home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.graphics
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!news.acns.nwu.edu!network.ucsd.edu!mvb.saic.com!unogate!stgprao
- From: stgprao@xing.unocal.com (Richard Ottolini)
- Subject: Re: Reversal of Hough Transform?
- Message-ID: <1992Jul23.151551.10685@unocal.com>
- Originator: stgprao@xing
- Sender: news@unocal.com (Unocal USENET News)
- Organization: Unocal Corporation
- References: <BrtA8E.Eq8@acsu.buffalo.edu>
- Date: Thu, 23 Jul 1992 15:15:51 GMT
- Lines: 19
-
- In article <BrtA8E.Eq8@acsu.buffalo.edu> lusardi@sybil.cs.Buffalo.EDU (Christopher Lusardi) writes:
- >A straight line can be parametrically described as:
- > rho = x * cos(phi) + y * sin (phi)
- >where rho is the normal distance of the line from the origin and phi is
- >the angle of the origin with respect to the axis.
- >
- >The Hough transform maps (rho, phi) coordinates to an array for determining
- >lines. I.e: If an array element has a specific number of hits in the
- >parametric domain then we have a line and not noise etc.
-
- The Hough transform is relative of a class of transforms called the generalized
- Radon transforms. These map information from one coordinate domain to another
- by integration of some of the data. There is much material on these transforms
- and their inversion in the medical tomography and geophysics literature.
-
- I don't know if there is an exact inverse (reversal), but there are near
- perfect inverses for practical purposes. The reason there doesn't seem to be
- an inverse is because even parameterization in one coordinate domain usually
- undersamples or oversamples some even part of the transform coordinate domain.
-