Combining local and global correlation for texture description

Local Binary Patterns (LBPs) and Covariance Matrices (CovMs) are two popular kinds of texture descriptors. However, local correlation brought by LBPs and global correlation brought by CovMs could not be directly combined to achieve enhanced discriminative power. This paper develops a powerful descriptor, named COV-LBP. Firstly, we propose a variant of LBPs on Euclidean space, named the LBP Difference feature (LBPD), which can be used to calculate any statistical image description. LBPD reflects how far one LBP lies from the LBP mean of a given image. It is simple, descriptive, rotation invariant, and computationally efficient. Secondly, by applying LBPD in multiple commonly used elementary features mapped from the original image, we provide a bank of discriminative features optional for CovMs. Consequently the information of LBPs and CovMs are embedded in a unified COV-LBP descriptor. Experimental results show that COV-LBP achieves promising performance on the public texture classification databases.