A Textile Image Classification based on Texture and Shape Features

Abstract

In the last few decades, with the progress of image processing textile designer used a most appropriate feature extraction method for both pattern design and analysis of fabric materials. In this proposed method, we have used an efficient features extraction method. Firstly, image is resized and converting to a gray scale level. Then features are extracted on the basis of shape and texture feature extraction methods. Discrete Wavelet Transform (DWT) and Local binary pattern (LBP) are used to find the texture features and for the shape features Invariant moments (IM) is used. Canny edge detection method is used to remove the noise to smooth the image. To apply a robust system, we used PCA to reduce the dimension of feature descriptors. In experimental results, we used 300 batik images for a same and different Pattern. Support Vector Machine (SVM) method is used for a Classification Method. The overall result shows that a combination of multi features gives efficient results as compared to single features extraction method. The accuracy of our system is about 97.6 % after applying the PCA algorithm.