Generalized Local Binary Patterns for Texture Classification

Recently, the orderless Bag-of-Words (BoW) approach has proven extremely popular and successful in texture classification tasks [1, 2, 3]. Due to its impressive computational efficiency and good texture discriminative property, the BoW-based approach LBP [1] has gained considerable attention. In order to employ the advantages of the method of Zhang et al. [3] in combining complementary local features, and those of LBP in computational efficiency and smaller support regions, we developed two intensity-based descriptors CI-LBP and NI-LBP, and two differencebased descriptors RD-LBP and AD-LBP. An overview of the proposed approach is illustrated in Fig. 1. 1. A Brief Review of LBP. Images are probed locally by sampling greyscale values at a central point x0,0 and p points xr,0, ...,xr,p−1 spaced equidistantly around a circle of radius r centered at x0,0. Formally,