A novel extended local-binary-pattern operator for texture analysis

The local-binary-pattern (LBP) operator has been proved to be a theoretically simple yet very effective multiresolution statistical texture descriptor in terms of the characteristics of the local structure, and has been applied in many areas. However, the extension to the LBP operator based on “uniform” patterns still has some shortcomings: it discards some important texture information and is sensitive to noise. A novel extended LBP operator for texture analysis is proposed in this paper. The new LBP operator classifies and combines the “nonuniform” local patterns based on analyzing their structure and occurrence probability. The new operator fully uses the texture information contained in the “nonuniform” local patterns, which is discarded by the classical LBP operators, and then becomes more robust against noise. Three experiments on the Brodatz texture database show the performance improvement of this novel extended LBP operator.