Compression orientée"Contenu" d'images.

Abstract

The representations currently used in image compression are based on separable transforms (DCT, wavelets, etc.). These do not capture the geometric regularity of the images along the contours. Thus, it is necessary to develop compressors that are able for the representation in capturing both this regularity along the contours and the regularity of zones. We are interested in this work, by an algorithm of compression of fixed images based on a new family of bases, the bandelets, for capturing singularities along contours. The proposed encoder uses also the embedded coding that provides the property of progressive transmission of the coded image. We propose in this work a new approach based on a new version of the SPECK algorithm "Set Partionning Embedded bloCK" applied to a fixed image transformed by bandelets . The proposed approach is characterized by its simplicity and gives better results compared to original SPECK because it exploits more the correlations between the bandelets coefficients .