Study of informative feature selection approaches for the texture image recognition problem using Laws' masks

In the paper, the efficiency of two methods for feature selection based on Laws' masks is studied. These are a method of feature ordering in accordance with the criterion of discriminant analysis and t-statistic and a method of iterations through all pairs and triplets of features. The experimental results show that the classification error of the best group for features based on the standard deviation does not exceed the classification error of the best group for features based on the average energy.