An Improved Pairwise Rotation Invariant Co-occurrence Local Binary Pattern Method for Texture Feature Extraction

An improved pairwise rotation invariant co-occurrence local binary pattern (PRICoLBP) algorithm blended with local texture information is proposed in regard to the problem of high feature dimensions to be calculated, poor rotation invariance and sensitivity to illumination change with the PRICoLBP algorithm for texture feature extraction. First, the binary sequence of image pixel points is maximized and minimized respectively, which yields the coordinates of two neighborhood pixel points, and the coordinates of co-occurrence point pairs are calculated from the coordinates of central pixel point and acquired coordinates of neighborhood pixel points. Second, texture information about each pixel point in the image is retrieved using the complete local binary pattern (CLBP) algorithm. With the same classifier, the proposed algorithm gets an increase of respectively 0.17, 0.24, 2.65, 2.39 and 2.04 percent in rate of identification over the PRICoLBP algorithm in the experiment of classifying Brodatz, Outex (TC10, TC12), Outex (TC14), CUReT and KTH_TIPS databases. Experiment results show that the proposed algorithm has better identification effectiveness when processing an image with rotation variant texture and diverse illumination conditions.