Weighted-guided-filter-aided texture classification using recursive feature elimination-based fusion of feature sets

In this work, a method is proposed for classification of texture images using a fusion of feature sets. Weighted guided filter based preprocessing technique has been performed using optimized cost function to enhance the discriminative property of different texture images. A hybrid model of normalized symmetrical gray level co-occurrence matrix parameters, histogram of oriented gradients, and Gabor features is used to extract the feature from the preprocessed images. The fusion model is fed to recursive feature elimination algorithm to select the appropriate feature sets for efficient classification. These feature vectors have been trained in two machine learning algorithms namely, multiclass support vector machine and extreme learning machine. It is experimentally demonstrated that proposed method achieves satisfactory efficiency on Outex, XU_HR, and UIUC texture image database. This method is also successfully applied on TEXDC database to identify the material of textile from images by recognizing the fibrous pattern of various textile images.