A Face Detection and Recognition System based on Rectangular Feature Orientation

[attachment=16526]

INTRODUCTION
In 2001 Viola and Jones [1] proposed a machine
learning approach for face detection which is capable of
processing images extremely rapidly and achieving high
detection rates. There are three main features of the Viola
and Jones face detection approach: (a) the use integral
image that allows for very fast feature evaluation, (b) they
construct a classifier by selecting a small number of
important features using AdaBoost [5], and © they
combine successively more complex classifiers in a cascade
structure which dramatically increases the speed of the
detector by focusing attention on promising regions of the
images.


. REVIEW OF PREVIOUS RESULTS
Viola and Jones [1] propose a concept of integral image
and construct a real-time face detection classifier based on
AdaBoost.



. EXPERIMENTAL RESULTS
A. Face data-base
In our experiment we build the face image database by
ourselves and all the face images are obtained by using a
Web-Cam. The images size is 640x480. The database
contains 15 individuals, among them images of 10 peoples
are considered as valid users and the other 5 peoples are
considered as invalid invaders. Total number of images is
500 (300 valid users and 200 invalid invaders). Our training
set contains 10 individuals, a total of 200 face images and
20 for each individual (obtained from 4 slightly different
poses). The testing set contains 5 individuals, a total of 100
face images and 20 for each individual (obtained from 4
slightly different poses).