ImagePro>clustering or maximum lihelihood support
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
ImagePro>clustering or maximum lihelihood support
- To: <imagepro-users[at]lists.mediacy.com> (Image-Pro Plus Users Email List)
- Subject: ImagePro>clustering or maximum lihelihood support
- From: "Michael Plonski" <mplonski[at]cox.net>
- Date: Mon, 4 Mar 2002 10:07:30 -0800
- List-Archive: <http://lists.mediacy.com:80/Lists/imagepro-users/List.html>
- List-Unsubscribe: <mailto:imagepro-users-off@lists.mediacy.com>
- Reply-To: <imagepro-users[at]lists.mediacy.com> (Image-Pro Plus Users Email List)
- Sender: <imagepro-users[at]lists.mediacy.com> (Image-Pro Plus Users Email List)
Is there any way to get to the internal clustering routines in image pro.
FOr example the Auto count function has to do some clustering on the
histogram to decide where to put the breakpoint. The autoclassification has
to use a 3d clustering routine to classify objects.
THe problem I am having is trying to get clean boundaries so everything in
the image is classified or segmented out. I only have about 5 types of
things in the image. I want the whole image to be forced into 1 of those 5
categories. If I pick sample data for the segmentation colors based on the
color-cube segmentation - I can get a reasonable partial segmentation. The
problem is I have to manually pick the colors - the second problem is the
segmentation routines do not force every pixel into 1 and only 1 class. I
can't really store off the color cube segmentation file as there are a lot
of lighting issues where the colors shifts around a little from image to
image.
The classic way to do this is to do some unsupervised clustering to find the
5 cluster centers and then some maximum likelihood classification on each
pixel based which cluster it is closest too. The maximum likelihood
classification is easy to implement if I can get some routine to hand me
back gaussian clusters (means and std. devs). Ideally I would like to be
able to give it the whole rgb image to find the clusters. I can probabily
get things to work just using the r & g channel as the blue doesn't provide
much variance. Even simple routines to find gaussian clusters on a simple
1D histogram would be useful. Any suggestions:
Other useful things would be:
1) principal component analysis
2) region growing tools - so I can take the partial segmentations made by
image pro and push them out to a more logical boundary that includes the
whole object.
I looked in the SDK, HIl and HAIL, but there doesn't seem to be any routines
for this.
Thanks for all the helpful comments on my last question on a macro file
including a macro file.
Mike Plonski
***********************************************************
Need an Image-Pro macro or driver? Find it at http://www.Solutions-Zone.com
Got an Image-Pro macro or driver? Add it to http://www.Solutions-Zone.com
***********************************************************
This message is sent to you because you are subscribed to <imagepro-users@lists.mediacy.com>.
To unsubscribe, email to: <imagepro-users-off@lists.mediacy.com>
To switch to the DIGEST mode, email to <imagepro-users-digest@lists.mediacy.com>
To switch to the INDEX mode, email to <imagepro-users-index@lists.mediacy.com>
Send administrative queries to <imagepro-users-request@lists.mediacy.com>
To subscribe or unsubscribe visit http://www.solutions-zone.com/ipednld/subscriber.asp