[Insight-developers] Statistics: Mean shift mode estimator

[Miller, James V (Research)]

Jisung, 

I would suggest doing a mean shift from a histogram as well.

Jim

-----Original Message-----
From: Jisung Kim [mailto:bahrahm at yahoo . com]
Sent: Friday, June 13, 2003 9:31 AM
To: Miller, James V (Research); Insight-developers (E-mail)
Subject: RE: [Insight-developers] Statistics: Mean shift mode estimator


Thanks Jim.

I would first try implement the mean shift mode
seeking technique using k-d tree. And then I would
implement a smoothing filter and a segmentation method
using the mode seeking technique previously
implemented.

And then I might extend the mode seeking technique
with other range search techniques. 

Does it sound OK?

--- "Miller, James V (Research)" <millerjv at crd . ge . com>
wrote:
> You go Jisung!
> 
> A couple of things to think about:
> 
> 1) Mean shift could be applied to histograms to
> identify the modes or
> classes
> 2) Mean shift can be applied to scattered "points"
> in a high dimensional
> space
> 3) Mean shift can be applied to filter (smooth) an
> image
> 4) Mean shift can be used for image segmentation
> 5) Mean shift can be used for "tracking" an object
> from slice to slice or
> frame to frame.  It has been used a lot for people
> trackers.  But could also
> be applied to vascular tracking, etc.
> 
> #1 and #2 are general statistical approaches that
> can be used by many
> applications.  The rest are applications of mean
> shift that may use the
> techniques developed in #1 and #2.
> 
> When applied to very high dimensional spaces (say
> 50), it sounds like people
> stray from k-d trees and use "locality sensity
> hashing".
> 
> Jim
> 
> 
> -----Original Message-----
> From: Jisung Kim [mailto:bahrahm at yahoo . com]
> Sent: Thursday, June 12, 2003 4:25 PM
> To: Miller, James V (Research); Insight-developers
> (E-mail)
> Subject: Re: [Insight-developers] Statistics: Mean
> shift mode estimator
> 
> 
> Hi. I read the paper. The filtering and segmentation
> methods seem to be a good example of kernel density
> techniques. My k-d tree implementation probably
> accelerate the mean shift vector calculation part
> (range search). Even the non-terminal nodes of the
> k-d
> tree store the vector sum and the number of
> measurement vectors below each node. If nobody wants
> to pursuit these methods, I want to try.
> 
> 
> --- "Miller, James V (Research)"
> <millerjv at crd . ge . com>
> wrote:
> > Does anyone have plans to add a "mean shift" mode
> > estimator to ITK?
> >  
> > http://www . computer . org/tpami/tp1995/i0790abs . htm
> >
> <http://www . computer . org/tpami/tp1995/i0790abs . htm> 
> >  
> >
>
http://www . caip . rutgers . edu/~comanici/Papers/MsAnalysis . pdf
> >
>
<http://www . caip . rutgers . edu/~comanici/Papers/MsAnalysis . pdf>
> > 
> >  
> >  
> >  
> > 
> > Jim Miller 
> > _____________________________________
> > Visualization & Computer Vision
> > GE Research
> > Bldg. KW, Room C218B
> > P.O. Box 8, Schenectady NY 12301
> > 
> > millerjv at research . ge . com
> > <mailto:millerjv at research . ge . com> 
> > 
> > james . miller at research . ge . com
> > (518) 387-4005, Dial Comm: 8*833-4005, 
> > Cell: (518) 505-7065, Fax: (518) 387-6981 
> > 
> >  
> > 
> 
> 
> =====
> Jisung Kim
> bahrahm at yahoo . com
> 106 Mason Farm Rd.
> 129 Radiology Research Lab., CB# 7515
> Univ. of North Carolina at Chapel Hill
> Chapel Hill, NC 27599-7515
> 
> __________________________________
> Do you Yahoo!?
> Yahoo! Calendar - Free online calendar with sync to
> Outlook(TM).
> http://calendar . yahoo . com
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at www . itk . org
>
http://www . itk . org/mailman/listinfo/insight-developers


=====
Jisung Kim
bahrahm at yahoo . com
106 Mason Farm Rd.
129 Radiology Research Lab., CB# 7515
Univ. of North Carolina at Chapel Hill
Chapel Hill, NC 27599-7515

__________________________________
Do you Yahoo!?
Yahoo! Calendar - Free online calendar with sync to Outlook(TM).
http://calendar . yahoo . com