Bilinear Interpolation: Extract Values to Points

[CraigEissler]

I am attempting to use the Spatial Analyst Tool in the Extraction Toolset called 'Extract Values to Points'.

One option is to assign the Cell Values to the Input Points using "Bilinear Interpolation".

I have read about this in the Resampling Help Doc, though this seems like it would be the reverse situation. Either way, it is not clear to me which Cells will be used and how. The Tool itself Help Doc is not clear to me either.

Is it going to use only the 4 "nearest" Cell center Values? If so, I assume that includes the one it falls within..?

Or is it going to use the 4 "adjacent" Cell center Values? If so, I assume those are the Cells that share a side with the Cell containing the Point..? if so, I assume that this is really going to be using 5 Cell Values to include the one it falls within..?

Or, is this Tool not restricted to 4 or 5 Cell Values, and will use all the "neighboring"cells which will amount to 8 or 9 Cell Values? ...In other words, the diagonal Cells will also be included since they share a vertex with the Cell containing the Point.

In either case, are these Cell Values "distance-weighted"? If so, is this technique explained how it weights..?

Thanks!

[curtvprice]

I am attempting to use the Spatial Analyst Tool in the Extraction Toolset called 'Extract Values to Points'.

One option is to assign the Cell Values to the Input Points using "Bilinear Interpolation".

In either case, are these Cell Values "distance-weighted"? If so, is this technique explained how it weights..?

Surfaces are handled with the values treated as points at the center of each cell.

Bilinear interpolation will use the four closest cell centers, distance weighted (1/x, not 1/x²)

ArcGIS Desktop 10.2 Help: Cell size and resampling in analysis

[CraigEissler]

Thanks -- yes, I read and understood the Help Doc on the Bilinear Resampling interpolation technique. ...So I assume you are indicating that this behavior is the same exact same thing as the Bilinear Extraction (Cell Value to Input Points) interpolation technique..?

That was the underlying confusion I had because the Help Doc for the Extraction Tool did not get into those specifics.

[CraigEissler]

...I guess I should do a better job of just getting straight to the question (although at least we have some context now):

Is the Bilinear 'Resampling' interpolation technique use the exact same methodology as the Bilinear 'Extraction (Cell Value to Input Points)' interpolation technique..?

Thanks!

[curtvprice]

[Does] the Bilinear 'Resampling' interpolation technique use the exact same methodology as the Bilinear 'Extraction (Cell Value to Input Points)' interpolation technique?

If you see otherwise, it's worth logging a bug. Bilinear interpolation is pretty straightforward; all the tools that use it (Resample, Project Raster, Sample, Extract Values To Points, to name a few) should work the same.

What is the reason for this question? Did you see some results that looked incorrect? You can easily measure the distances with the mouse in ArcMap and do the math yourself if you have any doubts.

[CraigEissler]

Thanks for your reply.

Yeah, the reason I ask is because Bilinear Interpolation for Resampling clearly states that the calculation will be based on a "weighted-distance of the 4 nearest Input Cell Centers".

But for Bilinear Interpolation "option" for Extraction [Cell Value to Input Points] I only see brief mention of this in the Help Doc and Tool Help that states the calculation will be based on the "adjacent Cells w/ valid Values". ...No mention of being "distance-weighted" and no mention of "4 closest Inputs".

...And so if Extraction really uses "adjacent cells", then what constitutes adjacency?
a) the 4 Cells sharing a Cell side
b) those 4, plus the Cell the Input Point is within, making it 5 Cells
c) the 8 Cells totally surrounding the Cell
d) those 8, plus the Cell the Input Point is within, making it 9 Cells

[curtvprice]

I still don't see why it would be anything else but the four closest cell centroids, as it is described in the help in the only place bilinear interpolation is described. Bilinear interpolation is distance weighted -- by definition! That's the "linear" part.

Again, have you seen any evidence that this is not the case?

[CraigEissler]

You're probably right, but I guess I'll just need to try it and do the calculations myself to confirm or deny.

On a related note, I just read the Bilinear interpolation description for 'Project Raster' and it states that the calculation is based on the "surrounding cells" -- even more ambiguous! ;0 haha

[curtvprice]

You're probably right, but I guess I'll just need to try it and do the calculations myself to confirm or deny.

"Trust but verify" applies to geoprocessing as well as arms control.

the Bilinear interpolation description for 'Project Raster' and it states that the calculation is based on the "surrounding cells" -- even more ambiguous! ;0 haha

I highly recommend that you click the Feedback button in the web help article and let them know - this goes directly to Esri's help authors and I've seen my own comments directly improve the help.

[CraigEissler]

Ok I will -- thanks for all your help