[Geotiff] Relationship Between Model Tie Point and Raster Space

[Tillerson, Clint]

Hello, 

I'm trying to understand the relationship between the model tie point
(upper left corner of the geographic domain) and the raster space
(PixelIsPoint vs. PixelIsArea) to ensure I am properly assigning
geographic coordinates to the data points in the file.  I am extracting
elevation data (NED or SRTM) for preprocessing for another program.

If someone could verify my interpretation is correct...

Suppose:

* model tie point = (0, 0, 0, -112.0, 39.0, 0),
* raster type = PixelIsArea
* pixel scale = 1 arc-second in X and Y directions
(0.000277777777777778, 0.000277777777777778, 0)

I interpret this to mean the upper left corner of the first pixel (0,0)
is located at lon = -112.0 and lat = 39.0.  However, the elevation for
that first pixel, bound by points (0,0) and (1,1) is represented at the
center of the pixel derived by adding 1/2 arc-second to the longitude
and subtracting 1/2 arc-second to the latitude (i.e., adjust lat/lon to
center of pixel).

Similarly, if the raster values for the model tie point were 0.5, 0.5,
0, this would mean that lat/lon supplied in the model tie point already
represents the center of the pixel and are the coordinates for the first
elevation.

In the original example above, if the raster type were PixelIsPoint,
then the first elevation is represented by the coordinates specified in
the model tie point (i.e, do not adjust the coordinates by 1/2 pixel).
If the raster values were 0.5, 0.5, I would need to derive the
coordinates for the first elevation by subtracting 1/2 arc-second from
the longitude and adding 1/2 arc-second to the latitude (i.e, adjust
geographic coordinates to raster point (0, 0).

Could someone verify my interpretation correct or show me the error of
my ways?  

Thanks so much

Clint



 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.maptools.org/pipermail/geotiff/attachments/20081031/f96613de/attachment-0001.html