Re: Rendering an image

• To: Don Murray <dmurray@xxxxxxxxxxxxxxxx>
• Subject: Re: Rendering an image
• From: Bill Hibbard <billh@xxxxxxxxxxxxx>
• Date: Mon, 29 Dec 2003 08:50:07 -0600 (CST)

Hi Don,

> > VisAD does make a geometry for an image, but also an array
> > of pixels for texture mapping. Shapes can include the
> > geometry but not the texture map array. You can reduce an
> > iamge to pure geometry, but it requires two triangles per
> > pixel which will be slow for large images. You can
> > experiment with speed and memory use by disabling texture
> > mapping off when you display an image.
>
> Where is the geometry made (class/method)?  Most of the
> images I'm dealing with are small (legends, icons), so I
> don't think speed will be an issue.

VisAD turns images into geometry in the make2DGeometry()
method of Gridded3DSet.java, which returns the geometry
as a VisADGeometryArray (usually a VisADTriangleStripArray).
However, the Shadow*Type classes do some related computation
like deriving a Gridded3DSet in Cartesian display
coordinates from the Gridded2DSet domain Set of the image
FlatField, and making the 'byte[][] color_values' argument
to make2DGeometry() from the range values of the image.

I don't *think* it would be too hard for you to construct a
Gridded3DSet from your image (maybe just set all ZAxis
values to 0.0, and scale the XAxis and YAxis values
appropriately), and construct a 'byte[][] color_values'
array to pass to the make2DGeometry() method of the
Gridded3DSet you construct. I suggest passing false to the
'boolean indexed' argument - otherwise it will return a
VisADIndexedTriangleStripArray which is needless complexity.

Good luck,
Bill

• References:
  • Re: Rendering an image
    • From: Don Murray