Face recognition with varying facial expression using local directional number pattern

This paper presents a method for face recognition using local directional number pattern (LDN). LDN adopts encoding the directional pattern of the face and produces a more selective code than methods listed in the literature. The directional patterns are computed using the compass mask and the information is encoded using the prominent direction indices (directional numbers) and sign — which distinguishes among similar structural features. The proposed method divides the face into several blocks, and computes the LDN features for each block. Later feature vector for each block is constituted based on these LDN features. The face descriptor is obtained by concatenating these feature vectors. An experimental result shows that the proposed LDN performs consistently under noise, illumination, expression, and time lapse variations. The performance is evaluated using FRET databases.