Texture classification by using co-occurrences of Local Binary Patterns

This paper proposes an efficient method for increasing the performance of Local Binary Patterns (LBPs). Although histogram of LBPs provides sufficient information of local pattern occurrences, it discards global interaction of these local patterns. We replace histogram of LBPs by making a Local Pattern Co-occurrence Matrix (LPCM) for the purpose of rotation and illumination invariant texture classification. Experimental results show significant improvement in terms of classification accuracy in comparison with conventional histogram based feature extraction method.