Re: images and efficieny

• To: "Campbell, Garrett" <Campbell@xxxxxxxxxxxxxxxxxx>
• Subject: Re: images and efficieny
• From: "From: Bill Hibbard <hibbard@xxxxxxxxxxxxxxxxx>" <billh@xxxxxxxxxxxxx>
• Date: Mon, 26 Nov 2001 14:38:45 -0600 (CST)

Hi Garrett,

The default for FlatFields (used to store images) is to
use 4-byte floats for values. However, the full FlatField
constructor signature includes an argument 'Set[] range_sets'.
If you use a Linear1DSet with no more than 255 values, then
values for the corresponding range RealType will be stored
in bytes.

I don't know if the McIDAS AREA file adapter uses this or not.
I don't think the Plain netCDF adapter uses it. You might be
able to read a files into a FlatField with floats, then copy
it into a FlatField with bytes.

Good luck,
Bill

On Mon, 26 Nov 2001, Campbell, Garrett wrote:

> I need to load 4 to 6 full disk satellite images at 5 km resolution and
> then compare
> patches of the images.  (2500 X 2500 pixels).
> I can load two of these, but then the memory overflows with mx384.  It
> appears that
> loading replaces all the 8 bit pixels with reals or doubles.  Is there a
> way to
> keep these arrays as bytes?
> 
> The comparisons occur on small areas within the images (100X100km or
> 20X20 pixels)
> which are extracted with image.resample(domain..).  That seems to work.
> But I need to step across the full disk image, resampling each piece so
> I can not afford
> to reload the 6 images for each matching step (each cloud).
> 
> Any suggestions would be appreciated.
> 
> G. Garrett Campbell Ph. D.
> Cooperative Institute for Research in the Atmosphere
> Colorado State University
> Ft. Collins, CO, USA
> 970 491 8497
> campbell@xxxxxxxxxxxxxxxxxx 
> 

• Follow-Ups:
  • Re: images and efficieny
    • From: Don Murray

• References:
  • images and efficieny
    • From: Campbell, Garrett