Applications

Image Segmentation-

    I.  Xerox censoring of text
        a.    With the location of the image and the text areas known, certain areas could be 
               omitted for media (newspaper, magazine) template design and other such 
               applications.
        b.    Areas with image could have higher xerox resolution while text areas have sharper 
               contrast to boost xerox quality.

   II.   Clutter Analysis
         a.    Large images can be segmented into specific objects to depending on what the 
                operator designates.  For example:  finding a tank in a military satelite image.
         b.    Noise and other non pertinent information can be sensored with many applications
                to military intelligence causing enhanced detection of missiles, tanks, airplanes, 
                etc.

  III.    Caricature Design 
         a.    By segmenting a photograph into its respective parts: eyes, ears, nose, etc. a 
                computer generated caricature can be produced. 
 


Caricature Design-

    I.    Enhanced Gaming Technology
          a.    The algorithm used for caricature design for a 2D image can be applied to a 3D 
                 image such as a computer graphic used in computer and console arcade 
                 games.  Such technology would allow any 3D image (say, of a person playing an 
                 arcade game) to morph into any cartoon character or gaming player depending 
                 on the specific algorithm.  Any person could become a character in a game.
          b.    Special research for games based on sports stars and teams could be cut down.
                 The algorithm could animate people based on a series of videos, opposed to the
                 long hours of artist and computer rendering used today.

   II.    Enhanced Animation Technology
          a.     Artists could spend less time developing political and other such cartoons based 
                   on the real humans.
 
 

Future Work
 
  • Use advances statistical models for segmenting text versus image
  • Find optimal number of filter banks to improve segmentation
  • Segment specific areas of images or specific text
  • Segment a photograph of a face and produce a matching caricature