UNE APPROCHE PARALLELE D'EVALUATION DES PARAMETRES DE TEXTURE POUR L'ANALYSE D'IMAGES DE GRANDE TAILLE
Résumé
The main objective of this paper is to develop a new image processing technique that could better be adapted to the processing of large and complex images, especially SAR images, in relation to the former methods of our works. We propose here a new approach of textural parameters evaluation which is about (Ng+1)n-1 times faster than the co-occurrence matrix approach considered as classical, where n is the order of the textural parameter and Ng being the maximum grey level in the image. This approach is based on a new modeling of textural parameters of a generic order n>1 equivalent to the classical formulation, but which is no longer based on frequency matrix (co-occurrence matrix in case of order n=2). The resulted model enables us a gain of about (Ng+1)n octets of memory space required by the computer for the processing of these images. Further more, this technique is a parallel approach of computation. It enables a possibility of real time computation independently to the image sizes.