ABSTRACT
We present in this paper a new method for image description based on the description of regions that contains. Image is firstly segmented in meaningful regions and in the next step these regions are represented semantically in a tree structure. Regions are described by means the geometry of their outline contours and the description of encompassed internal regions. An XML language XLWDI (XML Language for Writing Descriptors of Images) is proposed to write these descriptions. This method may be used for image coding in multimedia applications or for image description in computer vision applications. Experiments conducted over real images are presented.
