Texture Classification Based on Completed Modeling of Local Binary Pattern

Local Binary Pattern (LBP) algorithm is a typical texture analysis method combined with structural and statistical texture. LBP has a drawback of losing global spatial information, while global features preserving little local texture information. By extension of the standard LBP algorithm, this paper focuses on Completed modeling of Local Binary Pattern (CLBP), which is composed by the center gray level, sign components and magnitude components. Two experiments were carried out to test the classification ability of CLBP by Brodatz and UIUC image database. Results show that CLBP algorithm has the highest average classification accuracy of 86.63% (Brodatz) and 83.29% (UIUC). But the standard LBP only obtained the highest average classification accuracy of 79.97% (Brodatz) and 57.59% (UIUC). So the CLBP has a better texture feature extraction capabilities than standard LBP, and the different neighborhood scale of CLBP has a large influence to the classification accuracy.