[ Face analysis] A Compact Local Binary Pattern

A Compact Local Binary Pattern using Maximization of Mutual Information (MMI) for Face Analysis

Introduction

Although many variants of local binary patterns (LBP) are widely used for face analysis due to their satisfactory classification performance, they have not yet been proven compact. We propose an effective code selection method that obtains a compact LBP (CLBP) using the maximization of mutual information (MMI) between features and class labels. The derived CLBP is effective because it provides better classification performance with a smaller number of codes. We demonstrate the effectiveness of the proposed CLBP through several experiments of face recognition and facial expression recognition. Our experimental results show that the CLBP outperforms other LBP variants such as LBP, ULBP, and MCT in terms of the smaller number of codes and better recognition performance.

Keywords:

Local Binary Pattern, Feature Selection, Compact LBP, Maximization of Mutual Information

To get further details, check out our paper:
A compact local binary pattern using maximization of mutual information for face analysis, Pattern Recognition, 2010