Re: face recognition - NewbieQuestion

WSiler (wsiler@aol.com)
Tue, 6 Oct 1998 04:06:46 +0200 (MET DST)

The standard steps in processing an image of your type are (1) segmentation
into regions of interest; (2) extraction of features, numeric on non-numeric,
on these regions; (3) classification of these regions. You are presenting us
with the first step: segmentation.

There are a number of ways in which this can be approached. One of the easiest
is to use a cell automaton algorothm, in which contents of a pixel are modified
according to the contents of pixels immediately surrounding the one you are
working on. (This requires copying you image to a working area, so you don't
mess up the original image.) For example, on a black-and-white image, if the
majority of the pixels surrounding the one you are working on are above a
specified threshold, turn the pixel fully on, else turn it fully off. This will
tend to consolidate regions of similar density. Of course, there are many
variations of this scheme.

Anoter technique is to scan the image vertically and horizontally, getting some
statistics on local average density and density variability. A change in these
statistics might indicate that you are entering a different region. This can be
coupled with the second image created above, so that the pixel contents hold a
numerical reference to what region they are are in.

There are a lot of algorithms for edge detection, but in images like yours I've
been more successful in detecting regions rather than edges. But edge detection
is worth looking into.

And so on. The idea is to produce a second image in which the different regions
have pixels coded as to their region number. This second image can then be used
to mask the first.

There are some articles on use of AI in segmentation, but I forget where they
are. In region classification, fuzzy AI is almost a must.

William Siler

############################################################################
This message was posted through the fuzzy mailing list.
(1) To subscribe to this mailing list, send a message body of
"SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at
(2) To unsubscribe from this mailing list, send a message body of
"UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
to listproc@dbai.tuwien.ac.at
(3) To reach the human who maintains the list, send mail to
fuzzy-owner@dbai.tuwien.ac.at
(4) WWW access and other information on Fuzzy Sets and Logic see
http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
(5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html