Fractals in Multimedia Imaging: From Compression to Image Indexing to Watermarking

Thursday, January 18, 2001 - 2:00pm - 3:00pm
Keller 3-180
Jean-Luc Dugelay (Institut EURECOM)
Fractal compression is a particular form of block-based vector quantization and consists in expressing an image in terms of local self-similarities. But despite a series of optimizations, the elaboration of hybrid schemes, the possible extension to colour and moving pictures, etc., fractal based image compression has never reached the initial expected success. However some notions and properties included in the fractal theory can present a fair potential in some other domains of image processing such as zoom, indexing or also watermarking.

Indeed, the compression inspired by the fractal theory possesses some invariance properties by affine (geometric and photometric) transformations, such as change of scale factor, that can be useful to some applications.

Typically, this representation space is useful when one looks for images, that are globally similar or close to a given image sample, inside an image database, or also when one hides a message inside an image and would like to be able to recover it later in a reliable manner, even if the image has undergone one or several non destructive attacks.