Face recognition using extended local binary patterns and fuzzy information fusion

This paper presents a novel and efficient approach for face recognition based on extended local binary patterns (LBP) and fuzzy information fusion. Each facial image in training set is divided into a certain number of sub-regions after a simple image preprocessing, and all training sub-regions from the same position construct a new training subset. The extended LBP method is used to extract local feature of the new training subset independently and then a set of feature histogram vectors can be obtained. For a given unknown facial image, sub-feature histogram vectors of corresponding sub-region are gained after the same preprocessing and partition. The χ² distances between test sub-regions´ histogram and trainings´ are obtained to calculate their membership grade. After fuzzy classification of local sub-features, strategy of fuzzy integrate is adopted to fuse each of them. At last the result of classification is determined by the principle of maximum membership. Experimental results on the ORL and FERET databases show competitive performance.