home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!uknet!gdt!eesfd
- From: eesfd@gdr.bath.ac.uk (F Dudbridge)
- Newsgroups: comp.compression
- Subject: Re: Fractal compression
- Message-ID: <1992Aug13.232938.2854@gdr.bath.ac.uk>
- Date: 13 Aug 92 23:29:38 GMT
- References: <14026@chorus.fr>
- Organization: School of Electrical Engineering, University of Bath, UK
- Lines: 72
-
- I recently spent a week as a guest of Barnsley, so should like to
- comment on these recent postings:
-
- >Their performance has improved dramatically beyond what they were
- >talking about in BYTE a few years ago. Human assistance to the
- >compression is no longer needed and the compression time is
- >reasonable, although the more time and compute power you throw at the
- >compression, the smaller the resulting file for the same level of
- >quality.
-
- At the time of the Byte article they didn't have any realistic image
- compression algorithm. The block-coding technique they have patented
- is based on theory that was developed since that time. In fact, a lot
- of the work was done by Arnaud Jacquin, who has published these methods.
- Rumour has it that he and Barnsley had a bust-up over intellectual property
- rights: Barnsley's name is not on the papers, Jacquin's is not on the
- patents, so draw your own conclusions.
-
- >
- >now that they have obtained their patents they are talking
- >more about the method (since their 'competitive edge' is protected).
-
- The fact that Jacquin published before the patent was issued may yet cause
- a legal problem with the patent. In any case they are *still not* saying
- much about their methods. Despite the fact that everyone now knows roughly
- how it is done, I spent a week chatting to Barnsley and got nothing more
- from him than "it's block coding and involves searching". While such
- paranoia is somewhat understandable, it's fairly unusual to find a company
- where the programmers are prohibited from discussing work with the researchers,
- etc... which is the case at Iterated Systems.
-
- >Barnsley may have pissed a lot of people off because of the way he
- >presents (or doesn't) his work, but that doesn't mean that his method
- >is bogus. It is quite reasonable and it works.
-
- Is it true that he was made so unwelcome at SIGGRAPH that he left hurriedly?
-
- >>Compression ratios (unzoomed) seem to range from 20:1 to 60:1... The
- >>quality is considerably worse than wavelets or JPEG on most of the
- >>non-contrived images I have seen.
- >Iterated Systems Inc. has 10 full-time mathematicians who do nothing but work
- >on the fractal compression algorithms
- >The observation "quality is considerably worse" is quite dated by now.
-
- Yes the method does work. Yes, the results are good and are comparable to
- JPEG. However, the sales pitch does need to be looked at closely. Their
- compression ratios are often given by reconstructing the fractals at a higher
- resolution than the originals. Thus the compression ratios appear higher
- than they should. Also, the decoded colour may be in 8-bits, when the
- original was quantized from 24-bit data. Ie. the quoted compression is
- often reconstructed/compressed, as opposed to original/compressed.
- Since fractal images have infinite resolution, it is possible to reconstruct
- fractals to arbitrary resolution while avoiding pixelation, giving the
- impression of increased compression.
-
- Interestingly, I learned from one of the mathematicians that they had
- implemented their own wavelet compression algorithm, which outperformed
- the fractals. But as this defeated the point of the company, they preferred
- to ignore this result.
-
- There are 10 (very reputable) mathematicians working for Iterated Systems.
- One of these is Lyman Hurd who is co-authoring one of the new books with
- Barnsley. The other co-author, Louisa Anson, is Barnsley's girlfriend.
- When I met these people neither of the books were finished, so I think we'll
- have to wait a while.
- Having developed my own, different, algorithm for IFS compression I had
- been hoping to join the research team, but it seems that their expansion
- will now be in the marketing rather than research direction.
-
- Hope this is of interest to people
-
- -- Frank
-