gen standards like MPEG4, some fax standard (I dont remember exactly), but that does not mean dct has become redundant. If that was the case, all new engineers should directly jump to wavelet transform skipping everything since it is not used ...

Regards,

On Nov 25, 12:04 am, sbr487 <sharan.basa...@gmail.com> wrote:
> Maybe I was not clear enough.
> At a very low level, DCT is mainly mult/addition of spatial samples with co-effs, the question was whether the standard enforces these co-effs. Or it recommends them and one can be off (more/less accurate) depending on his needs.
>
> On your comment about DCT usage today, I think wavelet is probably used in newer gen standards like MPEG4, some fax standard (I dont remember exactly), but that does not mean dct has become redundant. If that was the case, all new engineers should directly jump to wavelet transform skipping everything since it is not used ...
>
> Regards,

The definitions of the forward and inverse DCTs including the
coefficients used in the calculation are specified in section A.3.3 on
page 27 of the document to which I referred you. The definitions are
in the direct form. Implementations are free to use equivalent fast
transforms.

Compliance testing is addressed in ITU-T T.83 which includes 3
diskettes of sample images.

Dale B. Dalrymple

dbd <dbd@ieee.org> wrote in message <b433a115-d433-4c85-9810-29fc6da55587@u1g2000pre.googlegroups.com>...
 
> When I hear wavelet advocates make the JPEG claim (and I haven't since
> last week), I always have to stop and ponder whether they are
> intentionally or ignorantly misleading and what either one must imply
> about the rest of their message.


Why? Is there a reason why wavelet compression shouldn't be able to outperform DCT compression?

On Nov 25, 9:45 am, "Matt " <x...@whatever.com> wrote:
> dbd <d...@ieee.org> wrote in message <b433a115-d433-4c85-9810-29fc6da55...@u1g2000pre.googlegroups.com>...
> > When I hear wavelet advocates make the JPEG claim (and I haven't since
> > last week), I always have to stop and ponder whether they are
> > intentionally or ignorantly misleading and what either one must imply
> > about the rest of their message.
>
> Why? Is there a reason why wavelet compression shouldn't be able to outperform DCT compression?

The point isn't about whether wavelet compression works, I think that
it does, but it has not achieved wide acceptance replacing .jpg in
photography. The point is about how much one should trust those who
consistently practice behavior that is either uninformed or dishonest
in the pursuit of their marketing interests.

Dale B. Dalrymple

dbd <dbd@ieee.org> wrote in message <2f5d3115-eede-476a-a251-48b407b7f548@f20g2000prn.googlegroups.com>...
> On Nov 25, 9:45?am, "Matt " <x...@whatever.com> wrote:
> > dbd <d...@ieee.org> wrote in message <b433a115-d433-4c85-9810-29fc6da55...@u1g2000pre.googlegroups.com>...
> > > When I hear wavelet advocates make the JPEG claim (and I haven't since
> > > last week), I always have to stop and ponder whether they are
> > > intentionally or ignorantly misleading and what either one must imply
> > > about the rest of their message.
> >
> > Why? Is there a reason why wavelet compression shouldn't be able to outperform DCT compression?
>
> The point isn't about whether wavelet compression works, I think that
> it does, but it has not achieved wide acceptance replacing .jpg in
> photography. The point is about how much one should trust those who
> consistently practice behavior that is either uninformed or dishonest
> in the pursuit of their marketing interests.
====================

I'm still not getting what the dishonest/uninformed behavior is. If someone can demonstrate that wavelet compression works better than DCT compression, then what is uninformed/dishonest about marketing it as such?

On Nov 25, 2:43 pm, "Matt " <x...@whatever.com> wrote:
> dbd <d...@ieee.org> wrote in message <2f5d3115-eede-476a-a251-48b407b7f...@f20g2000prn.googlegroups.com>...
> > On Nov 25, 9:45?am, "Matt " <x...@whatever.com> wrote:
> > > dbd <d...@ieee.org> wrote in message <b433a115-d433-4c85-9810-29fc6da55...@u1g2000pre.googlegroups.com>...
> > > > When I hear wavelet advocates make the JPEG claim (and I haven't since
> > > > last week), I always have to stop and ponder whether they are
> > > > intentionally or ignorantly misleading and what either one must imply
> > > > about the rest of their message.
>
> > > Why? Is there a reason why wavelet compression shouldn't be able to outperform DCT compression?
>
> > The point isn't about whether wavelet compression works, I think that
> > it does, but it has not achieved wide acceptance replacing .jpg in
> > photography. The point is about how much one should trust those who
> > consistently practice behavior that is either uninformed or dishonest
> > in the pursuit of their marketing interests.
>
> ====================
>
> I'm still not getting what the dishonest/uninformed behavior is. If someone can demonstrate that wavelet compression works better than DCT compression, then what is uninformed/dishonest about marketing it as such?

I didn't say 'better'. To discuss that we would have to discuss our
choice of metrics first. Differences in such choices may explain the
failure of wavelet JPEG to penetrate major portions of still camera
usage even though wavelets seem so swell at first glance. A recent
example of the uninformed behavior was last week when a wavelet book
and course author who also thought as you said "JPEG had gone to
wavelets years ago" assumed it was safe to make the false statement
that my still camera used wavelets for compression. It might be a
reasonable caution to carefully examine the predictions made by people
who don't show the ability to speak accurately to where we are now.

Dale B. Dalrymple