See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258790239

3D Face Recognition Using Radon Transform and Symbolic LDA

Article in International Journal of Computer Applications · April 2013

DOI: 10.5120/11380-6666

CITATIONS READS

5 96

2 authors:

Prakash Hiremath Manjunatha Hiremath

Gulbarga University Christ University, Bangalore
203 PUBLICATIONS 3,582 CITATIONS 14 PUBLICATIONS 89 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Prakash Hiremath on 21 February 2017.

The user has requested enhancement of the downloaded file.

International Journal of Electronics and Computer Science Engineering 2342

Available Online at www.ijecse.org ISSN- 2277-1956

3D Face Recognition Using Radon Transform and

Symbolic PCA
P. S. Hiremath 1, Manjunath Hiremath 2
12
Departmentof Computer Science
Gulbarga University, Gulbarga-585106
Karnataka, India
1
[email protected],2 [email protected]

Abstract- Three Dimensional (3D) human face recognition is emerging as a significant biometric technology. Research
interest into 3D face recognition has increased during recent years due to availability of improved 3D acquisition devices
and processing algorithms. A 3D face image is represented by 3D meshes or range images which contain depth
information. Range images have several advantages over 2D intensity images and 3D meshes. Range images are robust to
the change of color and illumination, which are the causes for limited success in face recognition using 2D intensity
images. In the literature, there are several methods for face recognition using range images, which are focused on the data
acquisition and preprocessing stage only. In this paper, a new 3D face recognition technique based on symbolic Principal
Component Analysis approach is presented. The proposed method transforms the 3D range face images using radon
transform and then obtain symbolic objects, (i.e. interval valued objects) termed as symbolic 3D range faces. The PCA is
employed to symbolic 3D range face image dataset to obtain symbolic eigen faces which are used for face recognition. The
proposed symbolic PCA method has been successfully tested for 3D face recognition using Texas 3D Face Database. The
experimental results show that the proposed algorithm performs satisfactorily with an average accuracy of 97% as compared
to conventional PCA method and is efficient in terms of accuracy and detection time.

Keywords –3D face recognition, range images, radon transform, principal component analysis, Symbolic PCA.

I. INTRODUCTION
Face Recognition and verification have been at the top of the research agenda of the computer vision community for
more than a decade. The scientific interest in this research topic has been motivated by several factors. The main
attractor is the inherent challenge that the problem of face image processing, face detection and recognition.
However, the impetus for better understanding of the issues raised by automatic face recognition is also fuelled by
the immense commercial significance that robust and reliable face recognition technology would entail. Its
applications are envisaged in physical and logical access control, security, man-machine interfaces and low bitrate
communication.

To date, most of the research efforts, as well as commercial developments, have focused on two dimensional (2D)
approaches. This focus on monocular imaging has partly been motivated by costs but to a certain extent also by the
need to retrieve faces from existing 2D image and video database. With recent advances in image capture techniques
and devices, various types of face-image data have been utilized and various algorithms have been developed for
each type of image data. Among various types of face images, a 2D intensity image has been the most popular and
common image data used for face recognition because it is easy to acquire and utilize. It, however, has the intrinsic
problem that it is vulnerable to the change of illumination. Sometimes the change of illumination gives more
difference than the change of people, which severely degrades the recognition performance. Therefore, illumination-
controlled images are required to avoid such an undesirable situation when 2D intensity images are used. To
overcome the limitation of 2D intensity images, Three Dimensional (3D) images are being used, such as 3D meshes
and range images. A 3D mesh image is the best 2D representation of 3D objects. It contains 3D structural
information of the surface as well as the intensity information of each point. By utilizing the 3D structural
information, the problem of vulnerability to the change of illumination can be solved. A 3D mesh image is suitable
image data for face recognition, but it is complex and difficult to handle.

ISSN 2277-1956/V1N4-2342-2349
 2343
3D Face Recognition Using Radon Transform and Symbolic PCA

A range image is simply an image with depth information. In other words, a range image is an array of numbers
where the numbers quantify the distances from the focal plane of the sensor to the surfaces of objects within the
field of view along rays emanating from a regularly spaced grid. Range images have some advantages over 2D
intensity images and 3D mesh images. First, range images are robust to the change of illumination and color because
the value on each point represents the depth value which does not depend on illumination or color. Also, range
images are simple representations of 3D information. The 3D information in 3D mesh images is useful in face
recognition, but it is difficult to handle. Different from 3D mesh images, it is easy to utilize the 3D information of
range images because the 3D information of each point is explicit on a regularly spaced grid. Due to these
advantages, range images are very promising in face recognition.

The majority of the 3D face recognition studies have focused on developing holistic statistical techniques based on
the appearance of face range images or on techniques that employ 3D surface matching. A survey of literature on the
research work focusing on various potential problems and challenges in the 3D face recognition can be found in the
survey[1-5]. Gupta et al.[6] presented a novel anthropometric 3D face recognition algorithm. This approach
employs 3D Euclidean and Geodesic distances between 10 automatically located anthropometric facial fiducial
points and a linear discriminant classifier with 96.8% recognition rate. Lu et al.[7] constructed many 3D models as
registered templates, then they matched 2.5D images (original 3D data) to these models using iterative closest point
(ICP). Chang et al. [8] describe a “multi-region” approach to 3D face recognition. It is a type of classifier ensemble
approach in which multiple overlapping sub regions around the nose are independently matched using ICP and the
results of the 3D matching are fused. Hiremath and Manjunath [15] have discussed the 3D face recognition by using
Radon Transform and PCA with recognition accuracy of 95.30%. Hiremath and Prabhakar [16] have employed
symbolic data analysis approach for 2D face recognition based on eigenfaces.

In the proposed method, the objective is to apply Symbolic Principal Component Analysis method (Symbolic PCA)
for 3D face recognition based on Radon transformation. The experimentation is done using the Texas 3D face
database [9].
II.MATERIALS AND METHODS

For experimentation, we consider the Texas 3D Face Database [9]. The 3D models in the Texas 3D Face recognition
Database were acquired using an MU-2 stereo imaging system. All subjects were requested to stand at a known
distance from the camera system. The stereo system was calibrated against a target image containing a known
pattern of dots on a white background. The database contains 1149 3D models of 118 adult human subjects. The
number of images of each subject varies from 2 per subject to 89 per subject. The subjects age ranges from
minimum 22 to maximum 77 years. The database includes images of both males and females from the major ethnic
groups of Caucasians, Africans, Asians, East-Indians, and Hispanics. The facial expressions present are smiling or
talking faces with open/closed mouths and/or closed eyes. The neutral faces are emotionless.

III.PROPOSED METHOD
A. Radon Transform
The Radon Transform (RT) is a fundamental tool in many areas. The 3D radon Transform is defined using 1D
projections of a 3D object f(x,y,z) where these projections are obtained by integrating f(x,y,z) on a plane, whose
r
orientation can be described by a unit vector α . Geometrically, the continous 3D Radon transform maps a function
r
 3 into the set of its plane integrals in  3 . Given a 3D function f ( x )  f ( x, y, z ) and a plane whose
r
representation is given using the normal α and the distance s of the plane from the origin, the 3D continous Radon
Transform of ffor this plane is defined by
∞ ∞ ∞
r r r r
ℜf (a , s ) = ∫∫∫ f ( x )δ ( x T α − s )dx
−∞ −∞ −∞

ISSN 2277-1956/V1N4-2342-2349
 IJECSE,Volume1,Number 4 2344
P. S. Hiremath and Manjunath Hiremath

∞ ∞ ∞
= ∫∫∫ f ( x, y, z )δ ( x sin θ cos φ + y sin θ sin φ + z cos θ − s )dxdydz
−∞ −∞ −∞ .
r r
where x = [ x, y , z ] , α =[ sin θ cos φ ,sin θ sin φ , cos θ ]T, and δ is Dirac’s delta function defined by
T

∞
r
δ ( x) = 0, x ≠ 0, ∫ δ ( x)dx = 1 . The Radon transform maps the spatial domain (x,y,z) to the domain ( α , s ) are
−∞
not the polar coordinates of (x,y,z). The 3D continous Radon Transform satisfies the 3D Fourier slice theorem[10].

B. Symbolic Faces

Consider the 3D range face images Γ1 , Γ 2 ,...Γ n , each of size N × M , from a 3D face image database. Let
Ω = {Γ1 , Γ 2 ,...Γ n } be the collection of n face images of the database, which are first order objects. Each object
Γ ∈ Ω, l = 1, 2,..., n , is described by a feature vector (Y% ,..., Y% ) , of length p=NM, where each component Y% ,
l 1 p j

j = 1, 2,..., p , is a single valued variable representing the range values of the 3D face image Γ l . An image set is
a collection of 3D range face images of m different subjects; each subject has same number of images but with
different facial expressions and illuminations. There are m number of second order objects (face classes) denoted by
c1 , c2 ,..., cm ,each consisting of different individual images Γ l ∈ Ω . We denote the set E = {c1 , c2 ,..., cm } and
ci ⊆ Ω , i = 1, 2,..., m . The feature vector of each face class ci ∈ E is described by a vector of p interval
variables Y1 , Y2 ,..., Yp , and is of length p = NM . The interval variable Y j of face class ci is declared by

Y j (ci ) = [ xij , xij ] , where xij and xij are minimum and maximum intensity values, respectively, among j th range
values of all the images of face class ci . This interval incorporates information of the variability of j th feature
face class. We denote X (ci ) = (Yl (ci ),..., (Y p (ci )) . The vector
th
inside the i X (ci ) of symbolic variables is
recorded for each ci ∈ E , and can be described by a symbolic data vector which is called as symbolic face :
X (ci ) = ( ai1 , ai 2 ,..., aip ) , where aij = Y j (ci ) , j = 1, 2,..., p .[13,14] We represent the m symbolic faces by a
m × p matrix :
 a11 ... a1 p 
 
X = . . .  = ( aij ) m× p
 am1 ... amp 

C. Symbolic PCA

The symbolic PCA takes as input the matrix X containing m symbolic faces pertaining to the given set Ω of
images. We use centers method, which essentially applies the conventional PCA method to the centers xij ∈ ℜ of
c

the interval [ xij , xij ] , that is,

xij + xij
xijc =
2
where j = 1,..., p and i = 1,..., m . The m × p data matrix X c containing the centers xijc of the intervals α ij for
m symbolic faces given by :

ISSN 2277-1956/V1N4-2342-2349
 2345
3D Face Recognition Using Radon Transform and Symbolic PCA

 x11c ... x1cp 

 
Xc = . . . 
 xc ... c 
xmp
 m1 
1 1 c
The mean vector ψ of X c is defined by ψ = [ψ j ] , where ψ j = ∑ xij , j = 1, 2,..., p . Each row vector of
m m
X c differs from the mean vector ψ by the vector Φi = ( xic1 , xic2 ,..., xipc ) −ψ . We define the matrix Φ as
Φ = [Φ1' , Φ '2 ,..., Φ 'm ] . The covariance matrix C is obtained as C = Φ ' Φ . Then, we calculate the eigenvalues
λ1 ≥ λ2 ≥,..., λm ≥ 0 and the corresponding orthonormalized eigenvectors y1 , y2 ,..., ym ∈ ℜ of the covariance
m

matrix C. The eigenvectors of the symbolic PCA can be obtained as Vm = ΦYm , where Ym = ( y1 ,..., ym ) is the
m × m matrix with columns y1 , y2 ,..., ym and Vm is the p × m matrix with corresponding eigenvectors
v1 , v2 ,..., vm , as its columns. The subspace is extracted from the p × m dimensional space by selecting S number
of eigenvectors, which contain maximum variance and are denoted by v1 , v2 ,..., vs , corresponding to eigenvalues
λ1 ≥ λ2 ≥,..., λs . The weights Wik for i th symbolic face, i = 1, 2,..., m, are computed as
Wik = V ( x −ψ ) .
k
T c
ij

where k = 1, 2,..., S . The weights of i

th
symbolic face from the feature vector (Wi1 ,Wi 2 ,..., Wis ) of the i th
symbolic face[11-13]. The weights of test image I test are computed by projecting the test image into face subspace
as :
Wtestk = vkT ( I test − Ψ ) .
D. Proposed Methodology

The Figure 1 shows the overview of proposed framework. The algorithms of the training phase and the testing phase
of the proposed method are given below:

Training Phase

Training Radon Symbolic Feature

Set Transform PCA Library

Preprocessing PCA Feature Library

Input 3D Symbolic Recognized

Radon 3D range
Range PCA Classifier
Transform image
Image Features

Testing Phase

Figure1. Overview of proposed framework

ISSN 2277-1956/V1N4-2342-2349
 IJECSE,Volume1,Number 4 2346
P. S. Hiremath and Manjunath Hiremath

Algorithm 1: Training Phase

1. Input the range image I1 from the training set containing M images.
2. Apply Radon transform, from 0o to 180o orientations (in steps of h), to the input range image I1 yielding a
binary image I2.
3. Superpose the binary image I2 obtained in the Step 2 on the input range image I1 to obtain the cropped
facial range image I3.
4. Repeat the Steps 1 to 3 for all the M facial range images in the training set.
5. Apply Symbolic PCA to the set of cropped facial range images obtained in the Step 4 and obtain M Eigen
faces.
6. Compute the weights w1 , w2 , ..., w p for each training face image, where p < M is the dimension of Eigen
subspace on which the training face image is projected.
7. Store the weights w1 , w2 ,..., w p for each training image as its facial features in the Symbolic PCA feature
library of the face database.

Algorithm 2: Testing Phase

1. Input the test range image Z1.
2. Apply Radon transform, from 0o to 180o orientations (in steps of h), to the input range image Z1 yielding a
binary image Z2.
3. Superimpose the binary image Z2 on Z1 to obtain the cropped facial image Z3.
Compute the Symbolic weights wi , i = 1, 2,..., p , for the test image Z1 by projecting the test image on
test
4.
the Symbolic PCA feature subspace of dimension p.
test
5. Compute the Euclidian distance D between the feature vector wi and the feature vectors stored in the
Symbolic PCA feature library.
6. The face image in the face database, corresponding to the minimum distance D computed in the Step 5, is
the recognized face.
7. Output the texture face image corresponding to the recognized facial range image of the Step 6.
IV.RESULTS AND DISCUSSION
For experimentation, we consider the Texas 3D face database [9]. The proposed method is implemented using Intel
Core 2 Quad processor @ 2.66 Ghz machine and MATLAB 7.9. In the training phase, 10 frontal face images with
different expression of each 50 subjects are selected as training data set. In the testing phase, randomly chosen 200
face images of the Texas 3D face database with variations in facial expressions are used. The sample training images
which are used for our experimentation are shown in the Figure 2, and their corresponding texture images are shown
in the Figure 3.

(a) (b) (c) (d) (e)

Figure2. Sample range images of the training set.

ISSN 2277-1956/V1N4-2342-2349
 2347
3D Face Recognition Using Radon Transform and Symbolic PCA

(a) (b) (c) (d) (e)

Figure3. The facial texture images corresponding to the training range images of the Figure2

The recognition rates obtained by the proposed (RT+ Symbolic PCA) approach is compared with PCA (alone),
Symbolic PCA and RT+PCA methods [15] is presented in the Table 1. The graph of recognition rates versus the
number of eigenfaces is shown in the Fig.4 for the proposed method (RT + Symbolic PCA) with other PCA,
Symbolic PCA and RT + PCA methods. It is observed that the recognition rate improves as the number of eigen
faces is increased. It is 97% for 40 eigen faces in case of proposed method. Further, the proposed method based on
RT and Symbolic PCA outperforms the PCA method.

Table -1. Performance comparison of PCA, Symbolic PCA, RT+PCA and RT+ Symbolic PCA

No. of Eigen PCA Symbolic RT + PCA RT +

Components [15] PCA [15] Symbolic
PCA
5 58.50 59 60.10 62
10 76.10 77 77.50 78
15 81.50 82 84.36 85.5
20 87.10 88 90.19 91
25 87.61 88.5 94.10 95
30 88.50 89.5 94.16 96
35 89.11 90 95.20 96.5
40 89.47 91 95.30 97

ISSN 2277-1956/V1N4-2342-2349
 IJECSE,Volume1,Number 4 2348
P. S. Hiremath and Manjunath Hiremath

Figure4. The recognition accuracy (%) versus the number of eigenfaces

of the proposed method and the other methods.

V.CONCLUSION
In this paper, we have proposed a novel hybrid method for Three Dimensional (3D) face recognition using Radon
transform with Symbolic PCA based features on 3D range face images. In this method, the Symbolic PCA based
feature computation takes into account face image variations to a larger extent and has advantage of dimensionality
reduction. The experimental results yield 97% recognition performance with low complexity and a small number of
features, which compares well with other state-of-the-art methods. The experimental results demonstrate the efficacy
and the robustness of the method to illumination and pose variations. The recognition accuracy can be further
improved by considering a larger training set and a better classifier.

Acknowledgement

The authors are grateful to the referees for their helpful comments and suggestions. Also, the authors are indebted to
the University Grants Commission, New Delhi, for the financial support for this research work under UGC-MRP
F.No.39-124/2010 (SR).

VI.REFERENCE

[1] R. Chellappa, C. Wilson, and S. Sirohey, “ Human and machine recognition of faces: A survey”, Proc. Of the IEEE, vol. 83, no. 5,
pp.704-740(May 1995).
[2] W. Zhao, R. Chellappa, P. J. Phillips, A. Rosenfeld, “Face Recognition: A Literature Survey”, ACM Computing Surveys, Vol. 35, No.
4, pp. 399–458(December 2003).
[3] Patil A.M., Kolhe S.R. and Patil P.M., “2D Face Recognition Techniques: A Survey”, International Journal of Machine Intelligence,
ISSN: 0975–2927, Volume 2, Issue 1, pp.74-83(2010).
[4] Andrea F. Abate, Michele Nappi, Daniel Riccio, Gabriele Sabatino, “2D and 3D face recognition: A survey”, Pattern Recognition
Letters 28, pp.1885–1906, (2007).
[5] Kevin W. Bowyer, Kyong Chang, Patrick Flynn, “A survey of approaches and challenges in 3D and multi-modal 3D + 2D face
recognition”, Computer Vision and Image Understanding 101, 1–15 (2006).
[6] Shalini Gupta, Mia K. Markey, Alan C. Bovik, “Anthropometric 3D Face Recognition”, Int. J. Computer Vis., Springer
Science+Business Media, LLC (2010).

ISSN 2277-1956/V1N4-2342-2349
 2349
3D Face Recognition Using Radon Transform and Symbolic PCA

[7] XiaoguangLu , Dirk Colbry , Anil K. Jain, “Matching 2.5D Scans for Face Recognition”, Int. Conf. Pattern Recog. (ICPR 2004),
pp.362-366.
[8] Kyong I. Chang , Bowyer, K.W., Flynn P. J., “Adaptive Rigid Multi-region Selection for Handling Expression Variation in 3D Face
Recognition”, Computer Vision and Pattern Recognition - Workshops, 2005.
[9] S. Gupta, K. R. Castleman, M. K. Markey, A. C. Bovik, “Texas 3D Face Recognition Database”, IEEE Southwest Symposium on
Image Analysis and Interpretation, May 2010, p 97-100, Austin, TX. URL: http://live.ece.utexas.edu/research/texas3dfr/index.htm.
[10] Amir Averbuch and YoelShkolnisky, “3D Fourier based discrete Radon transform”, Appl. Comput. Harmon. Anal. 15 , Elsevier Inc.
(2003), pp. 33–69.
[11] M. Turk, A. Pentland, “Eigenfaces for Recognition”, Journal of Cognitive Neurosicence, Vol. 3, No. 1, 1991, pp. 71-86
[12] H. Moon, P.J. Phillips, “Computational and Performance aspects of PCA-based Face Recognition Algorithms”, Perception, Vol. 30,
pp. 303-321(2001).
[13] Bock, H. H. Diday E. (Eds) : “Analysis of Symbolic Data”, Springer Verlag (2000).
[14] Carlo N. Lauro and Francesco Palumbo, “Principal Component Analysis of Interval Data: a Symbolic Data Analysis Approach”,
Computational Statistics, Vol.15 n.1 (2000) pp.73-87.
[15] P. S. Hiremath and ManjunathHiremath, “3D Face Recognition Using Radon Transform and PCA”, International Journal of Graphics
& Image Processing,Vol. 2,No. 2,(May 2012), pp. 123-128.
[16] P. S. Hiremath and C. J. Prabhakar, “Face Recognition Technique Using Symbolic PCA Method”, PReMI, LNCS 3776,Springer-
Verlang Berlin Hiedelberg, pp.266-271 (2005).

ISSN 2277-1956/V1N4-2342-2349

View publication stats