Face Recognition Using Local Multiple Zig-Zag Structure Pattern

This paper presents a novel algorithm local multiple zig-zag structure pattern (LMZSP) for face recognition. The proposed method modifies the Weber’s ratio and obtains the local multiple pattern (LMP) images by exploring the relationship between center and neighboring pixels. In order to make these patterns more robust to noise and intensity variations, local multiple zig-zag structure pattern (LMZSP) has been proposed by applying the local zig-zag graph structure (LZGS) on the four LMP images. Further, the resulting high dimensional feature vector is converted to low dimensional feature using linear discriminant analysis (LDA) algorithm. The robustness of feature is tested and analyzed on faces from uncontrolled environment. The recognition results clearly proves the effectiveness of the proposed method as compared to the other state-of-the-art approaches.