mtdetect(1)                                                        mtdetect(1)

       mtdetect - detect circularly symmetric features


       This program is a companion to Mtteach, the interactive program for
       defining a discriminant analysis for feature detection.  On each sec-
       tion, it follows the procedures of that program: it looks at either
       single pixels or sums of pixels within circular windows, seeking points
       with values beyond a threshold value, then applies the discriminant
       function derived in Mtteach to classify the points as features or non-
       features.  However, it also reads in the teaching points used by
       MTTEACH and uses them, as well as the points identified by the proce-
       dure just described, as a basis for searching from one section to the
       next for additional points.  On the first pass through the sections, it
       applies the procedure just described, then it examines the vicinity of
       known points on adjacent sections.  For each such position, it looks
       for points within a specified radius that have values greater than a
       relaxed threshold, and forms discriminant scores for those points; if
       there are points with scores better than the criterion (which may be
       relaxed by a specified amount), then the best such point is taken as a
       new feature point. On a second pass through the sections, it does only
       the checking from section to section.  If the images being examined are
       unaligned but an alignment is available, one can enter a file of g
       transforms to improve the section-to-section correspondence for the

       ENTRIES to the program:

       Image file name
       Name of file of piece coordinates if image is a montage, otherwise
       Name of file of teaching points
       Name of output file for list of detected points
       Name of file of discriminant analysis parameters and solutions
          produced by MTTEACH
       Name of file of g transforms that align the whole data stack, or
          Return if none is available

       IF you enter a file of g transforms, enter 4 lines:
          0 if images are unaligned, or 1 if images are aligned via these
          0 if teaching points were picked on unaligned images, or 1 if
             points were picked on aligned images
          0 to output points that would match unaligned images, or 1 to
             output points that match aligned images
          X and Y center coordinates of the transforms, or / to take the
             default of the center coordinates of the image file

       Minimum distance between centers of adjacent features

       Radius to search around the position of a feature for features in
          adjacent sections

       Amount to relax criterion discriminant score for the search in
          adjacent sections.  This is expressed as the number of standard
          deviations of the scores for true features; the program will first
          print out what the criterion is already in terms of the number of
          S.D.'s below the mean score for true features.

       Amount to relax threshold for choosing peaks in the search for
          features in adjacent sections.  Use the same units in which the
          threshold was specified in MTTEACH, i.e. an absolute or a relative

       Starting and ending section to analyze, or / to do all sections

       Written by David Mastronarde, May 1989
       4/29/90 - add ability to set range of sections to analyze

       Email bug reports to mast at colorado dot edu.

BL3DEMC                              4.7.3                         mtdetect(1)