Title:  Elder-Zucker image compressions: implementation & analysis
  
Chad Netzer         cnetzer@stanford.edu
Dion Monstavicius   dionm@stanford.edu
  

We propose implementing the still image compression scheme proposed by
Elder et al in ref 1.  This scheme is as follows:

Encoder:

1)  Perform a multi-resolution edge detection scheme to find relevent
edges in an image.

2)  At these edge boundaries, keep the intensity information and an
estimate of the first derivative for transmission.  Depending on the
image, this can reduce complexity by order N.    
                                                
3)  perform compression on this data for transmission.  Chad and I
may investigate possible implementations for this part.  (Elder does not
detail this part)

Decoder:

1) With the boundary information, reconstruct the image by solving
Laplace's equation using the edge information as a boundary condition.  We
expect this to be the most challenging part of the project.

2) Elder illustrates how this step still retains artifacts, and proposes
an additional processing step to supress these artifacts.


Project goals:
Project goals:

1)  Sucessful implementation of scheme.
2)  Rate distortion analysis of compression of technique.
3)  Proposals for practical encoding of edge information.



References:

1) J.H. Elder et al, "Are Edges Incomplete?" International Journal of
Computer Vision, 1999 vol 34 no 2-3 p 97.

2) J.H. Elder et al. "Local scale control for edge detection an dblur
estimation" IEEE transactions on pattern analysis and machine
intelligence, July 1998, vol 20, no 7, p 699.