LOCAL DIRECTIONAL PATTERN (LDP) FOR FACE RECOGNITION

This paper presents a novel local feature descriptor, the Local Directional Pattern (LDP), for recognizing human face. In the proposed feature, different edge responses around each pixel are meticulously encoded to distinguish all kinds of textures available in facial images. The LDP code is also robust enough to provide consistent representation in the presence of random noise and non-monotonic illumination variation. An LDP feature is obtained by computing the edge response values in all eight directions at each pixel position and generating a code from the relative strength magnitude. Each face is divided into small regions, from where LDP histograms are extracted and concatenated into a single feature vector to efficiently represent the facial image. The recognition is performed by template matching in accordance with CSU face identification evaluation system and evaluated with well studied FERET database. The recognition performance demonstrates the robustness of proposed LDP descriptor for representing appearance of facial image over other existing approaches including Local Binary Pattern (LBP).