International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13460-13465
© Research India Publications. http://www.ripublication.com
13460
A Galois Field based Texture Representation for Face Recognition
Shivashankar S.
1
, Medha Kudari
2
, Prakash S. Hiremath
3
1
Department of Computer Science, Karnatak University, Dharwad, Karnataka 580003, India.
2
Department of Computer Science, Karnatak University, Dharwad, Karnataka 580003, India.
3
Department of Computer Science (MCA), KLE Technological University, BVBCET, Hubballi, Karnataka 580029, India.
Abstract
This paper presents a Galois field based texture representation
for face recognition. A Galois field has been used to represent
texture in images. The facial images are divided into several
local regions. Each of these local regions is represented using
a novel Galois field based method. The bin values of the
normalized cumulative histogram forms the feature vector for
the region. These local features are concatenated to form the
face descriptor. Extensive experiments are performed on
FERET face database and extended Cohn Kanade face
database. The results clearly show that the proposed method
is better and effective as compared to Rotation Invariant Local
Binary Pattern and Log-polar transform.
Keywords: Galois Field, Texture, Face descriptor,
Classification
INTRODUCTION
One of the properties that distinguish objects and images is
texture, the other properties being shape and colour. Analysis
of texture is one of the many aspects in image processing and
computer vision. Texture analysis has been applied in the
fields of visual inspection, remote sensing imagery, pattern
recognition and image retrieval. Statistical [1], structural [2],
model based [3] and signal processing models [4] are the
common feature extraction methods. In [5], a survey of
texture descriptors for texture classification is available.
However, most of these approaches are sensitive to the
changes of orientations and scales of the texture pattern. On
the other hand, objects of interest under various orientations,
scales, illumination and occlusion are often encountered in
different applications, such as face recognition and signature
verification. Therefore there arises a requirement to develop
descriptors of texture which are insensitive or invariant to
changes in rotation, scale, illumination and so on. There have
been ongoing research activities to represent invariant texture.
Many researchers have devoted their energies for representing
texture in only rotation invariant environment. Some
researchers have addressed the issue of scale invariance in
images. Limited research is available on the area of rotation
and scale invariant texture representation. The major existing
approaches include psycho-physical transformation, multi-
resolution simultaneous autoregressive (MRSAR) model [6],
log-polar wavelet signatures [7], multichannel Gabor filtering
[8] and the Wold model [9] for invariant texture analysis.
Biometrics is the study of human biological measurements for
identification and verification. Biological measurements like
face, voice and fingerprints qualify as a biometric
characteristic because it has the properties of universality,
distinctiveness, permanence and collectability. In recent times,
more biological measurements have been considered as a
biometric like gait, signatures and iris [10]. Based on
physiological characteristics, biometric traits include face
[12], fingerprints [13], finger geometry, hand geometry [14],
hand veins [15], palm, iris [16], retina [17], ear [18] and
voice. Based on behavioural characteristics, biometric traits
include gait [19], signature [20] and keystroke dynamics
[21][11]. A biometric system is similar to a texture pattern
recognition system where the biometric data is obtained from
an individual, a feature set is extracted from the acquired data
and then this acquired feature set is compared against the
template set in the database. A biometric system may either
verify or identify an individual based on the type of
application the biometric system is being used. A biometric
system can be used for either identification or verification
purposes. In verification application, a user’s identity is
validated by comparing the user’s captured biometric features
against the user’s biometric features stored in the database. In
identification application, a user is recognized by searching
the templates of all the users in the database for a match [10].
Face recognition has emerged as a major ongoing research
area in pattern recognition and computer vision. Face
recognition is considered to be a difficult task than the usual
pattern recognition problems due to the presence of few
training samples (in some cases only one training sample) and
numerous testing samples. The sources of variation in facial
appearance can be categorized into two groups: intrinsic
factors and extrinsic factors. Intrinsic factors are due to purely
physical nature of the face and are independent of the
observer. Extrinsic factors cause the appearance of the face to
alter via the interaction of light with the face and observer like
illumination, pose, scale and imaging parameters [11].
Current face recognition systems perform well under
relatively controlled environments but tend to suffer when
variations in different factors like rotation, scale, pose,
illumination are present [24]. Thus finding good descriptors
for the appearance of local facial regions is an open and
ongoing issue. The researchers on texture analysis have
developed a variety of different descriptors for the appearance
of image patches. Heisele, Ho, Wu and Poggio developed a