Facial Gender Classification Using LUT-Based Sub-images and DIE

Authors:
Jong-Bae Jeon;Sang-Hyeon Jin;Dong-Ju Kim;Kwang-Seok Hong
Affiliations:
School of Information and Communication Engineering, Sungkyunkwan University, Kyungki-do, Korea 440-746;School of Information and Communication Engineering, Sungkyunkwan University, Kyungki-do, Korea 440-746;School of Information and Communication Engineering, Sungkyunkwan University, Kyungki-do, Korea 440-746;School of Information and Communication Engineering, Sungkyunkwan University, Kyungki-do, Korea 440-746
Venue:
ICDHM '09 Proceedings of the 2nd International Conference on Digital Human Modeling: Held as Part of HCI International 2009
Year:
2009

Citing 4
Cited 0

Learning Gender with Support Faces

IEEE Transactions on Pattern Analysis and Machine Intelligence
A Unified Learning Framework for Real Time Face Detection and Classification

FGR '02 Proceedings of the Fifth IEEE International Conference on Automatic Face and Gesture Recognition
Performance Evaluation of Teeth Image Recognition System Based on Difference Image Entropy

ICCIT '08 Proceedings of the 2008 Third International Conference on Convergence and Hybrid Information Technology - Volume 02
LUT-based Adaboost for gender classification

AVBPA'03 Proceedings of the 4th international conference on Audio- and video-based biometric person authentication

Quantified Score

Hi-index	0.01

Visualization

Abstract

This paper presents a gender classification method using LUT-based sub-images and DIE (Difference Image Entropy). The proposed method consists of three major steps; extraction of facial sub-images, construction of a LUT (Look-Up table), and calculation of DIE. Firstly, extraction of sub-images of the face, right eye, and mouth from face images is conducted using Haar-like features and AdaBoost proposed by Viola and Jones. Secondly, sub-images are converted using LUT. LUT-based sub-regions are constructed by calculation of one pixel and near pixels. Finally, sub-images are classified male or female using DIE. The DIE value is computed with histogram levels of a grayscale difference image which has peak positions from -255 to +255, to prevent information sweeping. The performance evaluation is conducted using five standard databases, i.e., PAL, BioID, FERET, PIC, and Caltech facial databases. The experimental results show good performance in comparison with earlier methods.