AForge.NET

  :: AForge.NET Framework :: Articles :: Forums ::

Execution time

The forum is to discuss topics related to different areas of image processing and computer vision.

Execution time

Postby Shirka45 » Thu Mar 19, 2009 9:52 am

Good morning

I have discovered your image processing library and your demo application. This is a very good job.

So, I tried the following process to estimate execution time :
- Dilate 4 times.

Image format : grayscale (8 bits)
Image size : 1392x1040 pixels.
Processor : Pentium M - 1.6Ghz
Memory : 2 GBytes.

Execution time is around to 3 seconds. It seems me quite slow. Do you have any idea ? :?: Is it normal ?
I tried the same process with Sherlock software (from DALSA-IPD) and execution time is around 20 ms :shock:
Thanks
Best regards
Shirka45
 
Posts: 3
Joined: Thu Mar 19, 2009 9:41 am

Re: Execution time

Postby andrew.kirillov » Thu Mar 19, 2009 10:12 am

Hello,

Well, it looks so for image of such size. At this point implementation of Dilatation routine is generic and works for any type of structuring element. Different optimizations are possible to for specific cases. Which structuring element did you try in Sherlok?

Will put that into TODO list ...
With best regards,
Andrew


Interested in supporting AForge.NET Framework?
User avatar
andrew.kirillov
Site Admin, AForge.NET Developer
 
Posts: 3436
Joined: Fri Jan 23, 2009 9:12 am
Location: UK

Re: Execution time

Postby Shirka45 » Thu Mar 19, 2009 11:22 am

Hello,

Here définition gived by sherlock documentation :
Dilate replaces each pixel with the highest valued pixel in its 3x3 neighborhood. This is a simple form of "thickening" or "opening" where thin or broken bright lines get dilated or thickened. Dilating a dark object thins it.
Sherlock uses IPP Run time may be that's increase performance.

But in all case your image processing library is very easy to use and very useful to develop alogrithm.

Thank again
Best regards.
Shirka45
 
Posts: 3
Joined: Thu Mar 19, 2009 9:41 am

Re: Execution time

Postby andrew.kirillov » Thu Mar 19, 2009 12:04 pm

This is the simplest case. The most simple. It was already in TODO list for a while and will be done for 2.0 version. The routine which is in AForge.NET Framework now is more generic, allowing to use custom structuring elements, what allows solving more tasks.

And of course IPP stuff adds performance boost. AForge.NET Framework does not use IPP, but it may use all cores in the system to boost dilatation for large images (will be in 2.0).
With best regards,
Andrew


Interested in supporting AForge.NET Framework?
User avatar
andrew.kirillov
Site Admin, AForge.NET Developer
 
Posts: 3436
Joined: Fri Jan 23, 2009 9:12 am
Location: UK

Re: Execution time

Postby andrew.kirillov » Fri Mar 20, 2009 1:47 pm

<em>andrew.kirillov</em> wrote:It was already in TODO list for a while and will be done for 2.0 version.


Here they are: Dilatation3x3 and BinaryDilatation3x3.

Will become available in 2.0 beta, which is coming ...
With best regards,
Andrew


Interested in supporting AForge.NET Framework?
User avatar
andrew.kirillov
Site Admin, AForge.NET Developer
 
Posts: 3436
Joined: Fri Jan 23, 2009 9:12 am
Location: UK

Re: Execution time

Postby Shirka45 » Fri Mar 20, 2009 2:51 pm

Hello,

Great, a good news.
Thanks a lot to be willing to lend a sympathetic ear.

Best regards
Shirka45
 
Posts: 3
Joined: Thu Mar 19, 2009 9:41 am




Return to Image Processing and Computer Vision