[R-sig-Geo] ENVI data and R

[Eric B. Peterson]

I don't know if this will exactly apply to your case, but you mention pixel by pixel analysis.  If that is really all you need and there are no texture/neighborhood analyses going on, then you could simply open the file and use readBin to grab an appropriately sized chunk, analyze it, writeBin to your output file, then move on.  

Right now I'm exporting about a dozen rasters from ArcMap to binary floating point files (.flt), reading one row at a time into R with readBin(filehandle, double(), size=4, n=ncols) for each raster, and looping through all rows to output a manually constructed classification.  I've done similar things with ENVI files in the past (which when not compressed are simple, straight-forward binary files).  I've just used single-band files for simplicity sake;  for multi-band images you would need to manage the interleaving pattern (e.g. BSQ, BIL).

-Eric

------------------------------

Message: 10
Date: Tue, 29 Jul 2008 16:15:34 -0400
From: "Guy Serbin" <guy.serbin at gmail.com>
Subject: Re: [R-sig-Geo] ENVI data and R
To: r-sig-geo at stat.math.ethz.ch 
Message-ID:
	<1ed1e01b0807291315p2b556089o7bfc03f2d3d4167f at mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1

Thank you all for the help- I successfully read an image into R using
these methods.

I did, however, encounter some problems when loading a hyperspectral
image cube into R as it was unable to allocate the 2.9 GB of volatile
memory that it needed.

Is there a way to improve memory management by R, so that it only
reads in the data when actually needed for processing, e.g., only read
in the bands I need, or conversely read in spectra on a per-pixel
basis?

Guy



---
Eric B. Peterson, Ph.D.
Vegetation Ecologist / Data Manager
California Native Plant Society
(916) 322-2926 (desk)
(775) 750-4628 (cell)