NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/ai/neuraln/4632

Internet Message Format | 1992-12-21 | 11.8 KB

Xref: sparky comp.ai.neural-nets:4632 sci.image.processing:1439
Newsgroups: comp.ai.neural-nets,sci.image.processing
Path: sparky!uunet!zaphod.mps.ohio-state.edu!usc!rpi!sassoj
From: [e-mail] (John J. Sasso Jr.)
Subject: RESPONSES TO IMAGE PROCESSING REFS
Message-ID: <[e-mail]>
Nntp-Posting-Host: jec301.its.rpi.edu
Organization: Rensselaer Polytechnic Institute, Troy NY
Date: Mon, 21 Dec 1992 16:45:15 GMT
Lines: 340

From [e-mail] Fri Dec  4 10:35:17 1992
Date: Fri, 4 Dec 1992 10:32-EST
From: [e-mail]
To: "John J. Sasso Jr." <[e-mail]>
Subject: Re: TEXTS ON IMAGE COMPRESSION?

>including advanced techniques)?  I am working with gray-scale images.
>The images I work with generally have no sharp variation points (they look
>as if they were low-pass filtered), so can anyone suggest a compression

Assuming that you do not require lossless compression, JPEG would
probably be a good bet.  See the JPEG FAQ in comp.graphics or
news.answers for software and references.  Also read the
comp.compression FAQ.

            regards, tom lane


/********************************************************************************/


From edsr!edsdrd![e-mail] Fri Dec  4 13:38:14 1992
Date: Fri, 4 Dec 92 13:25:21 EST
From: edsr!edsdrd![e-mail] (Matthew Maa)
To: uunet!aix.rpi.edu![e-mail]
Subject: Re: TEXTS ON IMAGE COMPRESSION USING NNs?

I don't think there is a book particularly devoted to the
subject you are looking for.  There are books covering
NN applications to image compression.  The following is
the abstract of a paper I recently wrote.  If it fits your
need, I can email you the complete paper in postscript
format.

Matthew Maa, Ph.D.
[e-mail]

--------------- Abstract start here ---------------------

               Neural Networks for Image Compression:
                           A Survey

                          Matthew Maa

                    Electronic Data Systems
                Detroit Research and Development


This paper surveys the neural network related methods being used for
image compression.  These methods can be categorized into two major
groups, direct methods and vector quantization methods.  Direct methods
mostly adopt three-layered, feed-forward networks in which input and
output layers have the same number of neurons and hidden layers have
fewer neurons.  These networks are trained via certain pre-selected
learning schemes during the training stage so as to generate at output
layers (approximately) the same images that were given to input layers.
The first halves of these networks, from input layers to hidden layers,
are encoders (compressors), while the second halves, from hidden layers
to output layers, are decoders (decompressors).  The outputs of neurons
of hidden layers are considered compressed images since hidden layers
have fewer neurons.  For vector quantization methods, networks are used
to generate a proper codebook of reference vectors, where a vector is a
subimage.  A subimage can then be encoded by using the index of its
best-match reference vector in the codebook.  Compression is achieved
as a vector index having a shorter binary representation that the
vector itself.  Other related image compression methods, including
fuzzy systems and iterated functions systems as implemented by neural
networks, will also be addressed.  All methods surveyed in this report
are lossy, that is, they only render from compressed files an approxi-
mate image to the original.


/********************************************************************************/

From [e-mail] Fri Dec  4 16:04:27 1992
Date: Fri, 4 Dec 92 15:04:12 CST
From: [e-mail] (Armando Manduca)
To: [e-mail]
Subject: image compression text


John -

A book I liked very much is "Digital Image Compression
Techniques" by M. Rabbani and P. Jones.  It's put out by
SPIE Press and is their tutorial text TT7.  It's very
well-written and thorough, starts out basic, and goes
thru fairly advanced.

--------------------------------------------------------------
Armando Manduca                Internet:   [e-mail]
Dept. of Physiology and Biophysics
Mayo Foundation                talk:       (507)-284-8163
Rochester, MN 55905            FAX:        [phone]
--------------------------------------------------------------

/********************************************************************************/

From [e-mail] Fri Dec  4 18:44:32 1992
Date: Fri, 4 Dec 92 18:44 EST
From: [e-mail] (Paul Vargo)
To: [e-mail]
Subject: Re: TEXTS ON IMAGE COMPRESSION?
Newsgroups: sci.image.processing
Organization: UMCC
Cc:

Try "Practical Image Processing In C" written by........whoops
forgot the name.  It's published by John Wiley And Sons, Inc.
605 Third Ave. New York, N.Y. 10158-0012

It covers the basics and a lil' on the advance techniques.
A very good book IMO.  Oh...Craig A. Lindley wrote it....
and he also works for Hewlett Packard.  So you may even
be able to reach him via e-mail.

I pretty proud of myself! :)  I can't believe I remembered all
that stuff....I'm usually poor at it. :)

That's a start for ya.

--
|================================================|========================|
| Paul J. Vargo, II            [e-mail] | "Love is a many        |
| AC Rochester Engineering & Research Center, GM | splintered thing" A.E. |
|================================================|========================|

/********************************************************************************/

From [e-mail] Sat Dec  5 17:19:55 1992
Date: Sat, 5 Dec 92 15:19:42 MST
From: [e-mail] (Joseph Provine)
To: [e-mail]
Subject: Re: TEXTS ON IMAGE COMPRESSION?
Newsgroups: sci.image.processing
Organization: ECE Department, U. of Calgary, Calgary, Alberta, Canada
Cc:

Hi!
Digital coding of waveforms
by Jayant and Noll

Digital Pictures
by Netravali and Haskell

Digital Image Processing
W. Pratt

Some texts are included in comp.compression-faq

I recently came across a book that includes recent developments as well:

Digital Image Compression
by Rabbani and Jones

SPIE's selected papers on Image Coding.

The new version of Gonzalez and Woods book (Digital Image Processing)
includes a better chapter on Image Coding compared to the previous version.

This is pretty much the popular (at least that is what I think) stuff.

Now about Entropy coding:
You can refer to Review papers on Image Compression by Jain (Proc. IEEE,
Mar. 1981, pp:349-389), Netravali and Limb (Proc. IEEE, Mar 1980, pp366-406).

The specific papers are:
Huffman's paper (Proc. I.R.E, Sep. 1952, pp. 1098-1101)
Arithmetic Coding:
Rissanen and Langdon (IBM J. Res. develop. Mar. 1979, pp. 149-162)
Witten, Neal and Cleary (Commn. of ACM, June 1987, pp. 520-540)

Also Welch's paper on LZW coding gives brief account of the above
techniques too(IEEE Computer, June 1984, pp. 8-19).

Hope this helps you...
Good Luck.
[e-mail]

/********************************************************************************/

From [e-mail] Sat Dec  5 22:38:17 1992
Date: Sat, 5 Dec 92 22:38:07 EST
From: [e-mail] (Bart Bartholomew)
To: [e-mail]
Subject: Re: TEXTS ON IMAGE COMPRESSION USING NNs?
Newsgroups: comp.ai.neural-nets
Organization: The Great Beyond
Cc:

    John, there was apparently a paper on
the topic of image compression using NNs presented at the
recently completed IJCNN conference in Peiking.  A summary
of the conference was published a week or so back.
    BartB

--
"I'll need a sample."  "Come again?"  "No, once is enough." -Joel to Holling
If there's one thing I just can't stand, it's intolerance.
*No One* is responsible for my views, I'm a committee. Please do not
infer that which I do not imply.  [e-mail]

/********************************************************************************/

From [e-mail] Mon Dec  7 15:47:12 1992
Date: Mon, 7 Dec 92 15:55:29 -0500
From: [e-mail] (Samuel S. Paik)
To: [e-mail]
Subject: Re: TEXTS ON IMAGE COMPRESSION?
Newsgroups: sci.image.processing
Organization: Digital Workstations Software Group
Cc:

In article <[e-mail]> you write:
>  Can anyone refer me to any texts on image compression/coding (possibly
>including advanced techniques)?  I am working with gray-scale images.
>The images I work with generally have no sharp variation points (they look
>as if they were low-pass filtered), so can anyone suggest a compression
>scheme which is best for compressing such images?  Also, what is entropy
>coding, and how is its performance?  Furthermore, any references to research
>articles on advanced image compression techniques would be greatly
>appreciated.
>
>                        John

I'd suggest you take a look at the comp.compression FAQ and the
JPEG FAQ sheet (also found in comp.compression).

As an intro text, I'd recommend "Digital Coding of Waveforms", N.S.
Jayant and Peter Noll, Prentice-Hall, 1984. (This is real hard to find
in a book store, but you might have luck in a good engineering
library).

A source for the latest in image compression is the ICASSP proceedings,
the "International Conference on Audio, Speech, and Signal Processing".
This is a mammoth conference (I think the last few years they've
had over 5000 pages each year), and you'll have to wade through a lot.
Much of it won't be understandable without a background however.

You can get a nice freely-available and distributable JPEG codec
via ftp.  Try looking on ftp.uu.net in the graphics directory.

Entropy coding maps from symbols to bits based on a probability model
of the source s.t. the number of bits produced is minimized.  Assuming
you have a good model of your source, an optimal entropy coding (such
as arithmetic coding) will produce the best compression on the
average.  (The catch, of course, is getting a good model of your
source).

Sam Paik

p.s.  Sorry, if this seems a bit nebulous, but my job has very little
      to do with image compression, so all my refs are at home.

/********************************************************************************/

From [e-mail] Tue Dec  8 03:55:13 1992
Date: Tue, 8 Dec 92 10:55:03 +0200
From: Jyrki Alakuijala <[e-mail]>
To: [e-mail] (John J. Sasso Jr.)
Subject: TEXTS ON IMAGE COMPRESSION?
Reply-To: [e-mail]

If you have printed text, try convolution of your
font set. This should be possible to implement by using FFT.

I would try to code hand written text with some primary elements such as
arcs and lines and so on... I don't know if it's a good method.

Then you have all the standard image compression methods, of course.

You might want to get rid of the low pass effect by some non-linear
filter...

You should scan through you librarys IEEE magazines... atleast on the
last issue of IEEE medical imaging, there was a nice compression
method based on linear prediction...

Why don't you just treshold the images and compress the binary image? :-)

--Jyrki Alakuijala

/********************************************************************************/

From [e-mail] Wed Dec  9 17:10:35 1992
Date: Wed, 9 Dec 92 15:10:54 MST
From: [e-mail] (sane sachin arvind)
To: [e-mail] (John J. Sasso Jr.)
Subject: Re: TEXTS ON IMAGE COMPRESSION?


It would be nice  to try Vector Qyuantization on these type
of images. I am also working in this area and VQ gives certainly
very good results.
If you want to try ready made programs go for JPEG. It's based
transform coding.
The book "Fundamentals of Digital Image Processing" By A.K. Jain
chapter12 might give to basic idea of image compression. You may
need to read basics from previous chapters. I have written code to
do image compression using VQ. If you want I can send you the code.

Best Luck!

- Sachin Sane


/********************************************************************************/