An Enhanced Histogram Matching Approach Using the Retinal Filter s Compression Function for Illumination Normalization in Face Recognition

An Enhanced Histogram Matching Approach Using the Retinal Filter s Compression Function for Illumination Normalization in Face Recognition Ahmed Salah-ELDin, Khaled Nagaty, and Taha ELArif Faculty of Computers and Information Sciences Ain Shams University Cairo, Egypt ahmedss2003@gawab.com, knagaty@asunet.shams.edu.eg, taha_elarif@yahoo.com Abstract. Although many face recognition techniques have been proposed, recent evaluations in FRVT2006 conclude that relaxing the illumination condition has a dramatic effect on their recognition performance. Among many illumination normalization approaches, histogram matching (HM) is considered one of the most common image-processing-based approaches to cope with illumination. This paper introduces a new illumination normalization approach based on enhancing the image resulting from the HM using the gamma correction and the Retinal filter s compression function; we call it GAMMA-HM-COMP approach. Rather than many other approaches, the proposed one proves its flexibility to different face recognition methods and the suitability for real-life systems in which perfect aligning of the face is not a simple task. The efficiency of the proposed approach is empirically demonstrated using both a PCA-based (Eigenface) and a frequency-based (Spectroface) face recognition methods on both aligned and non-aligned versions of Yale B database. It leads to average increasing in recognition rates ranges from 4 ~ 7 % over HM alone. 1 Introduction Face recognition from outdoor imagery remains a research challenge area, [1]. Evaluations of the state-of-the-art techniques and systems in FRVT2006 [2] conclude that relaxing the illumination condition has a dramatic effect on the performance. Moreover, It has been proven both experimentally [3] and theoretically [4] that the variations between the images of the same face due to illumination are almost always larger than image variations due to change in face identity. There has been much work dealing with illumination variations in face recognition. Generally, it can be classified into two categories: model-based and image-processing based approaches. Model-based approaches derive a model of an individual face, which will account for variation in lighting conditions [5,6,7]. Such algorithms require a training set with several lighting conditions for the same subject, which in addition to their highly computational cost make these algorithms unsuitable for realistic applications [8,9]. A. Campilho and M. Kamel (Eds.): ICIAR 2008, LNCS 5112, pp. 873 883, 2008. Springer-Verlag Berlin Heidelberg 2008

874 A. Salah-ELDin, K. Nagaty, and T. ELArif Image-processing-based approaches attempt to normalize the variation in appearance due to illumination by applying image transformations. Recognition is then performed using this canonical form [8,9,10,11]. Compared to the model-based approach, preprocessing has two main advantages: it is completely stand-alone and thus can be used with any classifier. Moreover, it transforms images directly without any training images, assumptions or prior knowledge. Therefore, they are more commonly used in practical systems for their simplicity and efficiency. The Image-processing-based approaches are applied either globally on the whole image or locally over blocks or regions. Local approaches have the disadvantage that the output is not necessarily realistic. Moreover, they are highly dependant on the local distribution of pixels within the image, which make them sensitive to any geometrical effects on the images such as translation, rotation and scaling rather than global approaches which are not affected by such geometrical changes. Histogram matching (HM) is considered one of the most common global imageprocessing-based approaches [11,12,13,14]. Some comparative studies in literature show the superiority of HM among other approaches [8,10]. For example, [8] compares five different illumination normalization approaches, namely histogram equalization HE, histogram matching HM, log transformation LOG, gamma intensity correction GIC and self quotient image SQI over three large-scale face databases which are FERET, CAS-PEAL and CMU-PIE. The results show that HM gives the best results among the four other approaches over FERET and CAS-PEAL, while comes after GIC over CMU-PIE. Moreover, HM has two main advantages: first, it can be applied with any face recognition method, second, it is insensitive to geometrical effects on the image and thus no additional alignment steps are required. Although enhancing the image resulting from the HM can lead to increase the recognition rates over using the HM alone, no attempts have been made to combine the HM with other image enhancement methods for illumination normalization. Also, the compression function of the Retinal filter as an image enhancement method has not been used in the literature. So, it s very interesting to combine the HM with other image enhancement methods as illumination normalization for face recognition. In this paper, we introduce a new illumination normalization approach based on enhancing the image resulting from HM. Four different image enhancement methods are used in this study. They are combined in two different approaches: First, after HM; on the resulting image from HM, Second, before HM; on the reference image before matching the input image on it. In addition, for each approach, we try to further enhancing the results by applying one of these four methods again. Finally, the proposed approach is chosen from these combinations based on the increase in recognition rates against using the HM alone regardless of the following conditions: 1. Face recognition method that the normalization approach is applied with it, 2. Face alignment within the image, 3. Number of training images, and the degree of illumination within these images. This ensures both the flexibility of the proposed approach among different face recognition methods and the ability to apply it on real-life systems in which perfect alignment of faces is difficult to achieve. The verifications of these conditions are described in detail later in this paper. All previous combinations are empirically demonstrated and compared over Yale B database [15] using two holistic face recognition methods, namely, standard Eigenface

An Enhanced Histogram Matching Approach 875 method [16] and Holistic Fourier Invariant Features (Spectroface) [17]. These two methods are chosen to represent the two widely holistic-based categories, PCA-based and frequency-based respectively [18]. The rest of this paper is organized as follows: section 2 contains the description of the four image enhancement methods. In section 3, the different approaches of applying these four methods to enhance the image resulting from HM are introduced. Section 4 is dedicated to describe the Yale B database and the 25 training cases. Experimental results showing the best combinations of HM with different image enhancement methods are presented in section 5. Finally, conclusions and future works are presented in section 6. 2 Image Enhancement Methods The principal objective of image enhancement is to process the original image to be more suitable for the recognition process. Many image enhancement methods are available in the literature. Usually, a certain number of trial and error experiments are required before a particular image enhancement method is selected [19]. In this study, four image enhancement methods are chosen. Three of them are common in literature, namely histogram equalization, log transformation and gamma correction, while the fourth method which called the compression function of the Retinal filter [20] is newly suggested to be used as an image enhancement method in this study. 2.1 Histogram Equalization (HE) It s one of the most common image enhancement methods [19]. It aims to create an image with uniform distribution over the whole brightness scale by using the cumulative density function of the image as a transfer function. Thus, for an image of size M N with G gray levels and cumulative histogram H(g), the transfer function at certain level T(g) is given as follows: H ( g) ( G 1) T(g) = (1) M N 2.2 Log Transformation (LOG) The general form of the log transformation [19] is: s = c log(1 + r) (2) Where r and s are the old and new intensity value, respectively and c is a gray stretch parameter used to linearly scaling the result to be in the range of [0, 255]. The shape of the log curve in Fig.1 shows that this transformation maps a narrow range of dark input gray-levels (shadows) into a wider range of output gray levels. The opposite is true for the higher values of the input gray-levels. 2.3 Gamma Correction (GAMMA) The general form of the gamma correction [19] is:

876 A. Salah-ELDin, K. Nagaty, and T. ELArif s = c r 1/γ (3) Where r and s are the old and new intensity value, respectively, c is a gray stretch parameter used to linearly scaling the result to be in the range of [0, 255] and λ is a positive constant. In our case, λ is chosen to be greater than 1 (empirically, it s chosen to be four) in order to map a narrow range of dark input values (shadows) into a wider range of output values, with the opposite being true for higher input values, see Fig.1. Fig. 1. Transformation functions of LOG and GAMMA (L: number of gray levels) 2.4 Compression Function of the Retinal Filter (COMP) A Retinal filter [21] acts as the human retina by inducing a local smoothing of illumination variations. It has been successfully used as an illumination normalization step in the segmentation of facial features in [20,22]. In this paper, we tried to use it as an illumination normalization step in face recognition. However, our empirical results over both Eigenface and Spectroface methods show that using the Retinal filter as an illumination normalization step leads to decrease the recognition rates of both methods. One possible reason is that the Retinal filter produces a non-realistic image which in turn may not be suitable for holistic-based face recognition methods. Therefore, in this study, we use only the compression function of the Retinal filter as an image enhancement method since it s applied globally and so produces more realistic image (for more details about the Retinal filter, see [21]). Let G be a Gaussian filter of size 15 15 with standard deviation σ = 2 [20]. Let I in be the input image and let I G be the result of G filtering of I in. Image X 0 is defined by: 0.1 + 410 I X 105.5 + I The definition of the compression function C is based on X 0 : C G 0 = (4) G (255 + X ) I I + X 0 in = (5) in 0 3 The Enhanced HM Approaches A total of 40 different enhancement combinations using the HM [19] combined with different enhancement methods are considered and compared in this study in order to

An Enhanced Histogram Matching Approach 877 enhance the results of applying the HM alone. Our reference image used for HM is constructed by calculating the average image of a set of well-lit images one for each subject which gives, by our experiments, better results than using a single well-lit image. Each of the four enhancement methods is applied in three different approaches; 1) After the HM, 2) Before the HM, 3) Further enhancing 1 and 2. 3.1 Enhancement After HM Each of the image enhancement methods, discussed in section 2, is applied on the result of HM in order to enhance it, as shown in Fig.2. This give us four combinations, noted by HM-HE, HM-LOG, HM-GAMMA, HM-COMP, corresponding to applying HE, LOG, GAMMA and COMP, respectively, on the result of HM. Fig.3 shows the effect of these combinations on an illuminated face. Average well-lit Reference Image Input Image Histogram Matching (HM) Matched Image Image Enhancement Output Image Fig. 2. Block diagram of applying the image enhancement method after the HM Illuminated HM HM-HE HM-LOG HM-GAMMA HM-COMP Fig. 3. Effects of applying the image enhancement methods after applying the HM 3.2 Enhancement Before HM Opposite to the approach in 3.1, each of the image enhancement methods is applied on the reference image before matching the input image on it, see Fig.4. This give us another four combinations, noted by HE-HM, LOG-HM, GAMMA-HM, COMP-HM, corresponding to applying HE, LOG, GAMMA and COMP respectively, on the reference image. Fig.5 shows the effect of these combinations on an illuminated face. Average well-lit Reference Image Image Enhancement Enhanced Reference Image Input Image Histogram Matching (HM) Output Fig. 4. Block diagram of applying the image enhancement method before the HM

878 A. Salah-ELDin, K. Nagaty, and T. ELArif Illuminated HM HE-HM LOG-HM GAMMA-HM COMP-HM Fig. 5. Effects of applying the image enhancement methods before applying the HM 3.3 Further Enhancement Here, we further enhancing the result of each combination using each of the four enhancement methods which give us 8 4 = 32 additional combinations. Fig.6 shows block diagrams for such enhancements. The effects of further enhancement on the HM-GAMMA using each of the four enhancement methods are illustrated in Fig.7. Input image Enhancement approach in 3.1/3.2 Enhanced image Image enhancement Output Image Fig. 6. Block diagram showing the further enhancement of combinations in 3.1 and 3.2 Illuminated HM 1st Enhance. Further Enhancement Using HM-GAMMA HE GAMMA LOG COMP Fig. 7. Effects of further enhancement on both HM-GAMMA and GAMMA-HM combinations using each of the four enhancement methods 4 Descriptions of Database and Training Cases We use the Yale B database [15] frontal images only for studying and comparing the 40 enhancement combinations. It consists of 10 subjects each with 65 (64 illuminations + 1 normal) images. Only 46 out of these 65 images are divided into four subsets according to the angle the light source direction makes with the camera s axis (12, 25, 50, and 77 ). We shall use only these four subsets in our experiments. All images are cropped in two different ways to include only the head portion: 1. Automatic cropping using the face detection function in Intel OpenCV library [23] to produce non-aligned version of the database; we call it YALE B-AUTO. 2. Manual cropping using the landmarks coordinates available on the Yale B website [24] to produce an aligned version of it; we call it YALE B-MANU. These two versions, shown in Fig.8, allow us to test the robustness of each enhancement combination against the geometrical changes in faces within the images. The better enhancement combination is the one that always enhances the recognition results either with or without aligning the faces inside images.

An Enhanced Histogram Matching Approach 879 YALEB-AUTO YALEB-MANU Fig. 8. Samples from Yale B auto and manually cropped The images of a subject in each subset are divided into training and testing images as follows: subset 1 is divided into 3 training images and 5 testing images; each of subset 2, 3, and 4 is divided into 4 training images and 8, 8 and 10 testing images, respectively. As a result, the training set consists of 15 images 10 subjects while the testing set consists of the remaining 31 images 10 subjects. Fig.9 shows the training images in each subset, randomly selected, and the light angle of each image. There are 25 different training cases used in the testing, as shown in table 1, in which the normal image is common in all cases. These training cases are chosen to cover both the training with each elementary subset namely subset 1, 2, 3, and 4, and the training with the seven combinations of these subsets where subset 1 is essential in all of them as it contains the lowest illumination. Each elementary subset is composed of training by the normal image and either the vertical, horizontal or both lighting. While each combination is composed of training by the normal image and either vertical lighting or vertical and horizontal lighting. These training varieties help us to test the robustness of each enhancement combination against the number of training images and the changes in illumination direction of these images. The better enhancement combination, the one that always increases the recognition rates regardless of the training case. 5 Experimental Results The aim of these experiments is to choose the best enhancement combination from the 40 different combinations described in section 3. Thus, each combination is applied four different times corresponding to the Eigenface and Spectroface methods over YALE B-AUTO and YALE B-MANU versions. In each time, a combination is tested over the 25 training cases and their average recognition rate is calculated and then compared with the one resulting from applying the HM alone. The best enhancement combination, the one that increases the recognition rates resulting from applying the HM alone in all of the following: 1. Both face recognition methods (Eigenface and Spectroface), 2. Over aligned and non-aligned versions (YALE B- MANU and YALE B- AUTO), 3. In all the 25 training cases. The first condition is to ensure the flexibility of the chosen combination among different face recognition methods. While the second ensures its suitability for real-life systems, in which perfect aligning of the faces inside images is not a simple task. Finally, by ensuring the increasing of recognition rates in all the 25 training cases, it proves that the chosen combination is not affected by either the number of training images or the changes in illumination direction of these images.

880 A. Salah-ELDin, K. Nagaty, and T. ELArif Subset 1 Subset 2 Subset 3 Subset 4 Normal Vertical Vertical Horizontal Vertical Horizontal Vertical Horiz & Ver (+ 10, - 10 ) (+ 25, - 25 ) (+ 20, - 20 ) (+ 50, - 50 ) (+ 45, - 35 ) (+ 70, - 70 ) (+ 65, ± 35 ) Fig. 9. Training images for one subject in the four subsets with the light angle of each image Table 1. The 25 different training cases used in testing Subsets 1 2 3 4 nor: normal ver: vertical Elementary Subsets Training Case(train. images/subject) nor only nor + 2 ver nor + 2 ver nor + 2 hor nor + 2 ver + 2 hor nor + 2 ver nor + 2 hor nor + 2 ver + 2 hor nor + 2 ver nor + 2 hor nor + 2 ver + 2 hor hor: horizontal Subsets 1, 2 1, 3 1, 4 1, 2, 3 1, 2, 4 1, 3, 4 1, 2, 3, 4 Seven Combinations Training Case(train. images/subject) nor + 4 ver nor + 4 ver + 2 hor nor + 4 ver nor + 4 ver + 2 hor nor + 4 ver nor + 4 ver + 2 hor nor + 6 ver nor + 6 ver + 4 hor nor + 6 ver nor + 6 ver + 4 hor nor + 6 ver nor + 6 ver + 4 hor nor + 8 ver nor + 8 ver + 6 hor As described in section 3, 32 out of the 40 enhancement combinations are for further enhancement. To see if further enhancement the image leads to further increasing the recognition rates or not, table 2 shows the number of further enhancement combinations that lead to average increasing in recognition rates more than those accomplished by the single enhancement combinations. It s clear from table 2 that further enhancement the image using any of the three traditional enhancement methods namely HE, GAMMA and LOG, doesn t lead to further enhancement in recognition rates of both the Eigenface and Spectroface methods especially in the YALE B-MANU version. Only COMP that s lead to further enhancement in recognition rates of both face recognition methods over the two database s versions. For example, in Spectroface method over the YALE B-MANU (last row in table 2), we can see that when applying HE as further enhancement after each of the eight single combinations, none of these combinations get further increasing in its average recognition rate due to further enhancement it with HE. The same thing is happened when applying either GAMMA or LOG as further enhancement. On other hand, when applying COMP as further enhancement, five out of the eight combinations get further increasing in their average recognition rates after applying the COMP over those accomplished before applying it. As a result, only five out of 40 enhancement combinations are satisfying the three previously mentioned conditions, their effect is shown in Fig.10:

An Enhanced Histogram Matching Approach 881 Table 2. The number of combinations that lead to increase the recognition rates after using each of the enhancement methods for further enhancement Face Recognition Method Eigenface Spectroface Further Enhancement Using: Database HE GAMMA LOG COMP (8 combinations) (8 combinations) (8 combinations) (8 combinations) YALE B-AUTO 0 5 5 8 YALE B-MANU 0 0 1 8 YALE B-AUTO 0 1 0 5 YALE B-MANU 0 0 0 5 1. GAMMA-HM, where gamma is applied before HM. 2. GAMMA-HM-COMP, where gamma is applied before HM, then the result is further enhanced by applying the compression function. 3. HE-HM-COMP, where equalization is applied before HM, then the result is further enhanced by applying the compression function. 4. COMP-HM-COMP, where compression function is applied before HM, then the result is further enhanced by applying it again. 5. HM-HE-COMP, where equalization is applied after HM, then the result is further enhanced by applying the compression function. Illuminated GAMMA-HM GAMMA-HM-COMP HE-HM-COMP COMP-HM-COMP HM-HE-COMP Fig. 10. Effects of the five enhancement combinations that satisfy the three conditions Table 3 (a, b) shows the results of using these combinations with the Eigenface and the Spectroface methods, respectively, over both versions of the Yale B database. It shows the average recognition rate of each combination over the 25 training cases and the difference between it and the average recognition rate of applying the HM alone. It appears from table 3 (a) that using the second enhancement combination, namely GAMMA-HM-COMP, with the Eigenface method gives the best difference of averages over the four other combinations in both database s versions. While in the Spectroface method, table 3 (b) shows that there are no significant differences between using any of the five combinations in each of the database s versions. As a result, we can choose the GAMMA-HM-COMP combination as the best enhancement combination over the 40 different combinations according to the criteria stated above. 5.1 Complexity of the Proposed Approach The GAMMA-HM-COMP approach based on applying three consecutive steps, namely GAMMA, HM and compression function of the Retinal filter. For an N N image, both GAMMA and HM take O(N 2 ). Since the compression function is based on Gaussian filtering the image by applying the 1D Gaussian filter twice, it takes order O(N 2 k), where k is the mask size. But since the mask size is fixed and equal to 15 in our case [12], the overall complexity of the GAMMA-HM-COMP approach remains O(N 2 ) which is equal to the complexity of using the HM alone.

882 A. Salah-ELDin, K. Nagaty, and T. ELArif Table 3. Results of using the best five combinations with (a) Eigenface and (b) Spectroface methods over the two Yale B versions. Average recognition rate is calculated over the 25 different training cases. (1: GAMMA-HM, 2: GAMMA-HM-COMP, 3: HE-HM-COMP, 4: COMP-HM-COMP, 5: HM-HE-COMP). (a) Eigenface method YALE B-AUTO YALE B-MANU HM 1 2 3 4 5 HM 1 2 3 4 5 Average Recognition Rate 66.4 72.5 73.3 71.3 70.4 71.1 90.8 95.7 96.6 95.4 95.1 95.3 Difference of Averages - 6.1 6.9 4.9 4.0 4.7-4.9 5.8 4.6 4.3 4.5 (b) Spectroface method YALE B-AUTO YALE B-MANU HM 1 2 3 4 5 HM 1 2 3 4 5 Average Recognition Rate 73.2 76.9 77.4 77.2 77.3 77.4 80.6 87.4 87.7 87.3 88.0 87.4 Difference of Averages - 3.7 4.2 4.0 4.1 4.2-6.8 7.1 6.6 7.4 6.7 6 Conclusions and Future Works This paper introduces a new image-processing-based illumination normalization approach based on enhancing the image resulting from histogram matching using the gamma correction and the Retinal filter s compression function, which we called GAMMA-HM-COMP approach. It is based on three consecutive steps: 1) Applying the gamma correction on the reference average well-lit image, 2) Histogram matching the input image to the result from 1, 3) Applying the Retinal filter s compression function to further enhancing the result of 2. Among 40 different enhancement combinations, GAMMA-HM-COMP approach proves its flexibility among different face recognition methods and suitability for practical/real-life systems in which perfect aligning of the face is not a simple task. It leads to average increasing in recognition rates over HM alone ranges from 4~7% in Eigenface and Spectroface methods using aligned and non-aligned versions of the Yale B database. However, in this study, the compression function of the Retinal filter is newly applied as an image enhancement method. It proves its suitability for further enhancement rather than the other three traditional enhancement methods which are the histogram equalization, gamma correction and log transformation. Our future work is to empirically compare the GAMMA-HM-COMP approach with other best-of-literature illumination normalization approaches, taking into consideration both the flexibility among different face recognition methods and the robustness for the face aligning problem. In addition, the influence of the GAMMA- HM-COMP approach on non-illuminated face images needs to be investigated. References [1] Phillips, P.J., Grother, P., Micheals, R.J., Blackburn, D.M., Tabassi, E., Bone, J.M.: FRVT 2002: Evaluation Report (2003), http://www.frvt.org [2] Phillips, P.J., Scruggs, W.T., O Toole, A.J., Flynn, P.J., Bowyer, K.W., Schott, C.L., Sharpe, M.: FRVT 2006 and ICE 2006 Large-Scale Results. National Institute of Standards and Technology, NISTIR 7408 (2007), http://face.nist.gov [3] Adini, Y., Moses, Y., Ullman, S.: Face recognition: The problem of compensating for changes in illumination direction. IEEE Tran. PAMI 19(7), 721 732 (1997)

An Enhanced Histogram Matching Approach 883 [4] Zhao, W., Chellappa, R.: Robust face recognition using symmetric shape-from-shading Technical report, Center for Automation Research, University of Maryland (1999) [5] Basri, R., Jacobs, D.W.: Lambertian reflectance and linear subspaces. IEEE Trans. on Pattern Analysis and Machine Intelligence 25(2), 218 233 (2003) [6] Gross, R., Matthews, I., Baker, S.: Eigen Light-Fields and Face Recognition Across Pose. In: Proceedings of the Fifth IEEE international Conference on Automatic Face and Gesture Recognition, Washington, USA (2002) [7] Georghiades, A., Belhumeur, P., Kriegman, D.: From few to many: Illumination cone models for face recognition under variable lighting and pose. IEEE Trans. on Pattern Analysis and Machine Intelligence 23(6), 643 660 (2001) [8] Du, B., Shan, S., Qing, L., Gao, W.: Empirical comparisons of several preprocessing methods for illumination insensitive face recognition. In: Proceedings ICASSP 2005, vol. 2 (2005) [9] Heusch, G., Rodriguez, Y., Marcel, S.: Local Binary Patterns as an Image Preprocessing for Face Authentication. In: Proceedings of the 7th international Conference on Automatic Face and Gesture Recognition (Fgr 2006), Washington (2006) [10] Santamara, M.V., Palacios, R.P.: Comparison of Illumination Normalization Methods for Face Recognition. In: Third COST 275 Workshop Biometrics on the Internet, Univ. of Hertfordshire, UK (2005) [11] Levine, M.D., Gandhi, M.R., Bhattacharyya, J.: Image Normalization for Illumination Compensation in Facial Images. Department of Electrical & Computer Engineering & Center for Intelligent Machines, McGill University, Montreal, Canada (Unpublished Report, 2004) [12] Yang, J., Chen, X., Kunz, W., Kundra, H.: Face as an index: Knowing who is who using a PDA. Inter. Journal of Imaging Systems and Technology 13(1), 33 41 (2003) [13] Jebara, T.: 3D Pose estimation and normalization for face recognition, Honours Thesis, McGill University, Canada (1996) [14] Dubuisson, S., Davoine, F., Masson, M.: A solution for facial expression representation and recognition. Signal Process. Image Commun. 17(9), 657 673 (2002) [15] Georghiades, A., Belhumeur, P., Kriegman, D.: From few to many: Generative models for recognition under variable pose and illumination. In: Proc. IEEE FG 2000, pp. 277 284 (2000) [16] Turk, M., Pentland, A.: Eigenfaces for recognition. Journal of Cognitive Neuroscience 3(1), 71 86 (1991) [17] Lai, J., Yuen, P., Feng, G.: Face recognition using holistic Fourier invariant features. Pattern Recognition 34(1), 95 109 (2001) [18] El-Arief, T.I., Nagaty, K.A., El-Sayed, A.S.: Eigenface vs. Spectroface: A Comparison on the Face Recognition Problems. In: IASTED Signal Processing, Pattern Recognition, and Applications (SPPRA), Austria (2007) [19] Gonzales, R.C., Woods, R.E.: Digital Image Processing, 2nd edn. Addison Wesley Publishing Company, Inc., New York (2001) [20] Hammal, Z., Eveno, N., Caplier, A., Coulon, P.: Parametric models for facial features segmentation. Signal Process 86(2), 399 413 (2006) [21] Beaudot, W.: The neural information processing in the vertebra retina : a melting pot of ideas for artificial vision, Phd thesis, tirf laboratory, Grenoble, France (1994) [22] Hammal, Z., Massot, C., Bedoya, G., Caplier, A.: Eyes Segmentation Applied to Gaze Direction and Vigilance Estimation. In: Singh, S., Singh, M., Apte, C., Perner, P. (eds.) ICAPR 2005. LNCS, vol. 3687, pp. 236 246. Springer, Heidelberg (2005) [23] Intel OpenCV Liberary, http://sourceforge.net/projects/opencvlibrary/ [24] Yale B face database, http://cvc.yale.edu/projects/yalefacesb/yalefacesb.html