From help-request at octave dot org Sat Jan  8 01:45:21 2005
Subject: Re: Efficiently accessing subarrays from an image file...
From: Miroslaw Kwasniak <mirek at zind dot ikem dot pwr dot wroc dot pl>
To: help at octave dot org
Date: Sat, 8 Jan 2005 08:48:31 +0100

On Fri, Jan 07, 2005 at 09:48:21PM -0800, Jonathan Greenberg wrote:
> I was hoping I could get some hints on the most efficient way to perform
> "window" analyses on an image:
> 
> inputs: large multiband image (satellite imagery), binary format
> analysis: perform some repeated algorithm using a moving window of a fixed
> size
>     - example: for each location in the matrix, calculate the variance of
> the 3 x 5 rectangle of matrix elements (e.g. Pixels) surrounding the current
> location

Rule no. 1 - always vectorize your algorithm, don't use loops ;)

You can do something like this: 

image=randn(1000); 
rectangle=ones(5,3); lr=length(rectangle(:)); 
tic,
local_mean = filter2( rectangle/lr, image ); 
local_variance = filter2( rectangle/lr, (image-local_mean).^2 ); 
toc
ans = 0.83769

> What I'm hoping are some ways to speed this up for use in a MP environment
> -- what optimizations can help me do this?

Look around MPI Toolbox for Octave: http://atc.ugr.es/javier-bin/mpitb

Mirek



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------