Features, Local Binary Pattern, Lcoal Features, Texture, threshold, Uniform Patterns. I. INTRODUCTION. A face recognition system is a computer application for.
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 6, June 2013)
Illustration of Usage of Local Binary Patterns for Feature Extraction in Face Recognition Namratha M1, Dr. S. Natarajan2 1
MTech in Software Engineering, Department of Information Science, PESIT, Bangalore, India Professor and Key Resource person, Department of Information Science, PESIT, Bangalore, India
2
Recognition algorithms[1] can be divided into two main approaches: 1. Geometric: which looks at distinguishing features 2. Photometric: which is a statistical approach that distills an image into values and compares the values with templates to eliminate variances. Feature extraction[2] in face recognition refers to a special form of dimensionality reduction. When the input data for the algorithm is too large to be processed, then the input data is transformed into a reduced representation set of features. Transforming input data into a set of features is called feature extraction. There are two kinds of features: 1. Domain specific features such as fingerprints, human face. 2. General features such as color, texture, shape. Texture refers to the surface of an object. Texture can be rough, smooth, horizontal or vertical. They capture patterns in an image such as repitiveness and granularity. Texture based feature extraction[3] is preferred because it is pose-invariant, is not affected by light variations or illumination and hence gives more accuracy.
Abstract - Face Recognition is a very challenging area and has a wide range of applications in various fields such as video surveillance systems, crime informatics, etc. Feature extraction is the basic step in any face recognition system. It is required because the number of inputs or size of input may be too large and hence we extract only features of interest. Texture based feature extraction is preferred due to it’s invariance to pose, light, reflection. Local Binary Pattern(LBP) are used for feature extraction where we segment the given input image into segments and apply LBP for each segment and then combine them to obtain the global features. Local features are used because they provide more accuracy then global features. Keywords - Face Recogntion, Feature Extraction, Global Features, Local Binary Pattern, Lcoal Features, Texture, threshold, Uniform Patterns
I. INTRODUCTION A face recognition system is a computer application for automatically identifying a person from a digital image or video frame . One of the methods to do this is by comparing selected facial features from the image and a facial database. It is used in security systems and can be compared to other biometrics such as fingerprint or eye iris recognition systems. Face recognition[1] is a very challenging area in computer vision and pattern recognition due to variations in facial expressions, poses, illumination. Face recognition is largely motivated by the need for access control, surveillance and security, telecommunication and digital library. Some face recognition algorithms identify facial features by extracting landmarks or features from an image of the face. For example, an algorithm may analyze the relative position, size, and/or shape of the eyes, nose, cheekbones, and jaw. These features are then used to search for other images with matching features.Some other algorithms normalize a gallery of face images and then compress the face data only saving the data in the image that is useful for face detection. A probe image is then compared with the face data.
II. B ASICS O F LOCAL B INARY P ATTERN (LBP) The LBP operator[4] is one of the best performing texture descriptors and it has been widely used in various applications. It has proven to be highly discriminative and its key advantages are its invariance to monotonic graylevel changes and computational efficiency which make it suitable for demanding image analysis tasks. The idea of using LBP for face description is motivated by the fact that faces can be seen as a composition of micropatterns which are well described by such operator. The LBP operator was originally designed for texture description. The operator assigns a label to every pixel of an image by thresholding the 3*3 neighborhood of each pixel with the center pixel value and considering the result as a binary number. Then, the histogram of the labels can be used as a texture descriptor.
329
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 6, June 2013) To be able to deal with textures at different scales, the LBP operator can be extended to use neighborhoods of different sizes. Defining the local neighborhood as a set of sampling points evenly spaced on a circle centered at the pixel to be labeled allows any radius and number of sampling points. Bilinear interpolation[5] is used when a sampling point does not fall in the center of a pixel. Another extension to the original operator is the definition of so-called uniform patterns. A local binary pattern is called uniform if the binary pattern contains at most two bitwise transitions from 0 to 1 or vice versa when the bit pattern is considered circular. For example, the patterns 00000000 (0 transitions), 01110000 (2 transitions) and 11001111 (2 transitions) are uniform whereas the patterns 11001001 (4 transitions) and 01010011 (6 transitions) are not. In the computation of the LBP histogram uniform patterns are used so that the histogram has a separate bin for every uniform pattern and all nonuniform patterns are assigned to a single bin. The reason for selecting the local feature-based approach[6] is that trying to build a holistic description of a face using texture methods is not reasonable since texture descriptors tend to average over the image area. This is a desirable property for ordinary textures, because texture description should usually be invariant to translation or even rotation of the texture and especially for small repetitive textures, the small-scale relationships determine the appearance of the texture and, thus, the largescale relations[7] do not contain useful information. For faces, however, the situation is different: retaining the information about spatial relations is important. This reasoning leads to the basic methodology of this work. The facial image is divided into local regions and texture descriptors are extracted from each region independently. The descriptors are then concatenated to form a global description of the face.
Step 6: Each of these values are then combined to obtain the global features with the help of a histogram [8],[9]. The histogram is plotted for each block separately and then cumulative sum is obtained. The original histogram is as shown in Figure 4. To ehance the contrast of the image histogram normalization is also done which is shown in Figure 5. LBP computation is also performed on each of the color channel in the given input image. Initially the given RGB image is converted to YCbCr for better performance. YCbCr color spaceis proposed because the Cr and Cb (chrominance) components are independent of the skin color, the human race and the lighting conditions (it is in the YcbCr color space that the luminance is decoupled from the color information). Furthermore, the Cb and Cr are the chrominance components used in MPEG and JPEG. After conversion LBP is computed for each color channel separately and then combined to obtain the final LBP. Furthermore, existing methods used LBP of each color component for feature extraction. Here opponent color LBP is also performed. The LBP computed for color channels are subtracted for example the LBP obtained for Y component and Cb component are subtracted to obtain a new value for LBP which is used for the extraction process. This is experimentally found to produce better results. We propose for our system to use the YCbCr color space,[10] because the Cr and Cb (chrominance) components are independent of the skin color, the human race and the lighting conditions (it is in the YcbCr color space that the luminance is decoupled from the color information). Furthermore, the Cb and Cr are the chrominance components used in MPEG and JPEG.
III. C ALCULATION O F LBP Step 1: Record the pixel values in a 3*3 matrix Step 2: The center pixel value is treated as the threshold value Step 3: A value 1 is assigned if the value in the cell is greater than or equal to the threshold. A value 0 is assigned if the value in the cell is lesser than the threshold. Step 4: The values are written from left to right in clockwise order to get the binary equivalent. Step 5: Similarly this procedure is applied for each of the blocks separately.
Figure 1: Input image
330
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 6, June 2013)
Figure 2: LBP for the given input image
Figure 3: LBP for each of the 16 blocks of the given input image
331
International Journal of Emerging Technology and Advanced Engineering Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 6, June 2013) Acknowledgement I would like to express my immense gratitude to Prof.Shylaja S S, Professor and Head of Department of ISE, PES Institute of Technology, Bengaluru, for her help and inspiration during the tenure of the course. REFERENCES [1]
S. Z. Li and A. K. Jain, Eds., Handbook of Face Recognition. Springer, 2005. [2] P. S. Penev and J. J. Atick, “Local feature analysis: A general statistical theory for object representation,” Network – Computation in Neural Systems, vol. 7, no. 3, pp. 477–500, August 1996. [3] T. Ojala, M. Pietika¨inen, and D. Harwood, “A Comparative Study of Texture Measures with Classification Based on Feature Distributions,” Pattern Recognition, vol. 29, no. 1, pp. 51-59, 1996. [4] T. Ojala, M. Pietika¨inen, and T. Ma¨enpa¨a¨, “Multiresolution GrayScale and Rotation Invariant Texture Classification with Local Binary Patterns,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 24, no. 7, pp. 971-987, July 2002. [5] T. Ahonen, A. Hadid, and M. Pietik¨ainen, “Face recognition with local binary patterns,” in Proc. 8th European Conference on Computer Vision, ser. Lecture Notes in Computer Science, vol. 3021. Springer, 2004, pp. 469–481. [6] T. Ojala, M. Pietik¨ainen, and D. Harwood, “A comparative study of texture measures with classification based on feature distributions,” Pattern Recognition, vol. 29, no. 1, pp. 51–59, 1996. [7] T. Ojala, M. Pietik¨ainen, and T. M¨aenp¨a¨a, “Multiresolution grayscale and rotation invariant texture classification with local binary patterns,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 7, pp. 971–987, Jul 2002. [8] B. S. Manjunath, J.-R. Ohm, V. V. Vasudevan, and A. Yamada, “Color and texture 5th June 2006 DRAFT 15 descriptors,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 11, no. 6, pp. 703–715, Jun 2001. [9] T. Ahonen, M. Pietik¨ainen, A. Hadid, and T. M¨aenp¨a¨a, “Face recognition based on appearance of local regions,” in Proc. 17th International Conference on Pattern Recognition, 2004, pp. III: 153– 156. [10] G. Zhang, X. Huang, S. Z. Li, Y. Wang, and X. Wu, “Boosting local binary pattern (LBP)- based face recognition,” in Proc. Advances in Biometric Person Authentication, ser. Lecture Notes in Computer Science, vol. 3338, 2004, pp. 179–186. [11] C. Zhu, C. E. Bichot, and L. Chen, “Multi-scale color local binary patterns for visual object classes recognition,” in Proc. IEEE ICPR, pp. 3065–3068, 2010. [12] Gholamreza Anbarjafari, “Face recognition using color local binary pattern from mutually independent color channels”, EURASIP Journal on Image and Video Processing, June 2013
Figure 4: Original histogram of the given input image
Figure 5: Normalized histogram
IV. CONCLUSION This paper presents the usage of texture based feature extraction for Face Recognition. We have adopted texture based feature extraction due to it’s accuracy as opposed to other techniques. Local Binary Pattern (LBP) is adopted and applied for each segment of the image separately and then fused to obtain global features. Since here we use local features, the accuracy is increased. Texture plays a very important role in Face Recognition because it is invariant to pose, lighting, reflection, shadow. Hence can accurately match the test image with the trained image and has applications in fields like crime informatics, video survillance.
332
