In fractal compression a signal is encoded by the parameters of a contractive transformation whose fixed point (attractor) is an approximation of the original data. Thus fractal coding can be viewed as the optimization problem of finding in a set of admissible contractive transformations the transformation whose attractor is closest to a given signal. The standard fractal coding scheme based on the collage theorem produces only a suboptimal solution. We demonstrate by a reduction from MAXCUT that the problem of determining the optimal fractal code is NP-hard. To our knowledge, this is the first analysis of the intrinsic complexity of fractal coding. Additionally, we show that standard fractal coding is not an approximating algorithm for this problem.
[1]
Y. Fisher.
Fractal image compression: theory and application
,
1995
.
[2]
James A. Storer,et al.
Optimal Pruning for Tree-Structured Vector Quantization
,
1992,
Inf. Process. Manag..
[3]
Jianhua Lin.
Vector quantization for image compression: algorithms and performance
,
1992
.
[4]
David S. Johnson,et al.
Computers and Intractability: A Guide to the Theory of NP-Completeness
,
1978
.
[5]
Pierluigi Crescenzi,et al.
A compendium of NP optimization problems
,
1994,
WWW Spring 1994.
[6]
Michael F. Barnsley,et al.
Fractals everywhere
,
1988
.
[7]
Hannes Hartenstein,et al.
Fractal Image Compression - An Introductory Overview
,
1997,
SIGGRAPH 1997.