/[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 1.61 by tjl, Sat Mar 29 11:19:17 2003 UTC revision 1.62 by jvk, Sat Mar 29 11:44:21 2003 UTC
# Line 1214  the black text, showing the blurred ligh Line 1214  the black text, showing the blurred ligh
1214  }  }
1215  \end{figure}  \end{figure}
1216    
1217  \subsection{Recognizability and memorizability}  \subsection{Recognizability experiment}
1218    
1219  %In our application visual texture discrimination is not as  %In our application visual texture discrimination is not as
1220  %much of an issue as memorizability and recognizability of  %much of an issue as memorizability and recognizability of
# Line 1227  the black text, showing the blurred ligh Line 1227  the black text, showing the blurred ligh
1227    
1228  Experiments on black-and-white %(faces,)  Experiments on black-and-white %(faces,)
1229  ink blots and snow crystals  ink blots and snow crystals
1230  \cite{goldstein71visualrecognition} show that  \cite{goldstein71visualrecognition} have shown that
1231  complex, unfamiliar pictures can be remembered and recognized  complex, unfamiliar pictures can be remembered and recognized
1232  and that recognition  and that recognition
1233  performance decreases very little over time.  performance decreases very little over time.
1234    
1235  In order to evaluate the recognizability of our procedurally generated  In order to evaluate the recognizability of our procedurally generated
1236  textures,  textures, we need to have an appropriate comparison point.
1237    %
1238    %- need to compare to something
1239    %- can't generate an infinite supply
1240    %  of images easily and there are too much semantics
1241    Pictures of natural objects would not provide a good comparsion,
1242    because they cannot be generated in infinite amounts from seed
1243    numbers and they would esasily yield undesirable semantic associations.
1244    Thus, we simply use plain solid color backgrounds as a baseline for comparison
1245    even though the colors of even a small set of randomly chosen colors would
1246    most likely not be discriminable.
1247    
1248  - need to compare to something  %even though know solid colors not really recognizable.
1249    
1250  - images not good, can't generate infinite supply easily, too much  Surely it would be hard to remember a large number of textures, too.
   semantics  
   
 - thus, use solid colors even though know solid colors not really recognizable.  
   
   
 Surely it would be hard to remember, e.g., 100 textures.  
1251  However, the user does not have to remember all the textures;  However, the user does not have to remember all the textures;
1252  it suffices to learn the textures of the most often used documents  it suffices to learn the textures of the most often used documents
1253  and distinguish them from the rest:  and distinguish them from the rest:
# Line 1267  some documents get accessed far more oft Line 1271  some documents get accessed far more oft
1271  Each square represents a document, and the area of each square is scaled  Each square represents a document, and the area of each square is scaled
1272  to its rate of accesses.  to its rate of accesses.
1273  The diagram shows 2000 documents weighted with Zipf's law with exponent 1.1.  The diagram shows 2000 documents weighted with Zipf's law with exponent 1.1.
1274  The 15 most important documents account XXX\% of the accesses.  The 15 most important documents account for 50\% of the accesses.
1275    % 0.50469672124463749
1276  }  }
1277  \end{figure}  \end{figure}
1278    
1279  Thus, the experiment measured the recognition of only 15 textures,  Thus, we chose to measure the recognition of only 15 target textures
1280  and  or colors in the experiment.
1281    Our hypthesis is that the texture backgrounds are more recognizable
1282    than the solid colors.
1283    
1284  Furthermore, because the texture appearance has no correlation  %and
1285    %
1286    
1287    XXX: Furthermore, because the texture appearance has no correlation
1288  with the document content,  with the document content,
1289  the textures of any two important files are similar only by chance.  the textures of any two important files are similar only by chance.
1290    
1291  We have conducted an experiment in a similar setting comparing  %We have conducted an experiment in a similar setting comparing
1292  the recognition performance of our textures with solid color backgrounds.  %the recognition performance of our textures with solid color backgrounds.
   
1293    
1294    \subsubsection{Method}
1295    
1296  Five participants performed the experiment in both conditions  \emph{Participants.}
1297  (in random order).  Five participants naïve to the hypothesis and our texturing work
1298    performed the experiment in both conditions.
1299    
1300    \emph{Design and materials.}
1301  15 target backgrounds and 15 distractor backgrounds were randomly  15 target backgrounds and 15 distractor backgrounds were randomly
1302  chosen for both conditions.  chosen for both conditions.
1303  The distribution of the solid colors was the same that is used for  The distribution of the solid colors was the same that is used for
# Line 1293  the colors of textures with the differen Line 1305  the colors of textures with the differen
1305  lighness tail was de-emphasized to increase the otherwise low  lighness tail was de-emphasized to increase the otherwise low
1306  discriminability of very light, unsaturated colors.  discriminability of very light, unsaturated colors.
1307    
1308    \emph{Procedure.}
1309    Each participant performed the test individually in both conditions
1310    in a random order.
1311  First, the 15 target backgrounds were shown sequentially,  First, the 15 target backgrounds were shown sequentially,
1312  in a random order, 5 seconds each.  in a random order, 5 seconds each.
1313  Then, recognition was tested by showing the 15 target backgrounds  Then, recognition was tested by showing the 15 target backgrounds
1314  and 15 distractor backgrounds in a random order and  and the 15 distractor backgrounds in a random order and
1315  having the participant answer ``old'' or ``new'' for each one.  having the participant answer ``old'' or ``new'' for each one.
1316  The time for answers was not limited.  The time for answers was not limited.
1317    
1318    \subsubsection{Results}
1319    
1320  The results are summarized in Table~\ref{tableresults}  The results are summarized in Table~\ref{tableresults}
1321    
1322  \begin{table}  \begin{table}
# Line 1330  the five subjects. Line 1347  the five subjects.
1347  %use mutual information or something like that. In colors, the subject  %use mutual information or something like that. In colors, the subject
1348  %was basically guessing!!!  %was basically guessing!!!
1349    
 % RT less for correct responses: [F(1,4) = 5.2, p = .082]  
1350  Thus, the recognition performance of the textures was good while  Thus, the recognition performance of the textures was good while
1351  the solid colors do not have enough variation for unambiguous  the solid colors do not have enough variation for unambiguous
1352  recognition (without additional cues).  recognition.
1353  The observed difference in recognition performance is statistically  A repeated measures analysis of variance indicates that
1354    the observed difference in the recognition performance is statistically
1355  significant [F(1,4) = 19.0, p = .012].  significant [F(1,4) = 19.0, p = .012].
1356    % RT less for correct responses: [F(1,4) = 5.2, p = .082]
1357    
1358  %Our experience shows that at least the most recurring textures  %Our experience shows that at least the most recurring textures
1359  %are not only recognized but can  %are not only recognized but can
1360  %also easily be associated with the document content.  %also easily be associated with the document content.
1361    
1362    The solid colors could be more recognizable with additional cues,
1363    such as the text of the document, but
1364    
1365    - fragments and not title page etc...
1366    
1367  XXX: refs?  XXX: refs?
1368    
1369  %\section{Software availability}  %\section{Software availability}

Legend:
Removed from v.1.61  
changed lines
  Added in v.1.62

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