Selective local texture features based face recognition with single sample per class

Local appearance-based methods have been successfully applied to face recognition and achieved state-of-the-art performance. In this paper we propose a local selective feature extraction approach based on Gabor filters and the Local Binary Pattern (LBP) approach to face recognition. A Gabor filter extracts the textural features from the face image and generates the binary face template using those features. The binary face template acts like a mask to extract the local texture information of the face image using a Local Binary Pattern technique. This selective local texture feature approach uses the histogram-based matching for face recognition. This method reduces the computation time considerably. This method also reduces the number of Local Binary Patterns into half compared to the existing LBP method. This proposed approach reduces the computation time for the FERET dataset by 45%. Experiments on well-known face databases such as FERET, Yale, Indian Faces and ORL show that this approach obtains consistent and promising results in the scenario of one training sample per person with significant facial variation.