| hist_nd for creative grid resampling on big arrays [message #48670] |
Tue, 09 May 2006 13:19 |
MarioIncandenza
Messages: 231
Registered: February 2005
|
Senior Member |
|
|
Hey IDL Geniuses!
It starts with one of these:
> help, MyBigGrid
> MYBIGGRID BYT [43200,21600]
That's 30-second data over the whole globe. For my application, I'd
like to have a 1-degree grid where each point is the histogram of
values within that area. With unlimited memory, I could do something
like this:
> Xs1 = (lindgen(43200)/120.)-180+(1/120.); longitudes of columns
> Ys1 = (lindgen(21600)/120.)-90+(1/120.); latitude of rows
> Xs2 = rebin(Xs1,43200,21600); longitude of cells
> Ys2 = rebin(transpose(Ys1),43200,21600); latitude of cells
> EndProduct = hist_nd([[[Xs2]],[[Ys2]],[[MyBigGrid]]],[1.,1.,1],min=[-180. ,-90.,0],max=[179.999,89.999,255])
Due to the size of the arrays, I don't have enough memory to quite pull
this off. But maybe there is some way to trick the histogram routine
into parcelling MyBigGrid according to Xs1|Ys1 without having to put
Xs2|Ys2 into memory. What do you think?
The workaround looks like this:
> EndProduct = lonarr(360,180,256)
> for ix=0,359 do for iy=0,179 do EndProduct[ix,iy,*] = $
> histogram((MyBigGrid[((ix*120):((ix*120)+119)),*])[*,((iy*12 0):((iy*120)
> +119))],min=0,max=100,bin=1)
My workaround has been running for a few minutes now. It probably
worked, but I'd like something more elegant (and faster) as this
problem comes up all the time.
Any histogram masters out there, a faster solution to this general
class of problem would be a big help.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
