---------------------------------------------------------------------------
FIASCO («F»ractal «I»mage «A»nd «S»equence «CO»dec)
Copyright (C) 1994-2000
Ullrich Hafner <hafner@bigfoot.de>, http://ulli.linuxave.net
---------------------------------------------------------------------------
FIASCO is an image and video compression system based on fractal
coding which outperforms the well known JPEG and MPEG
standards. FIASCO has been developed during my Ph.D. thesis "Low
Bit-Rate Image and Video Coding with Weighted Finite Automata", Mensch
& Buch Verlag, ISBN 3-89820-002-7.
FIASCO consists of the following components:
afiasco
An X11 application to analyse and visualize a compressed image,
see separate manual page afiasco(1) for details.
bfiasco
A tool to generate the binary tree of the compressed image,
see separate manual page bfiasco(1) for details.
cfiasco
Command line application to compress images and videos, see
separate manual page cfiasco(1) for details. (File extension for
FIASCO files: .fco)
dfiasco
An X11 application to decompress and display images and videos,
see separate manual page dfiasco(1) for details.
efiasco
A tool to edit and concatenate several FIASCO streams,
see separate manual page efiasco(1) for details.
libfiasco.so
fiasco.h
A compression library to provide compression and decompression
of FIASCO files for other applications, see separate README
file doc/README.LIB or manual pages fiasco_coder(3),
fiasco_decoder(3) for details.
gimpplugin
A plugin for GIMP 1.1.x to read FIASCO images and movies
FIASCO is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2 of the License, or (at your
option) any later version.
FIASCO is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License
along with FIASCO; if not, write to the Free Software Foundation,
Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
---------------------------------------------------------------------------
Low Bit-Rate Image and Video Coding with Weighted Finite Automata
ABSTRACT
---------------------------------------------------------------------------
Recently, the need for efficient image compression systems can be
seen, e.g., in the rapidly growing field of world wide web
applications: not only still pictures but also small video sequences
are used to enhance the design of private and commercial web pages.
However, the speed of most internet connections is very limited, so
the amount of data to be transmitted must be dramatically reduced in
order to give a practical solution.
Weighted Finite Automata (WFA) exploit self-similarities within single
pictures and also sequences of pictures to remove spatial, spectral
and temporal redundancies. Their implementation combines techniques
from hierarchical methods related to quadtrees, vector-quantization,
matching pursuit, and fractal coding.
This thesis discusses important modifications of the original WFA
codec introduced by Culik and Kari which, in total, increase the
compression performance for low bit-rates to be on a par with
state-of-the-art codecs like embedded zerotree wavelet coding for
still images or H.263+ for video streams.
A crucial aspect during the development of our codec was the
performance of the matching pursuit algorithm used to generate an
approximation of the sub-images: we must carefully choose the set of
dictionary vectors as well as the quantization parameters of the
approximation coefficients. Last but not least, all coefficients and
indices must be efficiently stored using specially tuned models.
Not as crucial as the latter aspect but still important is the degree
of nondeterminism used by a WFA. We show that without nondeterminism
the approximation quality will quickly degenerate, but, on the other
hand, only a very restricted usage of nondeterminism can be stored
efficiently.
Finally, the application to video coding is investigated in detail.
We show that due to their simple mathematical structure, WFA provide
an ideal platform for efficient hybrid compression methods. So we
combined a hierarchical block based motion-compensation with the still
picture coder to form an efficient video codec. After all, we
optimized the run-time of the WFA video decoder to provide real-time
software decoding on most personal computer platforms.
---------------------------------------------------------------------------
---------------------------------------------------------------------------
