Texture Classification Using Scale Invariant Feature Transform and Bag-of-Words

Abstract

Texture images can be characterized with key features extracted from images. In this way, they can be qualified with distinctive features. In this paper, a feature-based approach is presented for texture classification using Scale Invariant Feature Transform (SIFT) and Bag of Words (BoW) methods. The SIFT method is preferred because the features obtained by this method are invariant against such cases of rotation, angle of camera, ambient light intensity. UIUCTex and KTH-TIPS2-a data sets are selected which are widely used for classification. A success rate of 91.2% was obtained for the data set UIUCTex. This rate was determined as 72.1% for the data set KTH-TIPS2-a.