| Re: four corners for lat & long pixels [message #54779 is a reply to message #54753] |
Tue, 03 July 2007 03:40  |
James Kuyper
Messages: 425
Registered: March 2000
|
Senior Member |
|
|
titan wrote:
> I have a radar image and I know latitude and longitude of each pixels
> of this image
> I wonder how to get latitude and longitude of the four corners
> "surrounding" each pixels of the image
A good approximation would simply assign to each corner a latittude
equal to the average of the latitudes of pixels that meet at that
corner, and the same for longitudes. For the pixel corners on the
outer edge of the image, use bilinear extrapolation from nearby
pixels. Note that it takes special handling to make this approach work
properly if your image crosses the meridian at 180W = 180E. This
approach will unavoidably produce bad results near the north and south
poles.
For a more sophistocated analysis, you could use 2-D spline
interpolation. However, if you're going to get that sophisticated
about it, you should convert your latitudes and longitudes into map
coordinates for a map projection that reflects, with reasonable
accuracy, the way in which your image was collected. An ideal map
projection for your particular image would convert the latitudes and
longitudes of each row or column of your image into a straight line
of evenly spaced dots. Interpolate/extrapolate the projected
coordinates, just as described above for the latitude and longitude,
to get projection coordinates for pixel corners. Then invert the map
projection to get latitude and longitude for those corners. In
addition to getting slightly better accuracy, interpolation with a
well-chosen map projection will also automatically avoid the problems
at 180W=180E, and at the north and south poles, without requiring any
special handling in your code - the special handling is hidden inside
the map projection routines.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
