An Efficient Color Face Verification Based on 2-Directional 2-Dimensional Feature Extraction

TL;DR

This paper introduces a novel face verification framework that combines 2-directional 2D feature extraction with skin color information, resulting in high accuracy and speed on the XM2VTS database.

Contribution

It proposes a new uniform framework integrating 2D feature extraction with skin color features for improved face verification performance.

Findings

High verification accuracy achieved

Fast verification speed demonstrated

Effective use of skin color information

Abstract

A novel and uniform framework for face verification is presented in this paper. First of all, a 2-directional 2-dimensional feature extraction method is adopted to extract client-specific template - 2D discrimant projection matrix. Then the face skin color information is utilized as an additive feature to enhance decision making strategy that makes use of not only 2D grey feature but also 2D skin color feature. A fusion decision of both is applied to experiment the performance on the XM2VTS database according to Lausanne protocol. Experimental results show that the framework achieves high verification accuracy and verification speed.