A research of face feature extraction based on MB-LBP

The technology of face recognition is a well-known subject. This paper presents a face feature extraction method based on MB-LBP. In this method, by means of pre-locating the feature points, a Multi-Block Local Binary Pattern (MB-LBP) is intensively studied and related algorithm implemented. The proposed method is then tested by using Yale face database, and how MB-LBP scale (or size of training set) influences on the face recognition results is explored. Evaluation results confirm that the implemented algorithm is both valid and effective under certain conditions.