RE: [GMLJP2] NetCDF <--> GML/JPEG2000

• To: "John Caron" <caron@xxxxxxxxxxxxxxxx>
• Subject: RE: [GMLJP2] NetCDF <--> GML/JPEG2000
• From: "Ron Lake" <rlake@xxxxxxxxxxxxx>
• Date: Mon, 16 May 2005 10:19:30 -0700

Hi John:

See response below in-line.

Cheers

Ron

JPEG 2000 offers support for integers only. It is not clear that this
is a serious constraint since any observation must have only finite
precision and hence can be scaled. This may be a pain, but it is
common in all measurement systems.

Is there a standard way to specify scale and offset values?

Not presently - but this could be added NOW in an application schema and then 
added into a future version of GML valueObjects.

*Dimensionality of the coverage:*

* *

In general a coverage can be seen as a function *X* = f(*Y*). The
dimension of X can be some N (N=0,1,2, 3, 4 .. ). With the current
GMLinJP2K specification, the dimensionality of Y is 2 - that is at
each point *p* on a 2D surface (e.g. surface of the earth) we can have
vector quantities *X(p)* = (X1, X2, .. Xn) where each Xi will be in a
different JPEG 2000 codestream.

How do you specify what the Xi are? is that in the GML or in the JPEG ?

Can you specify different units for each ?

Yes this can be specified now in GML including different units for each.

So for representation of measurement information the restriction of
GMLJP2K is that it does not allow the points *p *to be in a volume -
note that the surface on which *p* lies can be in 3-space however.
This is not really a restriction of JPEG 2000, however, and I think we
can extend the GML description to allow the description of functions
(as above) over multi-dimensional "volumes".

If JPEG is 2D, doesnt that limit it to domain dimensionality = 2 ? Is there a 
way around that?

The domain is 2D yes - but the geometric points of the domain may be embedded 
in 3D now.

Thanks for all the info, its very interesting !