| interpolation question [message #48395] |
Wed, 19 April 2006 21:25  |
chen123.dian
Messages: 4
Registered: April 2006
|
Junior Member |
|
|
Hi, I want to use nearest-neighbor method to interpolate 2-D data.
After reading some posters here and idl references, I am still confused
how to do. I use a 1-D data to explain my problem. For example, there
is x-arrary, [1.0, 2.9, 5.1], the corresponding y-arrary is [10, 50,
80]. Now we want to get the y-values at x-value [0.5, 2.6, 3.1, 4.6]
using nearest-neighbor method. The result of y-values should be [10,
50, 50, 80]. Can the IDL function of interpolate.pro do this work? The
input of interpolate.pro, X,Y,Z, arrays of numeric type containing the
locations for which interpolated are desired, is very confuse. Is there
any function like the function interp1 (interp2 for 2-D data) for
matlab? Like, y_values=interp1(x, y, x-values, 'nearest', 'extrap').
Thanks in advance.
Chen
|
|
|
|
| Re: interpolation question [message #48484 is a reply to message #48395] |
Thu, 20 April 2006 18:57  |
JD Smith
Messages: 850
Registered: December 1999
|
Senior Member |
|
|
On Wed, 19 Apr 2006 21:25:15 -0700, chen123.dian wrote:
> Hi, I want to use nearest-neighbor method to interpolate 2-D data.
> After reading some posters here and idl references, I am still confused
> how to do. I use a 1-D data to explain my problem. For example, there
> is x-arrary, [1.0, 2.9, 5.1], the corresponding y-arrary is [10, 50,
> 80]. Now we want to get the y-values at x-value [0.5, 2.6, 3.1, 4.6]
> using nearest-neighbor method. The result of y-values should be [10,
> 50, 50, 80]. Can the IDL function of interpolate.pro do this work? The
> input of interpolate.pro, X,Y,Z, arrays of numeric type containing the
> locations for which interpolated are desired, is very confuse. Is there
> any function like the function interp1 (interp2 for 2-D data) for
> matlab? Like, y_values=interp1(x, y, x-values, 'nearest', 'extrap').
Doesn't it depend on how your 2D array is sorted? Are the values
along rows and columns always monotonically increasing, e.g.:
0.1 1.2 2.4
0.2 1.3 2.5
...........
Otherwise, the problem is undefined.
That said, I think you are not asking a value lookup problem at all,
but a simple 2D interpolation on a regular grid, where instead of
linear nearest neighbor intepolation, you want "plain" nearest
neihbor. VALUE_LOCATE can do this for you, assuming "nearest in x"
and "nearest in y" are separable.. Let 'yarray' be your data array
you'd like to intepolate on, and 'curve_x' be the monotonic, but
non-regular grid of x positions, and 'curve_y' similarly, e.g.:
IDL> yarray=randomu(sd,100,100)
IDL> curvex=findgen(100)^2
IDL> curvey=alog10(1.+findgen(100))
such that in the 2D 'yarray', all rows have the same X vector
coordinates 'curvex', and all the columns the same Y vector 'curvey'.
Now you have two vectors describing the X,Y positions in this funny
space where you would like nearest neighbor interpolates:.
IDL> x=(randomu(sd,10)*100)^2
IDL> print,x
6544.61 2552.47 2239.70 9695.07 144.029 919.768
933.261 3231.17 151.407 761.378
IDL> y=alog10(1.+randomu(sd,10)*100)
IDL> print,y
1.63828 1.52555 1.80391 0.983095 1.26320 1.86245
1.79494 1.67543 1.89285 1.37616
IDL> xpos=value_locate(curvex,x)
IDL> xpos=round(xpos+(x-curvex[xpos])/(curvex[xpos+1]-curvex[xpos ]))
IDL> ypos=value_locate(curvey,y)
IDL> ypos=round(ypos+(y-curvey[ypos])/(curvey[ypos+1]-curvey[ypos ]))
IDL> interp=yarray[xpos,ypos]
JD
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
