{\rtf1\mac\ansicpg10000\uc1 \deff4\deflang1033\deflangfe1033{\upr{\fonttbl{\f0\fnil\fcharset256\fprq2{\*\panose 02020603050405020304}Times New Roman{\*\falt Times};}{\f4\fnil\fcharset256\fprq2{\*\panose 02000500000000000000}Times;}
}{\*\ud{\fonttbl{\f0\fnil\fcharset256\fprq2{\*\panose 02020603050405020304}Times New Roman{\*\falt Times};}{\f4\fnil\fcharset256\fprq2{\*\panose 02000500000000000000}Times;}}}}{\colortbl;\red0\green0\blue0;\red0\green0\blue255;\red0\green255\blue255;
\red0\green255\blue0;\red255\green0\blue255;\red255\green0\blue0;\red255\green255\blue0;\red255\green255\blue255;\red0\green0\blue128;\red0\green128\blue128;\red0\green128\blue0;\red128\green0\blue128;\red128\green0\blue0;\red128\green128\blue0;
\red128\green128\blue128;\red192\green192\blue192;}{\stylesheet{\widctlpar\adjustright \f4\cgrid \snext0 Normal;}{\s1\keepn\widctlpar\outlinelevel0\adjustright \b\f4\cgrid \sbasedon0 \snext0 heading 1;}{\*\cs10 \additive Default Paragraph Font;}{
\s15\li720\widctlpar\adjustright \f4\cgrid \sbasedon0 \snext15 Body Text 2;}}{\info{\title Notes on Texture Segmentation:}{\author Marc Talusan}{\operator Marc Talusan}{\creatim\yr2003\mo10\dy2\hr17\min19}{\revtim\yr2003\mo10\dy2\hr17\min41}{\version4}
{\edmins22}{\nofpages2}{\nofwords594}{\nofchars3389}{\*\company MIT}{\nofcharsws4161}{\vern115}}\widowctrl\ftnbj\aenddoc\hyphcaps0\formshade\viewkind1\viewscale125\pgbrdrhead\pgbrdrfoot \fet0\sectd \linex0\endnhere\sectdefaultcl {\*\pnseclvl1
\pnucrm\pnstart1\pnindent720\pnhang{\pntxta .}}{\*\pnseclvl2\pnucltr\pnstart1\pnindent720\pnhang{\pntxta .}}{\*\pnseclvl3\pndec\pnstart1\pnindent720\pnhang{\pntxta .}}{\*\pnseclvl4\pnlcltr\pnstart1\pnindent720\pnhang{\pntxta )}}{\*\pnseclvl5
\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}{\*\pnseclvl6\pnlcltr\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}{\*\pnseclvl7\pnlcrm\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}{\*\pnseclvl8\pnlcltr\pnstart1\pnindent720\pnhang
{\pntxtb (}{\pntxta )}}{\*\pnseclvl9\pnlcrm\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\pard\plain \widctlpar\tx560\tx1120\tx1680\tx2240\tx2800\tx3360\tx3920\tx4480\tx5040\tx5600\tx6160\tx6720\adjustright \f4\cgrid {\tab \tab \tab \tab \tab \tab 
\tab \tab \tab \tab \tab Notes by Philip Meier
\par \tab \tab \tab \tab \tab \tab \tab \tab \tab \tab \tab pmmeier@mit.edu
\par \tab \tab \tab \tab \tab \tab \tab \tab \tab \tab \tab 9-30-03
\par 
\par }\pard \widctlpar\adjustright {\tab \tab \tab }{\scaps\fs28 Function of Early Vision}{
\par }\pard \fi720\li2880\widctlpar\adjustright {&
\par }\pard \fi720\li1440\widctlpar\adjustright {\scaps\fs28 Texture Segmentation I
\par }\pard \widctlpar\adjustright {
\par }{\b \ldblquote Bird wing- Brain Area\rdblquote  metaphor}{, Barlow argues that we should know the function of a brain area as we study it.  Studying a 
neural area without knowing its function is like studying a birds wing without knowing that the function of a wing is to enable flight.  What do neurons do? What is their purpose?  More specifically, what is the purpose of a \ldblquote sensory relay
\rdblquote  (like LGN for human vision)  
\par 
\par Barlow proposed three possible functions.
\par 
\par }{\b The function of a sensory relay is to detect features.
\par }{\tab Password hypothesis- signal is like a key or password that can open a locked door
\par \tab A cell fires if the stimulus has properties (x1,x2,\u8230\'c9xn)
\par }\pard\plain \s1\keepn\widctlpar\outlinelevel0\adjustright \b\f4\cgrid {The function of a sensory relay is to filter the signal
\par }\pard\plain \widctlpar\adjustright \f4\cgrid {\tab Controlled pass characteristics
\par }{\b The function of a sensory relay is to reduce redundancy.
\par }\pard \li720\widctlpar\adjustright {To convey the most amount of information with the least amount of metabolic energy
\par }\pard \widctlpar\adjustright {
\par Adelson added:
\par 
\par }{\b The function of a sensory relay is to simplify the problem for the next stage.}{\line (similar to last weeks contrast between lossless transmission and  useful transformation)
\par The question is \ldblquote useful for what?\rdblquote   The \ldblquote use\rdblquote  of the information depends on the task being performed and the resources available to the \ldblquote user.\rdblquote  (the next brain area)
\par 
\par }\pard\plain \s1\keepn\widctlpar\outlinelevel0\adjustright \b\f4\cgrid {Metabolic Energy
\par }\pard\plain \widctlpar\adjustright \f4\cgrid {
\par A brief side discussion was held on the relevance of metabolic energy for representing things in brains.  The primary conclusion was that there may be many constra
ints that operate to influence the representation format in a brain (metabolics, processing time, number of connections, etc.) , and we shouldn\rquote 
t get distracted by metabolics and should focus on representation itself.  However if someone did wan to investigate the effects of metabolics, these were some questions raised:
\par 
\par Is there a selective pressure for metabolic cost?
\par Is an increase in spiking a good measure of increased metabolic energy consumption?
\par How much energy does it take to maintain a concentration gradient?
\par Does is cost more energy to keep 200 cells alive with sparse activity, or 50 cells alive with continuous activity?
\par If you consider behaviors like walking, how does minimum energy movement compare to minimum energy control?
\par 
\par  }{\b Concerns about information theory and neurons.}{  (Raised by M. Histead)
\par 
\par How does one define \ldblquote independence\rdblquote  and \ldblquote redundancy?\rdblquote 
\par This can\rquote t be done because
\par }\pard\plain \s15\li720\widctlpar\adjustright \f4\cgrid {1) We don\rquote t know the temporal precision of spikes.  Infinitely precise times would result in infinite amounts of information in a single spike.
\par }\pard\plain \li720\widctlpar\adjustright \f4\cgrid {2) We don\rquote t know the neural code.  Multiple nerve fibers could contain information in the relation of their activity.
\par }\pard \widctlpar\adjustright {
\par Something that looks like a \ldblquote detector\rdblquote  can arise because of some unrelated constraint.
\par 
\par 
\par }{\b Adelson\rquote s Island Metaphor.}{  Trying to understand intermediate computational steps is like making a plan to get to Hawaii by hopping from island to island, when each intermediate island is only a potential island that might not exist.  [
intermediate computational step = island]   [what\rquote s in the black box = the length of the trip]  [the result of the computation = getting to Hawaii]
\par }\pard \li720\widctlpar\adjustright {
\par }\pard \widctlpar\adjustright {\tab 
\par }\pard\plain \s1\keepn\widctlpar\outlinelevel0\adjustright \b\f4\cgrid {
\par }\pard \s1\keepn\widctlpar\outlinelevel0\adjustright {TEXTURE SEGMENTATION  (Rosenholtz)
\par }\pard\plain \widctlpar\adjustright \f4\cgrid {\b 
\par Segmentation}{- seeing an edge between two textures
\par }\pard \fi720\widctlpar\adjustright {Pre-attentive (very fast 200-250- msec) vs. Attentive (slower, more cognitive)
\par }\pard \widctlpar\adjustright {\tab Two main models --- but really there\rquote s not much difference between the two
\par \tab \tab Statistical Models (Juesz, Beck)
\par \tab \tab Filter Based Models (}{\cf1 Malik}{\cf1 )
\par 
\par }{\b\cf1 When does segmentation occur?
\par }{\cf1 \tab Julesz, early \endash  \ldblquote when there is a distribution in the 2}{\cf1\super nd}{\cf1  order statistics\rdblquote  (recanted) 
\par \tab Julesz, late   }{\cf1 \endash  }{\cf1 \ldblquote }{\cf1 when there is a distribution in }{\cf1 textons, (\ldblquote quarks\rdblquote  of vision}{\cf1 )}{\cf1 
\par }\pard \fi720\li720\widctlpar\adjustright {Critique:}{\cf1  }{Center-surround filters can predict psychophysical results.}{\cf1 
\par }\pard \widctlpar\adjustright {\tab Beck-  }{\cf1 \ldblquote when there is a distribution in the }{\cf1 1}{\cf1\super st}{\cf1  }{\cf1 order statistics}{\cf1  of local features}{\cf1 \rdblquote }{\cf1  
\par }\pard \fi720\li720\widctlpar\adjustright {Critique: What is a feature?  
\par }\pard \widctlpar\adjustright {
\par }{\b Examples of Statistics using simple grayscale values of pixels.
\par }{1}{\super st}{ order statistics- probability of each pixel value occurring, mean, variance, histogram.
\par 2}{\super nd}{ }{order statistics- probability of }{a given pair of pixel values at a given location
\par 
\par 
\par }}