Face and Expression Analysis Using Local Directional Number Patter

Abstract

LDN encodes the directional information of the face’s textures in a compacted way, generating a more discriminative code than current methods. LDN encodes the structure of a local neighborhood by analyzing its directional information. The edge response in the neighborhood is computed in eight different directions with a compass mask. From all the directions the top positive and negative directions are chosen to produce a meaningful descriptor for different textures with similar structural patterns. The extracted information is encoded using the prominent direction indices such as directional numbers and sign which allows us to distinguish among similar structural patterns that have different intensity transitions.