A model of optimizing features of texture images based on three-class schema and two-step mapping architecture

Feature selection of confusable textural images is a difficult and challenging problem. Based on the theory of rough set and with classification error rate as the standard, a three-class schema and two-step mapping model aiming to reduce and to optimize feature set is proposed in this paper. Feature set is optimized with the proposed model and then the optimized feature set is applied to classification. In this paper, classification accuracy is used to evaluate the model. Finally test samples of texture images are processed with the proposed model and also texture features obtained by the model are compared in classification capacity with texture features that are obtained with other methods. Thus feasibility and simplicity of this model in feature optimization is confirmed.