/[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 1.53 by tjl, Mon Mar 24 16:24:27 2003 UTC revision 1.54 by jvk, Mon Mar 24 17:24:17 2003 UTC
# Line 33  Janne V. Kujala and Tuomas J. Lukka\\ Line 33  Janne V. Kujala and Tuomas J. Lukka\\
33  \begin{document}  \begin{document}
34  \maketitle  \maketitle
35    
36  \font\foofonti = cmfrak scaled 2000  %\font\foofonti = cmfrak scaled 2000
37  \font\foofontii = qhvb scaled 2000  %\font\foofontii = qhvb scaled 2000
38    
39    
40  \begin{abstract}  \begin{abstract}
# Line 264  other textures as a starting point Line 264  other textures as a starting point
264  perceptually, for visualizing surface orientation\cite{schweitzer83texturing,interrante97illustrating} and scalar or vector fields\cite{ware95texture},  perceptually, for visualizing surface orientation\cite{schweitzer83texturing,interrante97illustrating} and scalar or vector fields\cite{ware95texture},
265  and statistically, as samples from a probability distribution on a random field  and statistically, as samples from a probability distribution on a random field
266  \cite{cross83markov,geman84stochastic}.  \cite{cross83markov,geman84stochastic}.
267    % XXX: there's overlap between the enumerated cases
268    
269  %Textures have also been modeled statistically,  %Textures have also been modeled statistically,
270  %as samples from a probability distribution on a random field.  %as samples from a probability distribution on a random field.
# Line 273  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    
277  % In this article, we apply texture shading to synthesize a large number  %% XXX: this is not really texturing:
 % of unique textures for distinguishing virtual objects.  
   
 \subsection{Texture perception}  
   
 Psychophysical studies on texture perception have mostly concentrated  
 on pre-attentive  
 \emph{visual texture discrimination}\cite{julesz62visualpattern},  
 the ability of human observers to effortlessly discriminate  
 pairs of certain textures (see Bergen\cite{bergen91theories} for a review).  
 %The term is often used interchangably with \emph{texture segregation},  
 %the more specific task of finding the border between differently textured  
 %areas (different phases of local characteristics at the  
 %border can segregate otherwise indiscriminable textures).  
 %  
 %First experiments on computer-generated, unnatural textures in the 60s  
 %\cite{julesz62visualpattern} led to proposals of discrimination models  
 %based on the $N$th-order statistics of textures  
 %(the joint distributions of the values at the corners of a randomly  
 %placed (translated) $N$-gon for all different $N$-gons).  
 %%and connectivity structures of certain micropatterns.  
 %  
 First discrimination models were based  
 on the $N$th-order statistics of textures  
 (the joint distributions of the values at the corners of a randomly  
 placed (translated) $N$-gon for all different $N$-gons).  
 However, the order of similarity in the statistics did not  
 consistently explain discrimination performance, and certain  
 pre-attentive local features were conjectured.  
   
 Julesz\cite{julesz81textons} proposed that texture discrimination could be  
 explained by the densities of textons, fundamental texture elements, such as  
 elongated blobs, line terminators, and line crossings.  
 However, the textons are hard to define formally.  
   
 Much simpler filtering-based models can explain texture discrimination  
 just as well \cite{bergen88earlyvision}.  
 In these models, a bank of linear filters is applied to the texture followed  
 by a nonlinearity and then another set of filters to extract features  
 (see, e.g., \cite{heeger95pyramid} for an application).  
 %In \cite{heeger95pyramid}, new textures with appearance similar  
 %to a given texture are created by matching certain histograms  
 %of filter responses.  
   
 XXX: higher-level pre-attentive processes?  
   
 %XXX: texture perception reviews  
   
278  There have been studies on  There have been studies on
279  mapping texture appearance to an Euclidian texture space  mapping texture appearance to an Euclidian texture space
280  (see \cite{gurnsey01texturespace} and the references therein):  (see \cite{gurnsey01texturespace} and the references therein):
# Line 334  three dimensions have also been Line 288  three dimensions have also been
288  sufficient \cite{rao96texturenaming}, but often semantic connections cause the  sufficient \cite{rao96texturenaming}, but often semantic connections cause the
289  similarity to be context-dependant, making it hard to assess the  similarity to be context-dependant, making it hard to assess the
290  dimensionality.  dimensionality.
291  % XXX: this is something we should experiment with our textures  %% XXX: this is something we should experiment with our textures
292    
293    % In this article, we apply texture shading to synthesize a large number
294    % of unique textures for distinguishing virtual objects.
295    
296    %\subsection{Texture perception}
297    %
298    %Psychophysical studies on texture perception have mostly concentrated
299    %on pre-attentive
300    %\emph{visual texture discrimination}\cite{julesz62visualpattern},
301    %the ability of human observers to effortlessly discriminate
302    %pairs of certain textures (see Bergen\cite{bergen91theories} for a review).
303    %%The term is often used interchangably with \emph{texture segregation},
304    %%the more specific task of finding the border between differently textured
305    %%areas (different phases of local characteristics at the
306    %%border can segregate otherwise indiscriminable textures).
307    %%
308    %%First experiments on computer-generated, unnatural textures in the 60s
309    %%\cite{julesz62visualpattern} led to proposals of discrimination models
310    %%based on the $N$th-order statistics of textures
311    %%(the joint distributions of the values at the corners of a randomly
312    %%placed (translated) $N$-gon for all different $N$-gons).
313    %%%and connectivity structures of certain micropatterns.
314    %%
315    %First discrimination models were based
316    %on the $N$th-order statistics of textures
317    %(the joint distributions of the values at the corners of a randomly
318    %placed (translated) $N$-gon for all different $N$-gons).
319    %However, the order of similarity in the statistics did not
320    %consistently explain discrimination performance, and certain
321    %pre-attentive local features were conjectured.
322    %
323    %Julesz\cite{julesz81textons} proposed that texture discrimination could be
324    %explained by the densities of textons, fundamental texture elements, such as
325    %elongated blobs, line terminators, and line crossings.
326    %However, the textons are hard to define formally.
327    %
328    %Much simpler filtering-based models can explain texture discrimination
329    %just as well \cite{bergen88earlyvision}.
330    %In these models, a bank of linear filters is applied to the texture followed
331    %by a nonlinearity and then another set of filters to extract features
332    %(see, e.g., \cite{heeger95pyramid} for an application).
333    %%In \cite{heeger95pyramid}, new textures with appearance similar
334    %%to a given texture are created by matching certain histograms
335    %%of filter responses.
336    %
337    %XXX: higher-level pre-attentive processes?
338    %
339    %%XXX: texture perception reviews
340    %
341    %There have been studies on
342    %mapping texture appearance to an Euclidian texture space
343    %(see \cite{gurnsey01texturespace} and the references therein):
344    %in the reported experiments, three dimensions have been sufficient
345    %to explain most of the variation in the similarity judgements for
346    %artificial textures.
347    %However, the texture stimuli have been somewhat simple
348    %(no color, lack of frequency-band interaction, etc.).
349    %For some natural texture sets,
350    %three dimensions have also been
351    %sufficient \cite{rao96texturenaming}, but often semantic connections cause the
352    %similarity to be context-dependant, making it hard to assess the
353    %dimensionality.
354    %% XXX: this is something we should experiment with our textures
355    
356  \subsection{Focus+Context views}  \subsection{Focus+Context views}
357    
# Line 525  In order to design a distinguishable dis Line 542  In order to design a distinguishable dis
542  we have to take into account the properties of the human  we have to take into account the properties of the human
543  visual system.  visual system.
544    
545  % The seed for randomly choosing  Psychophysical studies on texture perception have mostly concentrated
546  % an easily distinguishable unique background from a  on pre-attentive
547  % distribution based on  \emph{visual texture discrimination}\cite{julesz62visualpattern},
548    the ability of human observers to effortlessly discriminate
549    pairs of certain textures (see Bergen\cite{bergen91theories} for a review).
550    Nevertheless,
551    discrimination models can provide insight on the pre-attentive
552    processes underlying global perception.
553    
554  %providing an infinite source of unique backgrounds.  Julesz\cite{julesz81textons} proposed that texture discrimination could be
555  %generating textures based on seed numbers [identity]  explained by the densities of textons, fundamental texture elements, such as
556  The first stages  elongated blobs, line terminators, and line crossings.
557  of visual perception  However, the textons are hard to define formally.
558  are fairly well known  
559  (see, e.g.,~Bruce et al\cite{bruce96visualperception}):  Much simpler filtering-based models can explain texture discrimination
560    just as well \cite{bergen88earlyvision}.
561    In these models, a bank of linear filters is applied to the texture followed
562    by a nonlinearity and then another set of filters to extract features
563    (see, e.g., Heeger\cite{heeger95pyramid}).
564    There is also physiological evidence of filtering processes:
565    %The first stages
566    %of visual perception
567    %are fairly well known
568  in the visual cortex, there are cells sensitive to different  in the visual cortex, there are cells sensitive to different
569  frequencies, orientations, and locations in the visual field.  frequencies, orientations, and locations in the visual field
570    (see, e.g.,~Bruce et al\cite{bruce96visualperception}).
571    
572  On a higher level, the correlations between local features are combined  On a higher level, the correlations between local features are combined
573  by forming contours and possibly  by forming contours and possibly
574  other higher-level constructions.  other higher-level constructions (see, e.g., \cite{saarinen97integration}).
575  These higher levels are not yet thoroughly understood;  These higher levels are not yet thoroughly understood;
576  some theories  some theories
577  (see, e.g., Biederman\cite{biederman87})  (see, e.g., Biederman\cite{biederman87})
578  assume certain primitive shapes whose  assume certain primitive shapes whose
579  structure facilitates recognition.  structure facilitates recognition.
580    
581    % The seed for randomly choosing
582    % an easily distinguishable unique background from a
583    % distribution based on
584    
585    %providing an infinite source of unique backgrounds.
586    %generating textures based on seed numbers [identity]
587    
588  \begin{figure}  \begin{figure}
589  \centering  \centering
590  %\fbox{\vbox{\vskip 3in}}  %\fbox{\vbox{\vskip 3in}}

Legend:
Removed from v.1.53  
changed lines
  Added in v.1.54

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