Re: fuzzy change

From: sam (sdomenic@attbi.com)
Date: Fri Jan 04 2002 - 17:24:35 MET

  • Next message: tim menzies: "extended deadline: WISET1: workshop innovative software engineering"

    Ola,

    I have had some experience with image processing. I think that everyone's
    first approach is to try to find differences by subtracting sequential
    images. This is usually not very productive. Even when using a fixed
    camera to image a fixed target, there will be differences you can attribute
    to thermal noise in the camera or amplifiers. Other sources of noise can
    introduce black and white spots (salt-and-pepper noise) caused by saturation
    or cut-off in the amplifiers, or periodic nonlinearities causing regular
    patterns in one or both images.

    When you are comparing frames from satellite images, I would think that
    there would be problems in at least the following ways:
    (1) finding corresponding fiducial (index) points in the two images
    (2) geometric distortion caused by differences in slant range (camera angle
    off the vertical) for the two images
    (3) time-of-day lighting changes between the two frames
    (4) cloud-cover differences between the two images

    I do not claim that subtraction cannot be used, merely that it is difficult.

    For Instance, if I we to try to compare two satellite images by subtraction,
    I might try to process the image in the following way.

    (1) convert both images from color to gray scale [decreases the work by a
    factor of three]
    (2) normalize the histograms of the two images by stretching one or both
    histograms.. (Make the blacks and whites the same in the two images, and
    hoping the distribution of grays will be similar) [takes care of factor 3,
    above]
    (3) Find corresponding features (near the four frame corners, hopefully) and
    use geometric transformations to make the two frames have the same geometric
    shape.[This takes care of factor 2, above]
    (4) Try to find contiguous (in some form) areas of high reflectivity in each
    image. This will be cloud cover. Mark these areas as "not available for
    examination"

    At this point, we still have differences attributable to noise, but do the
    subtraction anyway.

    Subtract one image from the other, ignoring the "not available of
    examination" areas. Areas that are identical and are lucky enough to have
    no noise will have a very small (sometimes zero) difference. Noise areas
    should be uncorrelated (random) and mostly small. A fuzzy comparison could
    probably be used directly.

    A more productive approach would probably be to extract features from each
    image (region growing to find contiguous areas, using a fuzzy criteria to
    identifiy these areas), and to compare the two images on the basis of
    differences in the size of the areas). I found this approach to be very
    usable.

    In my most successful case, I took the fourier transform of each of the two
    images and compared the coefficients of the ten most significant
    coefficients.

    I think you would find several good possiblities for features to examine in
    any textbook on image processing, looking at those chapters on feature
    extraction.

    Let me look through my library and see if I can't find some good advice. I
    will post if I can find it (unfortunately, these books were packed away when
    I took a database job). My library is probably 5 years or so out of date,
    so you may have more recent references available.

    I wish you luck in your project and hope you publish in some public venue
    which we can all examine.

    I hope I don't sound too overbearing. I have found some of this information
    useful in the past, and hope you do also.

    Take care,

    Sam

    --
    Sam Domenico
    (303) 278-4142
    sdomenic@attbi.com
    

    "Ola Hall" <ola.hall@humangeo.su.se> wrote in message news:006d01c17e6b$8dafe580$deadcc84@geog.umontreal.ca... > This is a multi-part message in MIME format. > > ------=_NextPart_000_006A_01C17E41.A46C0080 > Content-Type: text/plain; > charset="iso-8859-1" > Content-Transfer-Encoding: quoted-printable > > Hej > I'm about to start a new project about landscape change detection = > methods. The landscape is to be understood as "satellite data/imagery". = > The format is usually 8-16 bit byte or float. Change is typically viewed = > as a subtraction of images. However, it is very difficult to separate = > instrumental noise, atmospheric inference, etc from true landscape = > changes. Every chunk of change will then be a mixture of all these = > different types of changes. > I'm convinced that fuzzy theory will be successful in this project. Now, = > I ask you if you have any knowledge, ideas, papers, contacts etc that = > you'd like to share with me? > > sincerely, > Ola > > > > Ola Hall > Stockholm University > Dept. of Human Geography > and > University of Montreal > Geocomputing Laboratory=20 > Dept. of Geography > > > ------=_NextPart_000_006A_01C17E41.A46C0080 > Content-Type: text/html; > charset="iso-8859-1" > Content-Transfer-Encoding: quoted-printable > > <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"> > <HTML><HEAD> > <META http-equiv=3DContent-Type content=3D"text/html; = > charset=3Diso-8859-1"> > <META content=3D"MSHTML 5.50.4616.200" name=3DGENERATOR> > <STYLE></STYLE> > </HEAD> > <BODY bgColor=3D#ffffff> > <DIV><FONT face=3DArial size=3D2>Hej</FONT></DIV> > <DIV><FONT face=3DArial size=3D2>I'm about to start a new project about = > landscape=20 > change detection methods. The landscape is to be understood as = > "satellite=20 > data/imagery". The format is usually 8-16 bit byte or float. Change is = > typically=20 > viewed as a subtraction of images. However, it is very difficult to = > separate=20 > instrumental noise, atmospheric inference, etc from true landscape = > changes.=20 > Every chunk of change will then be a mixture of all these different = > types of=20 > changes.</FONT></DIV> > <DIV><FONT face=3DArial size=3D2>I'm convinced that fuzzy theory will be = > successful=20 > in this project. Now, I ask you if you have any knowledge, ideas, = > papers,=20 > contacts etc that you'd like to share with me?</FONT></DIV> > <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV> > <DIV><FONT face=3DArial size=3D2>sincerely,</FONT></DIV> > <DIV><FONT face=3DArial size=3D2>Ola</FONT></DIV> > <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV> > <DIV>&nbsp;</DIV> > <DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV> > <DIV><FONT face=3DArial size=3D2>Ola Hall<BR>Stockholm = > University<BR>Dept. of Human=20 > Geography<BR>and<BR>University of Montreal<BR>Geocomputing Laboratory = > <BR>Dept.=20 > of&nbsp; Geography<BR></FONT></DIV></BODY></HTML> > > ------=_NextPart_000_006A_01C17E41.A46C0080-- > > > ############################################################################ > 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 >

    ############################################################################ 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



    This archive was generated by hypermail 2b30 : Fri Jan 04 2002 - 17:27:55 MET