Extended Contrast Local Binary Pattern for Texture Classification

Jie Sima, Yongsheng Dong, Tianyu Wang, Lintao Zheng, Jiexin Pu

Considering the limitation that LBP only focuses on the sign feature in extracting the texture feature as well as its low recognition rate, we in this paper propose an extended contrast ratio local binary pattern for texture classification. The extracted features include its sign feature, energy feature and its center pixel feature, which aims at constructing the histogram based on the features of the sign energy center pixel gained before. Then we perform texture classification by employing the Chi-square distance and the nearest neighbor classifier. Experimental results reveal that our proposed method outperforms several representative texture classification methods.