The effect of face regions' weighting to face recognition performance with local binary patterns

Local Binary Patterns (LBP) is a successful feature extraction method used in pattern recognition problems such as face, texture, and gender recognition. LBP extracts features by comparing the gray level values of each pixel in the image and its neighboring pixels. It is known that some parts of the face such as eyes, nose and mouth contribute more to the classification performance than other regions. Therefore, the face image is divided into rectangular regions. Local features are obtained by generating LBP histograms for each region. It is aimed increase the classification performance by setting the weights to the local features according to the importance of the corresponding region. In the experimental studies, AR and Expanded Yale face Database B are used to determine the performance of the method. In the classification stage, the subspace-based Linear Regression Classifier (LRC) and the Chi-square (χ^2) statistics are used. As a result of the experimental studies, it is observed that the performance of the LRC classifier varies according to the database used, but χ^2 statistics achieves high classification performance independent of the database used.