Non Binary Local Gradient Contours for Face Recognition

TL;DR

This paper introduces a novel face recognition feature extraction method based on Non Binary Local Gradient Contours derived from Information sets, aiming to improve discriminability over traditional LBP and LDP techniques.

Contribution

The paper proposes a new feature extraction approach using Information sets to eliminate information loss during binarization, reducing feature dimensionality and enhancing face recognition performance.

Findings

Features outperform eigenface, fisherface, and LBP methods.

Reduced feature set due to non-overlapping windows.

Effective in discriminating facial textures.

Abstract

As the features from the traditional Local Binary Patterns (LBP) and Local Directional Patterns (LDP) are found to be ineffective for face recognition, we have proposed a new approach derived on the basis of Information sets whereby the loss of information that occurs during the binarization is eliminated. The information sets expand the scope of fuzzy sets by connecting the attribute and the corresponding membership function value as a product. Since face is having smooth texture in a limited area, the extracted features must be highly discernible. To limit the number of features, we consider only the non overlapping windows. By the application of the information set theory we can reduce the number of feature of an image. The derived features are shown to work fairly well over eigenface, fisherface and LBP methods.