A novel texture descriptor: Homogeneous Rotated Local Binary Pattern (HRLBP)
Abstract
Invariant rotation in the application of texture classification is generally beneficial due to the material: camera or auto rotation, which can affect objects captured by arbitrary angles. This letter, introduce a new, efficient rotation invariant descriptor for texture analysis appointed Homogeneous Rotated Local Binary Pattern (HRLBP). The goal of this novel method is to take more account of the intrinsic characteristics of the images in rotation changing by using the incidence of homogeneity tolerance h provide from General Adaptive Neighborhood (GAN). A significant features are generated from HRLBP by thresholding the center and each neighbor pixels with an homogeneity tolerance value which help to get more efficient and discriminating features for rotation variation further multi-scale changing owing to use a variation of the parameter of homogeneity tolerance h and radius R. The experiments are evaluated using two publicly available texture database OTC10, furthermore, HRLBP shown a high performance in classification accuracy for problem of rotation variation.