[gdal-dev] Why is the nearest neighbour resampling method the " worst interpolation "?

[Dylan Beaudette]

On Friday 28 March 2008, Nikos Alexandris wrote:
> On Tue, 2008-03-25 at 13:26 +0100, Nikos Alexandris wrote:
> [...]
>
> > I will report in case I find something useful trying out the different
> > methods to interpolate the SRTM3 data. But this depends also on the area
> > (the topography) I think.
>
> Just reporting back:
>
> I am going through a series of resampling tests. Not for elevation data
> but for smoothing MOD09 images.
>
> On a subset, South Greece, I tried nn, bilinear, cubic and cubic spline
> at 250, 200, 150 and 100 m for smoothing a  MOD09GQ and the MOD09GQ
> image of the same acquisition date (at 500, 250, 200, 150, 100m).
>
> Although I don't have any statistical measure to proove it, I "see" that
> the best method for smoothing these images ( Blue, Green, Red and NIR
> bands) is cubic resampling. Bilinear and Cubic Spline look too blurred.
>
> BTW:
> Is there any statistical method to evaluate this?
>

The smoothing that you are seeing can be attributed to a general term 
called "change of support". 

To quantify the effect you can compute summary statistics for the original 
data (the cell values only) i.e. mean, median, range, IRQ, min, max, etc. 
Also- an "empirical cumulative distribution function" plot can be helpful 
here.

Next, compute these stats for each resolution, and for each interp method-- 
compare them.

r.resamp.stats in GRASS 6.3 is useful for this.

Cheers,

Dylan


> Kind regards,
>
> Nikos
>
>
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/gdal-dev



-- 
Dylan Beaudette
Soil Resource Laboratory
http://casoilresource.lawr.ucdavis.edu/
University of California at Davis
530.754.7341