05-05-2011, 10:02 AM
Presented By
Ramakrishna Paruchuri
Abstract
This article is the brief summary based on the reference article “advances in the face
detection and recognition technologies” by Atsushi SATO, Hitoshi IMMOKA, SUZUKI
and HOSAI. This report briefly describes the advances in the authors face detection and
recognition technologies. For the face detection they proposed a combined scheme for
both face and eye based on the Generalized Learning Vector Quantization method. For
the face recognition a perturbation method has been improved to reduce the adverse
effects of both illumination and pose changes. In this report the first section gives an
introduction to the biometric authentication methods, the second section explains
different methods that are proposed by the authors for the face detection and recognition
and in the third section the experimental results of the proposed method has been
explained.
1. Introduction
In the recent years there have been great expectations in the biometric
authentication. Biometric authentication is the automatic identification of the individual
based on the physiological or behavioral characteristics such as finger print, iris, vein,
face and voice. This kind of authentication is commonly used in safeguarding country
borders, in control access to facilities and to enhance the computer security.
In the biometric authentication face recognition has special characteristics.
Following are the some of the advantages of the face reorganization technology.
Advantages:
- They do not need any physical contact
- They can be captured from a distance.
-some biometric authentication like voice detection depends mostly on the surrounding
environment. But for the face recognition they do not affect.
The same authors have developed lot of techniques for the face detection earlier
[1].this particular paper that I have discussed mentions some of the advances in the
authors face recognition technologies. For the face detection a combined scheme for the
both face and eye detection has been developed using generalized learning vector
quantization (GLVQ), for the face recognition a perturbation method has been improved
to reduce the adverse effects of both illumination and pose changes.
2. Face detection and alignment:
Face detection has mainly two important factors one is to determine all the
facial backgrounds on different backgrounds and the other is to determine the alignment
of each face such as position, rotation and size to obtain the better view. If the face is
rigid it has six parameters of freedom, three coordinate and three rotation freedoms. In
the front view the degrees of the freedom reduces to four as we have the compromise on
two rotational angles. so most detection algorithms uses detection of one in-plane
characteristic like eyes and one out of plane characteristic like detection on of the mouth
region in the face.
Lot of work has been done in this field earlier like detecting skin
regions for the face detection but the disadvantage with this method is its sensitivity for
the background light. The development of view based methods over came this problem.
However the problem with view based methods are they consume lot of time and the
facial alignment is not so clear in theses methods as they ignore high frequency
components.
Download full report
http://isnm.de/~rparuchu/Studies/Second%...tation.pdf
