Local Binary Pattern with Neighborhood Relationship (LBPNR) -A novel approach for Texture Classification

Abstract

Texture is inclusive in natural images and also a prime feature of the availability of virtually all natural surfaces. Classification of Texture has accepted noticeable attention during the past decades. A fruitful classification of texture analysis provides a valuable representation of image texture. Many sophisticated methods have been developed for expressing image texture. Application of Local Binary Pattern (LBP) in texture classification is one of the prominent methods. Modifying this to a new method namely Local Binary Pattern with Neighborhood Relationship (LBPNR) is proposed here for texture classification. It determines two binary patterns where the first binary pattern obtains by considering the relationships of a central pixel with its neighboring pixels and the 2nd one comes from the relationships among neighboring pixels. Ultimately the final binary pattern drives from the exclusive OR operation of these two binary patterns. By applying this method a unique local region can be boldly distinguished from all others and yields an improvement in the classification process. In this paper, we used Brodatz Dataset including 16 texture classes where everyone contains eight 256 × 256 root images. The proposed model outperformed with a classification rate of 93.27% than that of other methods in this domain.