Local Binary Complement Pattern for Color-Inversion Invariant Texture Classification

Abstract

AbstractExisting Local Binary Pattern (LBP) methods cannot well handle the color-inversion changes. To address this problem, we propose a novel color descriptor called Local Binary Complement Pattern (LBCP) for color-inversion invariant texture classification. The proposed LBCP consists of three local operators, i.e., LBCP_S, LBCP_C and LBCP_O, all of which use the complement information to achieve the color-inversion invariance. Specifically, LBCP_S encodes the sign information of local neighboring differences in each channel. LBCP_C and LBCP_O encode the binary color values and the ordering information of central pixels, respectively, across color channels. After encoding the image using these three operators, we obtain multiple histograms and concatenate them as the LBCP descriptor. Experiments on several color texture databases demonstrate the effectiveness of LBCP for texture classification under color-inversion changes.KeywordsLBPColorFeaturesTextureClassification