/[gzz]/manuscripts/Paper/paper.tex
ViewVC logotype

Diff of /manuscripts/Paper/paper.tex

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1.140 by jvk, Sat May 3 18:31:03 2003 UTC revision 1.141 by tjl, Sat May 3 19:08:30 2003 UTC
# Line 588  unique backgrounds. Line 588  unique backgrounds.
588    
589  %XXX: shorten by one half column  %XXX: shorten by one half column
590    
591  In this section, we discuss the principles underlying  In this section, we discuss the general principles
592    we have used to derive
593  our algorithm to generate unique  our algorithm to generate unique
594  background textures.  background textures.
595  To be useful, the unique backgrounds should be easily  The goals are that the unique backgrounds should be easily
596  distinguishable and recognizable, and should not  distinguishable and recognizable, and that they should not
597  significantly impair the reading of black text on top of it.  significantly impair the reading of black text drawn on top.
598    
599    Our approach is an extension of
600    the type of inversion approach that been used
601    by Ware and Knight\cite{ware95texture} for inverting
602    the earliest stage of the visual system%,
603    %the spatial frequency detectors,
604    in order to place
605    a particular vector or scalar field of data
606    in the texture ``channel''.
607    
608  The ability to distinguish a particular texture from a large set  The ability to distinguish a particular texture from a large set
609  depends on the distribution of textures in the set.  depends on the distribution of textures in the set.
# Line 736  should be correlated so as to maximize t Line 746  should be correlated so as to maximize t
746  %For example, pixels on a small area should correlate enough to  %For example, pixels on a small area should correlate enough to
747  %facilitate perception of contours.  %facilitate perception of contours.
748    
 A type of inversion approach has been used  
 by Ware and Knight\cite{ware95texture}, for inverting  
 the earliest stage of the visual system,  
 the spatial frequency detectors, in order to place  
 a particular vector or scalar field of data  
 in the texture ``channel''.  
   
749  % To understand why it is possible to learn to discriminate particular  % To understand why it is possible to learn to discriminate particular
750  % backgrounds easily, consider the task of learning {\em one} background texture.  % backgrounds easily, consider the task of learning {\em one} background texture.
751  % This is a two-class problem.  % This is a two-class problem.

Legend:
Removed from v.1.140  
changed lines
  Added in v.1.141

savannah-hackers-public@gnu.org
ViewVC Help
Powered by ViewVC 1.1.26