Olympus-OM
[Top] [All Lists]

[OM] Autostitch: Resolving the dearth of wide angles for DSLR's

Subject: [OM] Autostitch: Resolving the dearth of wide angles for DSLR's
From: Chuck Norcutt <chucknorcutt@xxxxxxxxxxx>
Date: Wed, 02 Feb 2005 08:26:39 -0500
The last digest arrived with much discussion around Barry's question 
regarding comparison of the E-1 and E-300.  Moose extolled the virtues 
of the C*n*n 300D and CH followed up by saying he had sold his 10D (a 
very similar camera) in favor of the E-1.  CH said this was due to the 
paucity of lenses able to deliver a truly wide angle view.  This, of 
course, is due to the less than full frame sensor on most DSLRs.

Not available to CH at the time with his 10D but new with the 300D and 
20D is a newly designed lens mount and mirror box that allows smaller, 
shorter focal length lenses ala the Olympus 4/3 solution.  So, C*n*n now 
offers a 10-22mm lens which is basically the same range as the Oly 
11-22mm 4/3 lens.  Both, however, cost about USD 800 at B&H.

My own DSLR solution is likely to be the 20D.  That is based on a 
different set of criteria than others have expressed here.  My most 
important criteria are low light autofocus, very low noise at high ISO 
and versatile daylight fill flash which calls for flash sync as fast as 
possible.

But I'm also faced with the wide angle problem.  The 20D will allow the 
use of the 10-22mm but I'm in no hurry to rush out and spend USD 800. 
But there is another way.  I was reminded to make this post by this 
panoramic shot that Chris posted of King's College made from 3 handheld 
E-1 images. <http://www.myfourthirds.com/document.php?id=6695>

If you have a perpetual, wide angle view of the world and use the OM 
18mm as your normal lens then building panoramas for every shot you take 
will not be attractive or even possible.  However, if you take really 
wide shots only occasionally then building panoramas is a solution. 
Especially if you have software (not PhotoShop) that makes it really, 
really easy to do.

For easy panoramas there is nothing else even approaching the ease of 
use of an experimental application called "Autostitch" which can be 
obtained free at <http://www.autostitch.net/>

Autostitch is truly almost magic.  In the case of Chris's King's College 
panorama he would only have needed to put the three images into a single 
folder and tell Autostitch the name of the folder.  Autostich would then 
automatically figure out how the images relate to each other, stitch 
them together and blend the stitched area if the exposure or lighting is 
slightly different.  In fact, if the folder contains images that are not 
part of the panorama, Autostitch is smart enough to figure that out as 
well but that will certainly slow it down.  The only thing left for you 
to do is make a final crop to remove the irregular outer bondaries of 
the final image.

Until the most recent release, Autostitch was limited by only allowing a 
very low quality JPEG as output.  The latest release has removed this 
restriction  by allowing user control of the quality level.  In fact, a 
friend was able to use it recently in constructing a two image panorama 
of the inside of a church with the final print being about 24" wide.

The enabling technology behind Autostitch is some patented computer 
vision, object recognition work done at the Artificial Intelligence Lab 
at the University of British Columbia, Canada.  Totally amazing, science 
fiction level stuff.  I didn't believe it myself until I tried it.

Just note that it is not foolproof.  Relatively featureless areas can 
cause it problems if there is not enough overlap.  Try it, you'll like 
it.  Maybe Chris will be inspired to try it with his same three images 
and report back on how it did.

ps:  Welcome back Barry!

Chuck Norcutt


==============================================
List usage info:     http://www.zuikoholic.com
List nannies:        olympusadmin@xxxxxxxxxx
==============================================

<Prev in Thread] Current Thread [Next in Thread>
Sponsored by Tako
Impressum | Datenschutz