A new texture classification using circular difference and Statistical Directional Patterns

Abstract

The Local feature detection and texture description have acquired a lot of interest in recent years. In this paper, we propose a novel textual approach for texture classification accuracy. It's called the Circular Difference and Statistical Directional Patterns (CDSDP) which combines the mean and standard deviation of the circular difference to improve the texture classification. Artificial Neural Network (ANN), Support Vector Machine (SVM) and K- Nearest Neighbors (KNN) are used for texture classification step. Experimental results are based on an available CURETGREY database. A comparison study has been carried with other texture classification approaches. The proposed scheme could significantly improve the classification accuracy and reduce the time of classification compared with other methods.