Path: news.mathworks.com!not-for-mail
From: <HIDDEN>
Newsgroups: comp.soft-sys.matlab
Subject: Re: conv2 of a large image
Date: Thu, 5 Mar 2009 13:35:02 +0000 (UTC)
Organization: Queen's University
Lines: 3
Message-ID: <gooke6$iek$1@fred.mathworks.com>
References: <gjhd65$7b2$1@fred.mathworks.com> <6cc68097-2e74-4872-bd05-e551acf0db01@u14g2000yqg.googlegroups.com> <2988496d-4770-4710-9213-6596da54c8ab@f11g2000vbf.googlegroups.com> <gjjq7p$6di$1@fred.mathworks.com> <gjjuvl$lbo$1@fred.mathworks.com> <gjk0v1$49g$1@fred.mathworks.com> <gjk1ef$4j9$1@fred.mathworks.com> <gjk32t$907$1@fred.mathworks.com>
Reply-To: <HIDDEN>
NNTP-Posting-Host: webapp-05-blr.mathworks.com
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-Trace: fred.mathworks.com 1236260102 18900 172.30.248.35 (5 Mar 2009 13:35:02 GMT)
X-Complaints-To: news@mathworks.com
NNTP-Posting-Date: Thu, 5 Mar 2009 13:35:02 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 943450
Xref: news.mathworks.com comp.soft-sys.matlab:522722


It may not help if it's a memory issue, but if you have the Image Processing Toolbox, imfilter takes advantage of zeros in the convolution kernel, which speeds up computation of the stationary/undecimated wavelet transform enormously.  As far as I can tell, conv2 doesn't, and so can be extremely slow.  I'm using R2008a - I don't know if this has been sorted out in a later release.

Pete (who has just been writing his own swt2 alternative using imfilter... it seems to be worth doing)