[Insight-users] Repost: Performance comparison

Miller, James V (Research) millerjv at crd.ge.com
Fri Sep 3 10:03:39 EDT 2004


There is no report on how the speed of ITK compares to other toolkits.
I suspect that in most cases ITK will be slower, partly due to the 
N-dimensional nature of ITK implementations and partly due to fact 
that performance has only been optimized as people have identified 
issues. 

For instance, the Bilateral filter in ITK is probably slower than the 
one in OpenCV because OpenCV always used a 3x3 kernel for the Bilateral
filter.

On a multiple processor system, I would guess ITK algorithms may outperform
MATLAB because an ITK algorithm may take advantage of the multiple
processors.

Jim


-----Original Message-----
From: Mark Asbach [mailto:mark.asbach at post.rwth-aachen.de]
Sent: Friday, September 03, 2004 8:12 AM
To: insight-users at itk.org
Subject: [Insight-users] Repost: Performance comparison


Sorry: This is a repost. Unfortunately the original message went out BEFORE
I was finally subscribed to the list. Since I haven't seen the original
message but others that went over the mailing list, I decided to send it
again (hope it's okay).


Hi ITK users,

at first: Hi all. I'm a CS student from Aachen (Germany) starting with my
PhD studies. I'll be working on the field of image analysis and that got me
interested in ITK.

What I would like to know (but didn't find via Google) is, what processing
speed / performance I can expect from ITK in relation to the hardware it
runs on and how that performance compares with - as an example - OpenCV or
Matlab where algorithms exist in each toolset. Is there some extensive (or
even a small) test report out somewhere?

If not, I would like to do one, but I'd need some ideas on reasonable
scenarios. For example I'd appreciate some hints which algorithms are mostly
dependant on

- processor speed (integer, float, vectorization?)
- memory throughput
- cache
- other factors (which?)

to put a setup together that really helps people making good estimates on
which software/hardware combination would lead to which execution time.

I already have some experience with VTK and even did some investigations on
execution speed on our university's machines (Sun Fire vs. x86 PCs vs.
PowerPC vs. MIPS, KCC vs. GCC vs. WSpro and different compiler settings),
that proved major differences in execution time (up to factor 10 on the same
machine with different compilers / compiler settings). However, these test
were for internal purposes and I only cared for my projects needs.

Thanks in advance,

Mark


_______________________________________________
Insight-users mailing list
Insight-users at itk.org
http://www.itk.org/mailman/listinfo/insight-users


More information about the Insight-users mailing list