Texture Classification Using Features from Multi-level Local Binary Patterns

This paper introduces a novel texture analysis method that combines simple histogram features with the local binary patterns (LBP) approach. By utilizing LBP to derive "pattern images" from a texture image, we obtain additional information sources to be explored using histogram analysis. We evaluated different combinations of "pattern images" as well as histogram features, yielding impressive results in three benchmarking databases. Our approach achieved accuracy rates of 99.54%, 99.25%, and 92.98% for the Vistex, Brodatz, and USPTex databases, respectively. These outcomes demonstrate the ability of the proposed hybrid method to generate a highly discriminative feature vector for effective texture classification.