region tracking

region tracking

Post by Raghu Hanumanthai » Thu, 30 Sep 1993 07:58:44



I am working on tracking regions in an image. Has anyone worked on such
a thing. My images consists of cells, what i am currently doing is
labelling the image and choose a region by clicking on it. Once the
region of interest is picked now i need to track this region in differrent
images. I have a program called utrack which will highlight a region
given it's number. The cells move around and may change a little bit in
shape. In every image i find the region of interest by finding the center
of mass and matching it with new center of masses of the regions. This
method works as long as the cells do not move so much or the number of
cells are fixed ( no cell creation or deletion ).

If any one has a better tracking algorithm or knows a reference to any
such work,  please send me an e-mail.

thanx
--
------
Raghu Hanumanthaiah
University of Texas at San Antonio

 
 
 

region tracking

Post by Raghu Hanumanthai » Thu, 30 Sep 1993 08:03:06


I am working on tracking regions in an image. Has anyone worked on such
a thing. My images consists of cells, what i am currently doing is
labelling the image and choose a region by clicking on it. Once the
region of interest is picked now i need to track this region in differrent
images. I have a program called utrack which will highlight a region
given it's number. The cells move around and may change a little bit in
shape. In every image i find the region of interest by finding the center
of mass and matching it with new center of masses of the regions. This
method works as long as the cells do not move so much or the number of
cells are fixed ( no cell creation or deletion ).

If any one has a better tracking algorithm or knows a reference to any
such work,  please send me an e-mail.

thanx
--
------
Raghu Hanumanthaiah
University of Texas at San Antonio


 
 
 

region tracking

Post by Mark Wrig » Thu, 30 Sep 1993 20:00:25



>I am working on tracking regions in an image. Has anyone worked on such
>a thing. My images consists of cells, what i am currently doing is
>labelling the image and choose a region by clicking on it. Once the
>region of interest is picked now i need to track this region in differrent
>images. I have a program called utrack which will highlight a region
>given it's number. The cells move around and may change a little bit in
>shape. In every image i find the region of interest by finding the center
>of mass and matching it with new center of masses of the regions. This
>method works as long as the cells do not move so much or the number of
>cells are fixed ( no cell creation or deletion ).

>If any one has a better tracking algorithm or knows a reference to any
>such work,  please send me an e-mail.

>thanx
>--
>------
>Raghu Hanumanthaiah
>University of Texas at San Antonio


A good way to do this is using active contours or "snakes", I know
that Leymarie and Levine have applied this to tracking cells. They had
a paper in PAMI January 92 but this was more on the use of the
technique for obtaining the medial axis but its a start. I have an
internal report February 1990 tr-cim-90-9 from McGill Research center
for intelligent machines " tracking and describing deformable objects
using active contour models", but I am sure they must have published
more widely by now.  As a last point it is computationally more
efficient to use spline based snakes rather than computing the
internal strain energies like early implementations.

-------------------------------------------------------------
Mark Wright
Speech, Vision and Robotics Group
Cambridge University Engineering Department
Trumpington Street, Cambridge, England CB2 1PZ
Tel: +44-223-332754 Fax:+44-223-332662

-------------------------------------------------------------

 
 
 

1. tracking regions

  I am working on a problem related to tracking regions ( blobs ) in a set of
  images. What I am currently doing is
   1) Segment the image in a suitable way, and label it. Each region i want to
          track will have a different region number after labelling them.

   2) Now I store the position of center of masses of all regions in an ascii
   file and when tracking a region ( whose number from the previous image is taken) I try to find the region whose position of center of mass is closest to
   the center of mass of the region i am trying to track.

   But the problem is

How do we decide to track the region when the difference in center of mass
is the same?

Velocity or acceleration does not seem to work, for our experiment due to
the fact that the direction of motion seems to change rapidly.

The problem with the direction is, just because the cell moved in the
south east direction last time does not imply that its going to move in
the south east direction now.

One Solution which I forsee is to have some kind of a logic built in i,e.,
we should have a series of directions which the region, we are trying to track,
could possible take and consult that series ( or table) when in doubt about the
center of mass.

Trying to track by using contours, or the concept of deformable objects does
not apply in our case, since our cells could change shape, size, direction of
motion etc, may be even all at one time.

We are not trying to track something whose shape is defined ( for example a
cell or a ring or nose etc). We are trying to track something purely by the
mechanism which the eye uses to tracking a object purely by vision. That is to follow the regions path of movement. To make my tracking program do that I
think I should some how provide it data about assumed movement.

Please take a look at this and mail me your comments. If any body has done
something similar or has any suggestions about the path I should take or
any references, please e-mail me or post to the newsgroup.

thanx in advance
--
------
Raghu Hanumanthaiah
University of Texas at San Antonio

2. Where to download textures?

3. Tracking regions

4. ___------- VISUALIZATION IN HOLLAND _________-----------

5. Region internals, calculating area of a region?

6. Expressions and particles

7. Help - Collapsing Regions (Was Disappearing Regions)

8. BMP displays in region smaller than specified region -- why?

9. Clip region

10. Drawing diagrams with selective shading of regions

11. Anyone know what Apples region type is ??