Email from Bob to Dan, 11/18/2001:

Subject:  SVP with two images and excerpts posted



Through our homepage:

http://www.ccs.neu.edu/home/futrelle/svp/vsp-index.html


In addition, I've posted a pic of a flea just plowing through
an abutting line.  It broadens briefly as it goes through
the line, then narrows right back down.

From a model point of view, this has some problems.  We need
to recognize the extreme local error.

Two ideas:

1.  A single flea for a straight line should hold a set of
segments, so it really will stop and hop.  It will notice a
local problem, end a segment there, hop ahead and start a 
new separate segment, but all part of the same flea object/result.
This also makes sense if a flea has to proceed in two opposite
directions from a starting point.  Since the line may be
absolutely clean at the point the two processes start, it
wouldn't make sense to say that they're two separate segments.

2.  We need to visualize the white border region of the line
model.  In the current fat pixel system (I haven't downloaded
your improved one yet), there's no visual indication of the
total extent of the flea on both sides. Somewhere in those
four subpixels you should be able to figure out what we need.
One way would be to have the yellow and blue occupy the same
spot as alternatives.

BTW:  The jpegs have gray level pixels, not sure how you're
handling that.

See you Wednesday in any event.

 -- RPF
 
 
PS: The almost perfect horizontal line below the "1" in the
good excerpt appears to slope down to the right with about
a 1% slope.  Small, but you can see it at large pixel sizes.