/[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 1.55 by jvk, Fri Mar 28 12:53:07 2003 UTC revision 1.56 by jvk, Fri Mar 28 13:54:17 2003 UTC
# Line 274  and statistically, as samples from a pro Line 274  and statistically, as samples from a pro
274  %depends only on the values of its neighborhood (local characteristics).  %depends only on the values of its neighborhood (local characteristics).
275  %XXX: resolution-dependency?  %XXX: resolution-dependency?
276    
 %% XXX: this is not really texturing:  
 There have been studies on  
 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,  
 three dimensions have also been  
 sufficient \cite{rao96texturenaming}, 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  
277    
278  % In this article, we apply texture shading to synthesize a large number  % In this article, we apply texture shading to synthesize a large number
279  % of unique textures for distinguishing virtual objects.  % of unique textures for distinguishing virtual objects.
# Line 561  just as well \cite{bergen88earlyvision}. Line 546  just as well \cite{bergen88earlyvision}.
546  In these models, a bank of linear filters is applied to the texture followed  In these models, a bank of linear filters is applied to the texture followed
547  by a nonlinearity and then another set of filters to extract features  by a nonlinearity and then another set of filters to extract features
548  (see, e.g., Heeger\cite{heeger95pyramid}).  (see, e.g., Heeger\cite{heeger95pyramid}).
549  There is also physiological evidence of filtering processes:  There is also physiological evidence of the filtering processes:
550  %The first stages  %The first stages
551  %of visual perception  %of visual perception
552  %are fairly well known  %are fairly well known
# Line 612  about recognizability. Line 597  about recognizability.
597  The simple model we use here assumes  The simple model we use here assumes
598  that at some point,  that at some point,
599  the results from the different pre-attentive feature detectors,  the results from the different pre-attentive feature detectors,
600  such as local and global shapes and colors,  such as different shapes and colors,
601  are combined to form an abstract \emph{feature vector}  are combined to form an abstract \emph{feature vector}
602  (see Fig.~\ref{fig-perceptual}).  (see Fig.~\ref{fig-perceptual}).
603  The feature vector is then used to compute  The feature vector is then used to compute
# Line 631  different instances Line 616  different instances
616  of noise would all yield  of noise would all yield
617  {\em almost  {\em almost
618  exactly the same feature vector} in the brain.  exactly the same feature vector} in the brain.
619  Noise has no global shapes because there is no correlation between  Noise has no shape because there is no correlation between
620  the random local features; it is simply perceived as the distribution  the local features; it is simply perceived as the distribution
621  of the local features, i.e., color and overall frequency  of texel colors and the overall frequency (the density of texels).
 (the density of texels).  
622    
623  From the  From the
624  model  model
# Line 653  To be distinguishable, \label{secfeature Line 637  To be distinguishable, \label{secfeature
637  % - seeing different parts of the texture?  % - seeing different parts of the texture?
638  % - ambiguous perception?  % - ambiguous perception?
639  \item The entropy of the feature vectors  \item The entropy of the feature vectors
640      over the distribution of textures, should be maximized.      over the distribution of textures should be maximized.
641  \end{itemize}  \end{itemize}
642    
643  % which facilitates recognition and memorization of images.  % which facilitates recognition and memorization of images.
# Line 673  To be distinguishable, \label{secfeature Line 657  To be distinguishable, \label{secfeature
657  %We call this the principle of saving bits.  %We call this the principle of saving bits.
658    
659  The last part means essentially  The last part means essentially
660  that if all square-like shapes were green, we would again be  that if all square-like shapes were green, we would be
661  wasting recognitive power.  wasting recognitive power.
662  There should also be as many possible features in the distribution  There should also be as many possible features in the distribution
663      as possible. For example, if there were no yellow textures,      as possible. For example, if there were no yellow textures,
# Line 853  Hues are chosen from a distribution whic Line 837  Hues are chosen from a distribution whic
837  them fairly close together around a uniformly chosen mean  them fairly close together around a uniformly chosen mean
838  but also allows far-away hues  but also allows far-away hues
839  in the same palette occasionally.  in the same palette occasionally.
840  The saturations are chosen from distribution emphasizing  The saturations are chosen from a distribution emphasizing
841  saturated colors: unsaturated colors can easily cause a too multicolored  saturated colors: unsaturated colors can easily cause a too multicolored
842  palette because the adaptive effects of the eye shift them  palette because the adaptive effects of the eye shift them
843  towards the complementary colors of the more saturated colors  towards the complementary colors of the more saturated colors
844  in the palette.  in the palette.
845    
846  XXX: note: maximum saturation of light colors is limited  XXX: note: maximum saturation of light colors is limited
847    
848  % To produce a palette of  % To produce a palette of
# Line 1411  and to make the ad hoc distributions mor Line 1396  and to make the ad hoc distributions mor
1396  It could also be worthwhile to experiment with other ways  It could also be worthwhile to experiment with other ways
1397  of visualizing identity, such as unique edge shapes.  of visualizing identity, such as unique edge shapes.
1398    
1399    There have been studies on
1400    mapping texture appearance to an Euclidian texture space
1401    (see \cite{gurnsey01texturespace} and the references therein):
1402    in the reported experiments, three dimensions have been sufficient
1403    to explain most of the variation in the similarity judgements for
1404    artificial textures.
1405    However, the texture stimuli have been somewhat simple
1406    (no color, lack of frequency-band interaction, etc.).
1407    For some natural texture sets,
1408    three dimensions have also been
1409    sufficient \cite{rao96texturenaming}, but often semantic connections cause the
1410    similarity to be context-dependant, making it hard to assess the
1411    dimensionality.
1412    %% XXX: this is something we should experiment with our textures
1413    
1414  % The new graphics chips, ATI R300 and NVIDIA NV30 support  % The new graphics chips, ATI R300 and NVIDIA NV30 support
1415  % a great deal more of procedural texturing and it will  % a great deal more of procedural texturing and it will
1416  % be interesting to apply the same criteria there.  % be interesting to apply the same criteria there.

Legend:
Removed from v.1.55  
changed lines
  Added in v.1.56

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