Face Recognition under Variation of Pose and ... - CiteSeerX

ICGST-GVIP, ISSN 1687-398X, Volume (8), Issue (IV), December 2008

Face Recognition under Variation of Pose and Illumination using Independent Component Analysis Kailash J. Karande.a Sanjay N.Talbar.b Department of Information Technology, Sinhgad Institute of Technology, Lonavala, Pune, Maharashtra State, India. [email protected] b Department of Electronics & Telecommunication, SGGS Institute of Engineering &Technology, Nanded, Maharashtra State, India. [email protected] a

(EBGM). In order to organize the vast field of face recognition, several approaches are conceivable. For instance, algorithms treating the face and its environment as uncontrolled systems could be distinguished from systems that control the lighting or background of the scene, or the orientation of the face. Or systems that use one or more still images for the recognition task could be distinguished from others that base their efforts on video sequences. The comparison of face recognition using PCA and ICA on FERET database with different classifiers [3] [8] are discussed and found that the ICA has better recognition rate as compared with PCA with statistically independent basis images and also with statistically independent coefficients. Their findings are based on the frontal face image datasets are encouraging with few face expressions. Marian S Bartlett used version of ICA [10] derived from the principle of optimal information transfer through sigmoidal neurons on face images from FERET database has proved that ICA representation gave the best performance on the frontal face images. Feature selection in the independent component subspace [4] which gives the benefits for face recognition with changes in illumination and facial expressions. Fusion of ICA features like Spatial, Temporal and Localized features [6] [7] for Face Recognition are considered as optimization method. The illumination has great influence on how a face image looks. Researchers have proved that for a face image, the difference caused by illumination changes has even exceeded the difference caused by identity changes [12]. The big challenge of face recognition lies in dealing with variations of pose, illumination, and expression. Also there is need to address the problem of identity changes using structural components. In this paper we propose the face recognition using ICA with large rotation angles with poses and variation in illumination conditions. We used the database which has the large rotation angles up to 1800 change between the images of person while looking right and or left. We considered the face images having various orientations of

Abstract This paper addresses the problem of face recognition under variation of illumination and poses with large rotation angles using Independent Component Analysis (ICA). Face recognition using ICA, based on information theory concepts, seek a computational model that best describes face, by extracting most relevant information contained in that face. ICA approach used here to extract global features seems to be an adequate method due to its simplicity, speed and learning capability. The preprocessing is done by Principle Component Analysis (PCA) before applying the ICA algorithm for training of images. The independent components obtained by ICA algorithm are used as feature vectors for classification. The Euclidian distance classifier is used for testing of the images. The variation in illumination and facial poses up to 1800 rotation angle is used by the proposed method and result shows that the recognition improved significantly. Keywords: Face recognition, Independent component analysis (ICA), Principle component analysis (PCA), Pose variance, Illumination variance.

1. Introduction Face recognition is one of biometric methods identifying individuals by the features of face. Research in this area has been conducted for more than 30 years; as a result, the current status of face recognition technology is well advanced. Many commercial applications of face recognition are also available such as criminal identification, security system, image and film processing. Humans often use faces to recognize individuals and advancements in computing capability over the past few decades now enable similar recognitions automatically. There are two predominant approaches to the face recognition problem: geometric (feature based) and photometric (view based). As a researcher interest in face recognition continued and many different methods are proposed, like well studied methods using Principle Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Elastic Bunch Graph Matching

1


the face as shown in figure 2 (i.e.: looking front, looking left, looking right, looking up, looking up towards left, looking up towards right, looking down). In this study we considered the samples of individual person which consist of sufficient number of images having expressions, changes in illumination and large rotation angles. For illumination variation the effects of light on face image from left, right, top and bottom sides are considered as shown in figure 1. The paper is organized as follows. In Section 2 we introduce the ICA and need of the preprocessing before applying the ICA algorithm. The modified Fast ICA algorithm is presented in the Section 3. Experimental results are discussed in Section 4 and accordingly the conclusions are drawn in Section 5.

from the previous one, since after estimating A, its inverse gives W. It can be observed that the problem is well defined, that is the model in equation (1) can be estimated if and only if the components si are nongaussian. This is the fundamental requirement that also explains the main difference between ICA and factor analysis, in which the nongaussianity of the data is not taken into account.

2.1 Preprocessing by PCA There are several approaches for the estimation of the independent component analysis (ICA) model. In particular, several algorithms were proposed for the estimation of the basic version of the ICA model, which has a square mixing matrix and no noise. Practically when applying the ICA algorithms to real data, some practical considerations arise and need to be taken into account. To overcome these practical considerations we implemented a preprocessing technique in this algorithm that is dimension reduction by principal component analysis. That may be useful and even necessary before the application of the ICA algorithms in practice. Overall face recognition benefits from feature selection of PCA and ICA combination [4]. A common preprocessing technique for multidimensional data is to reduce its dimension by principal component analysis (PCA) [3]. Basically, the data is projected linearly onto a subspace

2. Introduction to ICA Independent component analysis (ICA) is a method for finding underlying factors or components from multivariate (multidimensional) statistical data. There is need to implement face recognition system using ICA for facial images having face orientations and different illumination conditions, which will give better results as compared with existing systems [10], [11] and [13]. What distinguishes ICA from other methods is that, it looks for component that are both statistically independent and nongaussian [10]. The ICA is similar to blind source separation problem [1] that boils down to finding a linear representation in which the components are statistically independent. In practical situations, we cannot in general find a representation where the components are really independent, but we can at least find components that are as dependent as possible. This leads the definition of ICA. Given a set of observations of random variables (x1(t),x2(t),……,xn(t)), where t is the time or sample index, assume that they are generated as a linear mixture of independent components: ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝

x 1 (t ) ⎞ ⎛ s1 (t ) ⎟ ⎜ x 2 (t ) ⎟ ⎜ s 2 (t ) ⎟ ⎜. . ⎟ = A⎜ . ⎟ ⎜. ⎟ ⎜ . ⎟ ⎜. ⎜ s (t ) x n ( t ) ⎟⎠ ⎝ n

⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠

~ X = En x

Where Ex is the eigenvalue matrix and x is the Eigen vector matrix obtained from covariance matrix. So that the maximum amount of information (in the leastsquares sense) is preserved. Reducing dimension in this way has several benefits. First, let us consider the case where the number of independent components (ICs) n is smaller than the number of mixtures; say m. Performing ICA on the mixtures directly can cause big problems in such a case, since the basic ICA model does not hold anymore. Using PCA we can reduce the dimension of the data to n. After such a reduction, the number of mixtures and ICs are equal, the mixing matrix is square, and the basic ICA model holds. The question is whether PCA is able to find the subspace correctly, so that the n ICs can be estimated from the reduced mixtures. This is not true in general, but in a special case it turns out to be the case. If the data consists of n ICs only, with no noise added, the whole data is contained in an n-dimensional subspace. Using PCA for dimension reduction clearly finds this n-dimensional subspace, since the eigenvalues corresponding to that subspace, and only those eigenvalues, are nonzero. Thus reducing dimension with PCA works correctly. In practice, the data is usually not exactly contained in the subspace, due to noise and other factors, but if the noise level is low, PCA still finds approximately the right subspace. In the general case, some weak ICs may be lost in the dimension reduction process, but PCA may still be a good idea for optimal estimation of the strong ICs.

(1)

where A is some unknown matrix. Independent component analysis now consist of estimating both the matrix A and si(t),when we only observe the xi(t). Note that we assumed here that the number of independent components si is equal to the number of observed variables. This is a simplifying assumption that is not completely necessary. Alternatively, we could define ICA as follows, find a linear transformation matrix ‘W’ as given by following equation ⎛ y1 (t ) ⎜ ⎜ y 2 (t ) ⎜. ⎜ ⎜. ⎜ ⎜. ⎜ y (t ) ⎝ n

⎞ ⎛ x1 (t ) ⎞ ⎟ ⎜ ⎟ ⎟ ⎜ x 2 (t ) ⎟ ⎟ ⎜. ⎟ ⎟ = W ⎜ ⎟ ⎟ ⎜. ⎟ ⎟ ⎜ ⎟ . ⎟ ⎜ ⎟ ⎟ ⎜ x (t ) ⎟ ⎠ ⎝ n ⎠

(3)

(2)

so that the random variables yi; i=1,….,n are as independent as possible. This formulation is not really very different

2


variation with large rotation angles. In second part the face images are used having variation in illumination conditions. We also presented the result of traditional method of face recognition using PCA, these results are used for comparison purpose with face recognition using ICA algorithm. The algorithm is developed in the MATLAB environment. Two standard databases are used for evaluation of these results. The first database is published by IIT Kanpur [5] widely used for research purpose known as Indian face database. In this database images of 60 persons with 10 sample images with different orientations and views are available. The second database known as Asian face image database [2] is from Intelligent Multimedia research laboratories having face images of 56 male persons with 10 samples each; which consist of variation in illumination conditions and different views. The resolution of all images we used in the algorithm is 128 x 128 for computational purpose. Few face images are shown in Figure 1 and Figure 2 from both the databases with various views. In this paper, we presented the results for face recognition using PCA and ICA methods. We obtained the results of PCA method and compared with ICA method as shown in Table 1. The first 25 principle components and independent components are shown in the Figure.3 and Figure.4 respectively

3. ICA Algorithm Here we used the modified FastICA algorithm [1] for computing of independent components. Before computing the independent components we use to calculate the principle components and then whitened those components to reduce the size of matrix. In the previous section we have seen the need to use PCA before applying ICA. After finding principle components we whiten the eigenvector matrix to reduce the size of matrix and make it square. To estimate several independent components, we need to run the ICA algorithm several times with weight vectors w1, ...,wn. To prevent different vectors from converging to the same maxima we must decorrelates the outputs wT1x, ...,wTn x after every iteration. A simple way of achieving decorrelation is a deflation scheme based on a Gram-Schmidt-like decorrelation [14]. It is a sequential orthogonalization procedure. This means that we estimate the independent components one by one. When we have estimated p independent components, or p vectors w1, ...,wp, we run the one-unit fixed-point algorithm for wp+1, and after every iteration step subtract from wp+1 the projections wT p+1 wj wj , j = 1, ..., p of the previously estimated p vectors, and then renormalize wp+1: p

W p +1 = W p +1 − ∑ W T p +1W jW j

(4)

W p +1 = W p +1 / W T p +1W p +1

(5)

j =1

In certain applications, however, it may be desired to use a symmetric decorrelation, in which no vectors are privileged over others. This can be accomplished, e.g., by the classical method involving matrix square roots,

W = (WW T ) −1 / 2 W

(6) where W is the matrix (w1, ...,wn)T of the vectors, and the inverse square root (WWT )−1/2 is obtained from the eigenvalues decomposition of WWT = FDFT as (WWT )−1/2 = FD−1/2FT . A simpler alternative is the following iterative algorithm,

W = W / || WW T ||

(7)

Figure 1. Various view of face images with different illumination conditions in Asian face database

As we are using the ICA algorithm the training time required is less as compared to other methods. For the algorithm developed in this paper required around few seconds time as training time. For testing we used the Euclidean distance classifier, for calculating the minimum distance between the test image and image to be recognized from the database. Euclidean distance to determine closeness reduces the problem to computing the distance measures: j =1,2,……..,W. (8) D j ( x) =|| x − m j || If the distance is small, we say the images are similar and we can decide which the most similar image in the database is.

4. Experimental results The experimental results presented in this section are divided into two parts. Both parts evaluate the face representations using PCA [9] as well as ICA methods. In first part the face images are used having the pose

Figure 2. Various view of face images with different face orientations in Indian face database

3


Figure 7

Figure 3 First 25 Eigenimages. (Principle Components)

Figure 8

The recognition of individual images using the algorithm is shown in the Figure 5 to Figure 11 respectively. Figure 5, 6, 7 and Figure 8 gives the input (Image to be recognized) and output images (Recognized image) of the algorithm under face orientations. In Figure 5 and Figure 6 the rotation angle is maximum (approximately 1800), and with this large rotation angle images are recognized. With this much large facial pose variation the results are encouraging. When we choose the first 50 or 100 independent components the achieved result is 100 %, means there is no any false alarm with respect to recognizing the input face image. Second part consist the face images from Asian database with different illumination conditions. The results are taken with different datasets having number of principle and independent components varying from 50 to 200. Figure 9, 10 and Figure 11 gives the result of the algorithm under changing illumination conditions.

Figure.4 First 25 Independent Components

Face images used in the first part are from Indian face database with large rotation angles up to 1800. On this set of images we applied both the algorithms PCA and ICA. We used the different sets of principle and independent components varying from 50 to 200. If we observe in Table 1 the recognition rate achieved by ICA method is more as compared with PCA.

Figure 5

Figure 9

Figure 6

Figure 10

4


of first part. This implies that ICA on face images is less sensitive to face orientations as compared to illuminations as shown in results.

6. References [1] Aapo Hyvärinen and Erkki Oja “Independent Component Analysis: Algorithms and Applications” Neural Networks Research Centre Helsinki University of Technology P.O. Box 5400, FIN-02015 HUT, Finland, Neural Networks, 13(4-5):411-430, 2000 [2] Asian face image database from Intelligent MultimediaLaboratory www.nova.postech.ac.kr / special / imdb /paper_pdf.pdf. [3] Bruce A. Draper, Kyungim Baek, Marian Stewart Bartlett, “Recognizing faces with PCA and ICA”, Computer Vision and Image Understanding 91 (2003) 115-137. [4] H.K.Ekenel, B.Sankur, “Feature Selection in the Independent Component Subspace for Face Recognition”, Pattern Recognition Letters 25 (2004) 1377-1388. [5] Indian face database www.cs.umass.edu / ~vidit / face database. [6] Jiajin Lei, Chao Lu, “Face recognition by Spatiotemporal ICA using Facial Database Collected by AcSys FRS Discover System”, Proceedings of the Seventh ACIS International Conference on Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed Computing (SNPD’06). [7] Jiajin Lei, Chao Lu, “Fusion of ICA Spatial, Temporal and Localized Features for Face Recognition”, Proceedings of the Seventh ACIS International Conference on Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed Computing (SNPD’06). [8] Jian Yang, David Zhang, Jing-yu Yang, “Is ICA Significantly Better than PCA for Face Recognition?” Proceedings of the Tenth IEEE International Conference on Computer Vision (ICCV’05) 15505499/05. [9] M.A.Turk and A.P. Pentaland, “Face Recognition Using Eigenfaces”, IEEE conf. on Computer Vision and Pattern Recognition, pp. 586-591, 1991. [10] Marian Stewart Bartlett, Javier R. Movellan, Terrence J. Sejonowski, “Face Recognition by Independent Component Analysis”, IEEE Transactions on Neural Networks, vol-13, No-6, November 2002, PP 1450-1464. [11] Pong C.Yuen, J.H.Lai, “Face representation using independent component analysis”, Pattern Recognition 35 (2002) 1247-1257. [12] R. M. Bolle, J. H. Connell, and N. K. Ratha, “Biometric perils and patches,” Pattern Recognition vol. 35, pp. 2727 – 2738, 2002. [13] Tae-Kyun Kim, Hyunwoo Kim, Wonjum Hwang, Josef Kittler, “Independent component analysis in a local facial residue space for face recognition”, Pattern Recognition 37 (2004) 1873-1885. [14] Aapo Hyvarinen, Juha Karhunen, Erkki Oja “Independent Component Analysis” Book by A Wiley Interscience Publication, John Wiley & sons, inc, New York.

Figure 11

The light condition on the face images are from left and right sides as well as from top and down directions. Under different illumination conditions the results are encouraging as shown in table 1.

Part I-Pose variation with large rotation angles Part IIVariation of illumination conditions.

Recognition rate using ICA (%)

Recognition rate using PCA (%)

No of Components used (PCs & ICs)

Database used

Conditions of images used for algorithm

Table 1: Results of the Face Recognition Algorithm.

Indian face database.

50 100 150 200

100 100 95.33 90.5

100 100 96 92.5

Asian face database.

50 100 150 200

100 97 92 90.5

100 94 89.33 86.5

In this study we have explored feature selection techniques on ICA and PCA bases for face recognition. Feature selection techniques are warranted especially for ICA features, since these are devoid of any importance ranking based on energy content as the PCA components. The study is carried out on the face database that contains both facial expression with pose variation and illumination variations. We implemented the algorithms for face recognition using PCA and ICA on the different conditions of face images with different set of images. These set of images we used are by selecting the first 50, 100, 150 or 200 principle and independent components. If we compare the results of existing face recognition methods using ICA as discussed in [3], [4], [10], [11] and [13] our algorithm gives better results under various conditions of face images.

5. Conclusion In this paper an independent component analysis of face images has been discussed and used for face recognition. Two standard databases are used, which contains the face images with different orientations, expressions and change in illumination. The performance of the algorithm suggested produces very good recognition rate varying from 86% to 100%. Applying ICA on face images for recognition; selection of ICs are required in order to give better performance. We adopted modified FastICA algorithm for computing the independent components. If we observe the results of face recognition using ICA under pose variation with rotation angles up to 1800, the results are very good and encouraging. The results of the second part with ICA method are also close to the results 5


Biographies:

Kailash J. Karande born in Maharashtra, India, in 1971. He received his B.E degree in Electronics & Telecommunication from Mumbai University in 1994 and M. Tech degree in Electronics & Telecommunication from Dr BAT University, India in 2004. Currently working as Assistant Professor at Sinhgad Institute of Technology, Lonavala Pune, India. His research interests are in Embedded Systems, Image Processing and Bio-metrics

Sanjay N. Talbar received his B.E and M.E degrees from SGGS Institute of Technology, Nanded, India in 1985 and 1990 respectively. He obtained his PhD (Highest degree) from SRTM University, India in 2000. He received the Young Scientist Award by URSI, Italy in 2003. He has Collaborative research programme at Cardiff University Wales, UK. Currently he is working as Professor in Electronics & Telecommunication Department of SGGS Institute of Technology Nanded, India. He is a member of many prestigious committees in academic field of India. His research interests are Signal & Image processing and Embedded Systems.

6