Title :  Segmentation of Text in printed pages
 Name  :  Ngin Choo, Lim  ( lngincho@stanford.edu )
 
 
 Description
 ------------
 
 Printed pages contain text, pictures and images information.  The purpose 
 of the project
 is to segment the text blocks from the non-text blocks (picture and 
 images).  There are 2
 main parts :-
 
 a)  Block segmentation in printed pages.  This step basically picks out the
      information content of the pages in blocks.  To do this, I will be 
 smearing
      the digitized printed pages first followed by edge detection to 
 determine the
      regions of information.  The parameter that can affect performance 
 will be the
      neighbourhood of smearing.
 
 b)  The next step is the classification of text and non-text 
 regions.  DCT-based, edge based
      classification techniques have been developed.  I plan to exploit the 
 regularity in text
      blocks (regular spacing between text lines, words and alphabets) in 
 doing the
      classification.   This regularity might be reflected in the 
 auto-correlation matrix of
      text blocks.