🚀 go-pugleaf

In article <Aug.30.14.39.07.1992.23731@pilot.njin.net>, psomu@pilot.njin.net
(Prabhakar Somu) writes:
> 
>    Hi,
> 
>    Can anyone refer me to some literature on particle seperation?
>    Specifically, this is what I am interested in:
> 
>    I have a binary image which has in it some closed regions (blobs).
>    Some of the blobs are touching each other. I would like to
>    seperate the blobs which are touching each other, without altering
>    the blobs that are not touching. Operations such as erosion on the
>    whole image would seperate the touching blobs, but would also
>    change the parameters of the ones that are not touching (for
>    example, the area and perimeter of the isolated blobs would
>    change). In other words, what I am looking for is some
>    non-destructive way of seperation. Any ideas or references would be
>    gladly appreciated. Thanks,

Here's an idea:

1.  Erode the image n times until all blobs are separated.
2.  Change the scale of the image by a factor of 2 but keep 
    the size of the blobs the same.  I.e. make the blobs further
    apart but the same shape.  You can still keep track of their
    centroid location since you know the scale.
3.  Dilate the image n times to try to recover area lost by the erosions. 
4.  Be sure to keep track of seeds that may be created by eroding small
    isolated blobs into nothing.

Another well-established way to separate particles is a technique called
"Watershed Segmentation".  It might be described in Serra or Giardina
and Dougherty.  Also see Image Processing Handbook, J.C. Russ, 1992, CRC
Press.

--
+--------------------------------------------------------------------+
|Dick Wallingford      | Internet: dickw@iastate.edu                 |
|Center for NDE        | Packet  : wd0anb@n0an.ia.usa.na             |
|Iowa State University | "This river don't go to Aintree" - S.H.     | 
+--------------------------------------------------------------------+