LBP Texture Analysis Based on the Local Adaptive Niblack Algorithm

Local Binary Pattern (LBP) proposed by Ojala T is a typical texture analysis algorithm. It is widely used in the texture classification domain and etc. In this paper, we consider the local pixels’ effects among them efficiently when the LBP algorithm is used. In our proposed LBP algorithm, the central pixel in the local region was added into the procedure to calculate the local threshold using the Niblack algorithm. The histogram equalization is always used to actualize the illumination invariability in LBP algorithm. But the histogram equalization can not avoid the variety of local grayscale caused by the illumination. In order to avoid this weakness, a kind of illumination factor is brought into the Niblack algorithm. The values of the illumination factor can be got adaptively with the changes of illumination. By using the improved Niblack algorithm, the local threshold can be calculated adaptively. The proposed LBP algorithm in this paper can get more texture information and it has illumination adaptability.