Nouvelle méthodologie d'évaluation des paramètres de texture d'ordre trois
Many investigators have demonstrated that texture data improve the precision of a classification when they are combined with grey level information. However, the computation of textural parameters of order 2 often takes far too long. The problem is more complex when we have to compute textural parameters of order 3, which can grant a better precision in comparison with parameters of a lower order. A method, developed by Kourgli and Belhadj-Aissa allows a reduction in the computation time of the evaluation of second order textural parameters, based on the grey level co-occurrence matrix. This method is based on a technique allowing a double summation to pass to a simple summation. We have pursued the study further by developing textural parameters of order 3, showing how it is possible to express them by a simple summation instead of a triple summation. This method has been developed and tested with a Synthetic Aperture Radar (SAR) ERS-1 image, on a computer equipped with a Pentium processor of 800 MHz. The results obtained show that the new method of computation is at least two times faster than the classical method of computation of the co-occurrence grey level matrices.
Document Type: Research Article
Affiliations: 1: École Nationale Supérieure Polytechnique de Yaoundé, 653 rue 3386, BP 8390 Yaoundé, Cameroun; e-mail: [email protected], [email protected], [email protected] 2: Université de Marne-La-Vallée, Institut Francilien des Géosciences, 5 Boulevard Descartes, Champs-Sur-Marne, 77454 Marne-La-Vallée, France; e-mail: [email protected]
Publication date: 2003-05-01