🚀 go-pugleaf

In article <sjreeves.920810102554@eng.auburn.edu> sjreeves@eng.auburn.edu (Stan Reeves) writes:
>
>I read in SP Magazine that "coding of halftone images, once thought to
>have great promise, has not been successful."  I would like to find out
>more information about this.  Can someone summarize what's been done in
>this area and just how "unsucessful" the efforts have been?
>
>Stan Reeves
>Auburn University, Department of Electrical Engineering, Auburn, AL  36849
>INTERNET: sjreeves@eng.auburn.edu


I can't give an exhaustive summary, but can perhaps get you started.
I'll assume that you're referring to binary (monochrome) rather than
color halftoning.  There seems to be relatively more research activity
in the area of halftoning per se than there is in the area of
compression of halftoned images.  If you are interested in halftoning
by itself, see the Proceedings of the ICASSP-91, ICASSP-92, and
SPIE-92 conferences for some of the most recent proposed methods.  A
few (not all) of the relevant names are Neuhoff & Pappas, Zakhor, and
Allebach.  There is also Ulichney's book, _Digital Halftoning_, put
out by MIT Press (1987), which presents a variety of analyses on a
variety of halftoning techniques.

There are two forms of compression that you might be interested
in---lossless and lossy.  For lossless (noiseless, no information
loss) compression, the most promising method (in my opinion) is IBM's
Q-coder, which is documented by about five articles in the _IBM
Journal of Research and Development_, November, 1988 (Vol. 32, No. 6).
These articles have other useful references as well.  IBM is working
to have their system established as part of the Joint Binary Image
Group (JBIG) standard for compression.  It seems to be an excellent
noiseless coding scheme, which routinely produces coding rates that
achieve or beat the first-order entropy for a variety of halftoned
images. 

For lossy compression, there have been a couple of published works in
which the halftoning and compression are done separately (Neuhoff &
Pappas, ICASSP-91; Y.-M. Chao, "An Investigation into the Coding of
Halftone Images," Ph.D. thesis, MIT, 1987).  I co-authored a paper
(with Chou, Riskin, and Gray) at
ICASSP-92 showing that a better idea is to do the halftoning and
compression jointly.  In that paper, we showed an image compressed at
about 11:1 which looks pretty decent (we think!).  We are continuing
to work on this problem here at Stanford.  Hope this helps....

Rick A. Vander Kam
rvk@isl.stanford.edu