A New Local Binary Pattern in Texture Classification

The e LBPs extract local structure information by establishing a relationship between the central pixel and its adjacent pixels. However, it is very sensitive to the change of the central pixel .In this paper,we choose a circle with the radius of 1 instead of a single center. The method is proposed for texture classification by comparing the information between the neighbors and the new center pixel. In order to decrease the feature size and increase the classification accuracy, both LBC-like feature and CLBP feature were used in the proposed method .Experiments are carried out on Outex and UIUC databases. The experimental results demonstrate that this method perform effectively.