Genetic Approach and AURAY Criterion for Optimizing the Attribute Vector: Application to a Supervised Classification of Texture Images

摘要

Selecting the parameters for the classification is a delicate procedure. We present in this work a new method for selecting the parameters based on the genetic approach which optimizes the choice of parameters by minimizing a cost function. This function is defined by a new criterion that we have proposed. The proposed criterion is inspired from the AURAY criterion and the correlation approach. AURAY criterion estimates the discriminate power of the set of parameters based on a measurement of the classes separability and compacity. We have introduced in our criterion the correlation approach as an additional factor for selecting the parameters. This factor made it possible to avoid the information redundancy which is a major inconvenient for the selection criteria based on a measure of the classes separability and compacity. The proposed approach is validated on some texture images. The experimental results show the good performance of the proposed method.