home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.compression
- Path: sparky!uunet!centerline!noc.near.net!news.Brown.EDU!qt.cs.utexas.edu!cs.utexas.edu!zaphod.mps.ohio-state.edu!rpi!psinntp!psinntp!newstand.syr.edu!mothra.cns.syr.edu!dlwood
- From: dlwood@mothra.cns.syr.edu (David L Wood)
- Subject: Re: Compression technique mentioned in PCW
- Message-ID: <1992Aug19.010028.451@newstand.syr.edu>
- Keywords: snake oil, something for nothing
- Organization: Syracuse University, Syracuse, NY
- References: <1992Aug18.155518@sees.bangor.ac.uk>
- Date: Wed, 19 Aug 92 01:00:28 EDT
- Lines: 37
-
- In article <1992Aug18.155518@sees.bangor.ac.uk> mather@sees.bangor.ac.uk (Paul Mather) writes:
-
- [ bit deleted ]
- >Sounds like hot air? (Yes, IMHO.:)
- Yes.
-
- >... the decimal fraction was compressed by
- >determining two numbers, a and b, such that a/b = N. To transmit the
- >original file, one need only transmit a and b. Decompression would
- >involve calculating a/b and enumerating the result.
-
- Well, If N = a/b. then a= b*N. You also know that N < 1 meaning that
- a < b. If you factored out of N the value of the power of ten (negative)
- that makes it a mantissa then you have a factoring problem. The point
- of this now is to find out all the factors of N and a sufficiently small
- value for a. For some cases, sure this MAY produce fewer bytes than the
- original N.
-
- You could try and try to brute force calculate a/b to approximate N
- (I did this earlier today, eventually deciding on a technique which
- always hopped between just over, and just under the real value of N,
- which is a linear approach, in most cases.) --- or, since the values of
- a and b that you get as a result are more easily calculated as factors,
- then you're better off throwing this out and choosing a tree priented approach.
-
- If this is at all similar to arithmetic encoding (which I have no books on :()
- then it seems that you would have to 'tag' those groups of bytes that no
- a/b combination could be found for.
-
- If this isn't clear, think of the a/b combination for 0.127. only 127/1000
- will produce that decimal. 127 is a prime, no factors, no savings.
-
- --
- +-----------------------------+-----------------------------------------+
- | dlwood@mailbox.syr.edu | Gigs and gigs of NiCad memory, bummer.. |
- | Cybernaut, with a thought. | Why buy the leading brand; 90% hate it. |
- +-----------------------------+-----------------------------------------+
-