Tom -
> -----Original Message-----
> From: Tom Whittaker [mailto:tomw@xxxxxxxxxxxxx]
> Sent: 20. mars 2003 15:23
> To: Kjell Roang
> Cc: visad-list@xxxxxxxxxxxxx; Tor Christian Bekkvik; Richard D Gonzalez;
> Oyvind Breivik
> Subject: Re: Resampling from lat/lon to projection
>
>
> Kjell:
>
> > The reason I want to do the transformations (projections)
> ourself, is to be
> > able to support more output projections (to the user). We will
> probably use
> > the projection package in geotools. The interpolation we gladly
> let VisAd
> > handle.
>
> If 'geotools' can convert an array of points from one projection to
> another, you could create a VisAD CoordinateSystem that simply
> interfaces with that.
>
This could be possible.
> > The JGrib package now handles Lambert and Stereographics grids
> and convert
> > it to lat/long. Probably we will be able to support more
> coordinate systems
> > in the future. (JGrib is an open source project at
> > http://jgrib.sourceforge.net/ ).
>
> I would like to continue to expand the projections that are supported in
> GRIBCooridnateSystem, especially when it comes to the "grid number" ones
> - right now it only knows about 201,202,203,211,212,215,236.
>
> >
> > Another question related to interpolation: If one wants to
> interpolate to a
> > pixel sized grid, is it more efficient to first interpolate to
> a "coarse"
> > rectangular grid (both geometry and topology) and from there
> interpolate to
> > the pixel sized grid?
>
> If the reason you want to interpolate to a "pixel sized grid" is to be
> able to show a grid as an image, then I would suggest just letting VisAD
> do the display using your coarse grid. If there is some other reason,
> I'd be curious to know...?
>
> I would avoid a double interpolation for the reason that you would be
> essentially filtering your data twice and might introduce side effects
> that are unpleasant. It certainly could be more computationally
> efficient to initially interpolate to every N-th point in your "pixel
> grid" and then do a linear interpolation in between these points...but
> it does beg the question about why you'd like to use a 'pixel sized
> grid' in the first place.
>
> tom
Well. The coarse (20 x 20 pixel) grid could be used as input to contouring.
The pixel sized grid should be displayed as an image. The reasons to do it
in two step was (as you said), to save computation time. One might not have
artifacts if the coarse grid is finer than the original (lat, lon) grid.
Kjell R
>
> --
> Tom Whittaker (tomw@xxxxxxxxxxxxx)
> University of Wisconsin-Madison
> Space Science and Engineering Center
> Cooperative Institute for Meteorological Satellite Studies
> Phone/VoiceMail: 608.262.2759
>
