Similarity assessment with local binary patterns

Identification from human face plays an important role in social interaction, such as recognition and security. Thus facial information processing is an active research area in pattern recognition. The similarity of a child´s face to parent faces is evaluated in this paper. Intermediate face images are generated by morphing parents face images in specific proportions as model input images to the method. Then, feature vectors of the generated model images are obtained using the local binary patterns (LBP) and saved to a database. Euclidean, Manhattan, Chebyshev distances and Chi square statistic are used to measure the distance between given child´s face and generated model faces using feature vectors and Borda voting is used to evaluate the similarity.