Writers authentication and fractal compression

In this paper we propose a new approach that allows authenticating a writer. To achieve this goal the autosimilarity properties in writings are used, that is to say we are extracting some invariant pieces of the writing, invariant patterns that are characterizing the texts of a writer. From a practical point of view, these invariant shapes are determined by adaptation of techniques that have been developed in the fractal compression process. During a learning phase the invariant patterns are extracted and in a second step they are used as a reference base in order to analyze an unknown writing. The results of the analysis are quantified by measurement of the quality of a compression/decompression process elaborated using the reference base associated with different.