Compression d'images par fractales basée sur la triangulation de Delaunay. (Fractal image compression based on delaunay triangulation)

This thesis deals with fractal compression of still images based on the theory of iterated function systems IFS After an overview of the main lossy and lossless still image compression methods we introduce the IFS theory Then we detail the main fractal compression algorithms proposed in the literature A coder involves compari sons between two sets of blocks A block in the partition of the original image must be approximated by a collage block with a contractive function We therefore present di erent partitioning schemes used for the modelisation of the local similarities in the images Thereafter our algorithm based on the Delaunay triangulation is ex plained which constitutes the major contribution of our work This exible scheme allows to construct di erent triangulations in a image content dependant way In order to improve the coder decoder we propose di erent solutions The rst method allows to reduce the encoding complexity by the use of a scheme for quantizing the blocks The second method makes use of a mixed partition composed of triangles and quadrilaterals in order to improve the visual decoding quality at low bit rates To conclude this work we compare decoding results computed on di erent partitio ning schemes and we do a comparison between the block based fractal methods and recent hybrid methods merging fractal image compression and wavelet or Fourier transform methods