/[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 1.41 by tjl, Thu Mar 20 14:16:54 2003 UTC revision 1.42 by jvk, Thu Mar 20 15:32:25 2003 UTC
# Line 445  Now, Line 445  Now,
445    
446  %XXX: shorten by one half column  %XXX: shorten by one half column
447    
 XXX: - visual discrimination experiments: \cite{julesz62visualpattern}  
   
 XXX: simple models (filtering) can have good explanatory power  
 on texture discrimination\cite{bergen88earlyvision}.  
   
448  In this section, we discuss the methods to generate unique  In this section, we discuss the methods to generate unique
449  background textures on an abstract level.  background textures on an abstract level.
450    
# Line 1125  The bleaching is unnoticeable and works Line 1120  The bleaching is unnoticeable and works
1120  we start from a fairly light background (>80 XXX).  we start from a fairly light background (>80 XXX).
1121  On a darker background, this approach would probably look silly.  On a darker background, this approach would probably look silly.
1122    
1123  \subsection{Recognizability and memorizability}  \subsection{Texture perception}
   
 % XXX: shorten by half  
   
 JVK  
1124    
1125  Psychophysical studies on texture perception have mostly concentrated  Psychophysical studies on texture perception have mostly concentrated
1126  on \emph{visual texture discrimination}\cite{julesz62visualpattern},  on \emph{visual texture discrimination}\cite{julesz62visualpattern},
# Line 1172  of features (see, e.g., \cite{heeger95py Line 1163  of features (see, e.g., \cite{heeger95py
1163    
1164  In our application visual texture discrimination is not as  In our application visual texture discrimination is not as
1165  much of an issue as memorizability and recognizability of  much of an issue as memorizability and recognizability of
1166  previously seen textures.  previously seen textures, because the textures are ... discriminable..
1167    
1168    
1169    There have been studies on
1170    mapping texture appearance to an Euclidian texture space
1171    (see \cite{gurnsey01texturespace} and the references therein):
1172    in the reported experiments, three dimensions have been sufficient
1173    to explain most of the variation in the similarity judgements for
1174    artificial textures.
1175    However, the texture stimuli have been somewhat simple
1176    (no color, lack of frequency-band interaction, etc.).
1177    For some natural texture sets (see, e.g., \cite{rao96texturenaming}),
1178    three dimensions have also been
1179    sufficient, but often semantic connections cause the
1180    similarity to be context-dependant, making it hard to assess the
1181    dimensionality.
1182    % XXX: this is something we should experiment with our textures
1183    
1184    \subsection{Recognizability and memorizability}
1185    
1186  The textures created by our algorithm,  The textures created by our algorithm,
1187  although repeating, are more like complete images  although repeating, are more like complete images
1188  than the statistical microstructure studied in most texture perception work.  than the statistical microstructure studied in most texture perception work.
# Line 1213  Many studies of web cache statistics Line 1223  Many studies of web cache statistics
1223  have shown that file popularity approximately follows Zipf's law  have shown that file popularity approximately follows Zipf's law
1224  (see, e.g. [xxxref]) so that a small number of documents accounts  (see, e.g. [xxxref]) so that a small number of documents accounts
1225  for most of the use.  for most of the use.
1226  Furthermore, because the texture appearance has no semantic connection  
1227  with the document,  Furthermore, because the texture appearance has no correlation
1228    with the document content,
1229  the textures of any two important files are similar only by chance.  the textures of any two important files are similar only by chance.
1230    
1231  %Our experience shows that at least the most recurring textures  %Our experience shows that at least the most recurring textures
1232  %are not only recognized but can  %are not only recognized but can
1233  %also easily be associated with the document content.  %also easily be associated with the document content.
1234    
 XXX  
   
 Mapping texture appearance to an Euclidian texture space  
 (see \cite{gurnsey01texturespace} and the references therein):  
 in the reported experiments, three dimensions have been sufficient  
 to explain most of the variation in the similarity judgements for  
 artificial textures.  
 However, the texture stimuli have been somewhat simple  
 (no color, lack of frequency-band interaction, etc.).  
 For some natural texture sets (see, e.g., \cite{rao96texturenaming}),  
 three dimensions have also been  
 sufficient, but often semantic connections cause the  
 similarity to be context-dependant, making it hard to assess the  
 dimensionality.  
 % XXX: this is something we should experiment with our textures  
   
1235  XXX: refs?  XXX: refs?
1236    
1237  %\section{Software availability}  %\section{Software availability}

Legend:
Removed from v.1.41  
changed lines
  Added in v.1.42

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