/[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 1.145 by jvk, Sat May 3 20:10:48 2003 UTC revision 1.146 by jvk, Sat May 3 21:03:10 2003 UTC
# Line 55  Janne V. Kujala and Tuomas J. Lukka\\ Line 55  Janne V. Kujala and Tuomas J. Lukka\\
55    
56  \begin{abstract}  \begin{abstract}
57    
 %FIXME: more references of texture generation  
   
58  %FIXME: Justify why adding unique background will help recognise a document  %FIXME: Justify why adding unique background will help recognise a document
59  %from a fragment and enhance text readability. (Optional)  %from a fragment and enhance text readability. (Optional)
   
 %FIXME: Add a concise description of how fig. 3 works in the paper.  
   
 %FIXME: Make the titles of figures shorter, and explain the details in the  
 %context.(Optional)  
   
 %FIXME: Revise the spelling and grammar errors, esp. grammar of long sentences.  
   
   
60  %FIXME: more concise and exact explanations in some sections  %FIXME: more concise and exact explanations in some sections
61  %FIXME: what are this paper's specific contributions? (e.g., in Sec. 4)  %FIXME: what are this paper's specific contributions? (e.g., in Sec. 4)
62    
# Line 360  There is also physiological evidence of Line 349  There is also physiological evidence of
349  %of visual perception  %of visual perception
350  %are fairly well known  %are fairly well known
351  in the visual cortex, there are cells sensitive to different  in the visual cortex, there are cells sensitive to different
352  frequencies, orientations, and locations in the visual field%(see, e.g.,~Bruce et al  frequencies, orientations, and locations in the visual field%(see, e.g., Bruce et al
353  \cite{bruce96visualperception}.  \cite{bruce96visualperception}.
354    
355  On a higher level, the correlations between local features are combined  On a higher level, the correlations between local features are combined
# Line 502  dimensionality. Line 491  dimensionality.
491  %    %  
492  %  For instance, consider a published scientific article  %  For instance, consider a published scientific article
493  %  available through several different URLs on the WWW,  %  available through several different URLs on the WWW,
494  %  e.g.,~through ResearchIndex\cite{giles98citeseer%,lawrence99digital  %  e.g., through ResearchIndex\cite{giles98citeseer%,lawrence99digital
495  %  },  %  },
496  %  the ACM portal and the author's web site.  %  the ACM portal and the author's web site.
497  %  Now, when a user downloads and saves one copy, the relationship between  %  Now, when a user downloads and saves one copy, the relationship between
# Line 626  As seen for example in \cite{olson02vstm Line 615  As seen for example in \cite{olson02vstm
615  a limited  a limited
616  number of different features detected can be grouped into  number of different features detected can be grouped into
617  objects, indicating that the spatial resolution of the feature  objects, indicating that the spatial resolution of the feature
618  vector is quite low -- as a well-known example, conjunction  vector is quite low --- as a well-known example, conjunction
619  coding is not preattentive -- red squares are hard to find among  coding is not preattentive --- red squares are hard to find among
620  green squares and red and green circles.  green squares and red and green circles.
621    
622  The feature vector is then used to compute  The feature vector is then used to compute
# Line 651  This configuration is commonly used in n Line 640  This configuration is commonly used in n
640  The qualitative model of visual perception used to create  The qualitative model of visual perception used to create
641  the algorithm.  the algorithm.
642  The visual input is transformed into a feature vector,  The visual input is transformed into a feature vector,
643  which contains numbers (activation levels) corresponding to  which contains numbers (activation levels) corresponding to,
644  e.g.~colors, edges, curves and small patterns.  e.g., colors, edges, curves and small patterns.
645  The feature vector is matched against the memorized textures.  The feature vector is matched against the memorized textures.
646  In order to generate recognizable textures, random seed  In order to generate recognizable textures, random seed
647  values should produce  values should produce
# Line 740  In a sense, the model of perception shou Line 729  In a sense, the model of perception shou
729  in order to produce a unique background from  in order to produce a unique background from
730  a random vector.  a random vector.
731  Features that are orthogonal for human perception  Features that are orthogonal for human perception
732  (e.g.,~color and the direction of the fastest luminance change)  (e.g., color and the direction of the fastest luminance change)
733  should be independently random, and features not orthogonal  should be independently random, and features not orthogonal
734  (e.g., colors of neighbouring pixels)  (e.g., colors of neighbouring pixels)
735  should be correlated so as to maximize the entropy.  should be correlated so as to maximize the entropy.
# Line 1572  appearance designed by a human designer; Line 1561  appearance designed by a human designer;
1561  magazines and web sites have long used skillfully designed  magazines and web sites have long used skillfully designed
1562  graphical elements to make themselves recognizable.  graphical elements to make themselves recognizable.
1563  However, our algorithm is able to generate an unlimited  However, our algorithm is able to generate an unlimited
1564  amount of unique backgrounds cheaply, making it possible  amount of unique backgrounds cheaply, making it possible,
1565  e.g. for all academic articles with similar typography  e.g., for all academic articles with similar typography
1566  to have their own background.  to have their own background.
1567    
1568  %\subsection{Further work}  %\subsection{Further work}
# Line 1610  have not even tapped the full potential Line 1599  have not even tapped the full potential
1599  % to use repeating units for non-repeating backgrounds --- but use more than one,  % to use repeating units for non-repeating backgrounds --- but use more than one,
1600  % which are not rationally related.  % which are not rationally related.
1601    
1602  We report an initial recognition experiment showing that the textures  We have conducted an initial recognition experiment
1603    showing that the textures
1604  generated by our algorithm are indeed recognizable.  generated by our algorithm are indeed recognizable.
1605  Carrying out usability more tests is necessary, both  Carrying out more usability tests is necessary, both
1606  to measure how well textures can be remembered  to measure how well textures can be remembered
1607  and to make the ad hoc distributions more experimentally based.  and to make the ad hoc distributions more experimentally based.
1608    

Legend:
Removed from v.1.145  
changed lines
  Added in v.1.146

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