Slide 1

Face Recognition CPSC 4600/5600 @ UTC/CSE

Slide 2

2 Face Recognition Introduction Introduction Face recognition algorithms Face recognition algorithms Comparison Comparison Short summary Short summary

Slide 3

3 Face Recognition Algorithms We will introduce We will introduce Eigenfaces Eigenfaces Fisherfaces Fisherfaces Elastic Bunch-Graph Matching Elastic Bunch-Graph Matching

Slide 4

4 Eigenfaces Developed in 1991 by M.Turk Developed in 1991 by M.Turk Based on Principal Component Analysis (PCA) Based on Principal Component Analysis (PCA) Relatively simple Relatively simple Fast Fast Robust Robust

Slide 5

5 Eigenfaces PCA seeks directions that are efficient for representing the data PCA seeks directions that are efficient for representing the data efficient not efficient Class A Class B Class A Class B

Slide 6

6 Eigenfaces PCA maximizes the total scatter scatter Class A Class B

Slide 7

7 Eigenfaces PCA reduces the dimension of the data PCA reduces the dimension of the data Speeds up the computational time Speeds up the computational time

Slide 8

8

Slide 9

9 Eigenfaces, the algorithm Assumptions Assumptions Square images with Width = Height = N Square images with Width = Height = N M is the number of images in the database M is the number of images in the database P is the number of persons in the database P is the number of persons in the database

Slide 10

10 Eigenfaces, the algorithm The database The database

Slide 11

11 Eigenfaces, the algorithm We compute the average face We compute the average face

Slide 12

12 Eigenfaces, the algorithm Then subtract it from the training faces Then subtract it from the training faces

Slide 13

13 Eigenfaces, the algorithm Now we build the matrix which is N 2 by M Now we build the matrix which is N 2 by M The covariance matrix which is N 2 by N 2 The covariance matrix which is N 2 by N 2

Slide 14

14 Eigenfaces, the algorithm Find eigenvalues of the covariance matrix Find eigenvalues of the covariance matrix The matrix is very large The matrix is very large The computational effort is very big The computational effort is very big We are interested in at most M eigenvalues We are interested in at most M eigenvalues We can reduce the dimension of the matrix We can reduce the dimension of the matrix

Slide 15

15 Eigenfaces, the algorithm Compute another matrix which is M by M Compute another matrix which is M by M Find the M eigenvalues and eigenvectors Find the M eigenvalues and eigenvectors Eigenvectors of Cov and L are equivalent Eigenvectors of Cov and L are equivalent Build matrix V from the eigenvectors of L Build matrix V from the eigenvectors of L

Slide 16

16 Eigenfaces, the algorithm Eigenvectors of Cov are linear combination of image space with the eigenvectors of L Eigenvectors of Cov are linear combination of image space with the eigenvectors of L Eigenvectors represent the variation in the faces Eigenvectors represent the variation in the faces V is Matrix of eigenvectors

Slide 17

17 Eigenfaces, the algorithm A: collection of the training faces U: Face Space / Eigen Space

Slide 18

18 Eigenfaces Eigenface of original faces Eigenface of original faces

Slide 19

19 Eigenfaces, the algorithm Compute for each face its projection onto the face space Compute for each face its projection onto the face space Compute the threshold Compute the threshold

Slide 20

20 Eigenfaces: Recognition Procedure To recognize a face To recognize a face Subtract the average face from it Subtract the average face from it

Slide 21

21 Eigenfaces, the algorithm Compute its projection onto the face space U Compute its projection onto the face space U Compute the distance in the face space between the face and all known faces Compute the distance in the face space between the face and all known faces

Slide 22

22 Eigenfaces, the algorithm Reconstruct the face from eigenfaces Reconstruct the face from eigenfaces Compute the distance between the face and its reconstruction Compute the distance between the face and its reconstruction

Slide 23

23 Eigenfaces, the algorithm Distinguish between Distinguish between If then its not a face; the distance between the face and its reconstruction is larger than threshold If then its not a face; the distance between the face and its reconstruction is larger than threshold If then its a new face If then its a new face If then its a known face because the distance in the face space between the face and all known faces is larger than threshold If then its a known face because the distance in the face space between the face and all known faces is larger than threshold

Slide 24

24 Eigenfaces, the algorithm Problems with eigenfaces Problems with eigenfaces Different illumination Different illumination

Slide 25

25 Eigenfaces, the algorithm Problems with eigenfaces Problems with eigenfaces Different head pose Different head pose Different alignment Different alignment Different facial expression Different facial expression

Slide 26

26 Fisherfaces Developed in 1997 by P.Belhumeur et al. Developed in 1997 by P.Belhumeur et al. Based on Fishers Linear Discriminant Analysis (LDA) Based on Fishers Linear Discriminant Analysis (LDA) Faster than eigenfaces, in some cases Faster than eigenfaces, in some cases Has lower error rates Has lower error rates Works well even if different illumination Works well even if different illumination Works well even if different facial express. Works well even if different facial express.

Slide 27

27 Fisherfaces LDA seeks directions that are efficient for discrimination between the data LDA seeks directions that are efficient for discrimination between the data Class A Class B

Slide 28

28 Fisherfaces LDA maximizes the between-class scatter LDA maximizes the between-class scatter LDA minimizes the within-class scatter LDA minimizes the within-class scatter Class A Class B

Slide 29

29 Fisherfaces, the algorithm Assumptions Assumptions Square images with Width=Height=N Square images with Width=Height=N M is the number of images in the database M is the number of images in the database P is the number of persons in the database P is the number of persons in the database

Slide 30

30 Fisherfaces, the algorithm The database The database

Slide 31

31 Fisherfaces, the algorithm We compute the average of all faces We compute the average of all faces

Slide 32

32 Fisherfaces, the algorithm Compute the average face of each person Compute the average face of each person

Slide 33

33 Fisherfaces, the algorithm And subtract them from the training faces And subtract them from the training faces

Slide 34

34 Fisherfaces, the algorithm We build scatter matrices S 1, S 2, S 3, S 4 We build scatter matrices S 1, S 2, S 3, S 4 And the within-class scatter matrix S W And the within-class scatter matrix S W

Slide 35

35 Fisherfaces, the algorithm The between-class scatter matrix The between-class scatter matrix We are seeking the matrix W maximizing We are seeking the matrix W maximizing

Slide 36

36 Fisherfaces, the algorithm If S W is nonsingular ( ): Columns of W are eigenvectors of Columns of W are eigenvectors of We have to compute the inverse of S W We have to compute the inverse of S W We have to multiply the matrices We have to multiply the matrices We have to compute the eigenvectors We have to compute the eigenvectors

Slide 37

37 Fisherfaces, the algorithm If S W is nonsingular ( ): Simpler: Simpler: Columns of W are eigenvectors satisfying Columns of W are eigenvectors satisfying The eigenvalues are roots of The eigenvalues are roots of Get eigenvectors by solving Get eigenvectors by solving

Slide 38

38 Fisherfaces, the algorithm If S W is singular ( ): Apply PCA first Apply PCA first Will reduce the dimension of faces from N 2 to M Will reduce the dimension of faces from N 2 to M There are M M -dimensional vectors There are M M -dimensional vectors Apply LDA as described Apply LDA as described

Slide 39

39 Fisherfaces, the algorithm Project faces onto the LDA-space Project faces onto the LDA-space To classify the face To classify the face Project it onto the LDA-space Project it onto the LDA-space Run a nearest-neighbor classifier Run a nearest-neighbor classifier

Slide 40

40 Fisherfaces, the algorithm Problems Problems Small databases Small databases The face to classify must be in the DB The face to classify must be in the DB

Slide 41

41 PCA & Fishers Linear Discriminant

Slide 42

42 PCA & Fishers Linear Discriminant

Slide 43

43 Comparison FERET database FERET database best ID rate: eigenfaces 80.0%, fisherfaces 93.2% best ID rate: eigenfaces 80.0%, fisherfaces 93.2%

Slide 44

44 Comparison Eigenfaces Eigenfaces project faces onto a lower dimensional sub- space project faces onto a lower dimensional sub- space no distinction between inter- and intra-class variabilities no distinction between inter- and intra-class variabilities optimal for representation but not for discrimination

Slide 45

45 Comparison Fisherfaces Fisherfaces find a sub-space which maximizes the ratio of inter-class and intra-class variability find a sub-space which maximizes the ratio of inter-class and intra-class variability same intra-class variability for all classes same intra-class variability for all classes

Slide 46

46 Local Feature Analysis -- Elastic Bunch-Graph Matching

Slide 47

47 Face Features Facial recognition utilizes distinctive features of the face including: distinct micro elements like: Facial recognition utilizes distinctive features of the face including: distinct micro elements like: Mouth, Nose, Eye, Cheekbones, Chin, Lips, Forehead, Ears Mouth, Nose, Eye, Cheekbones, Chin, Lips, Forehead, Ears Upper outlines of the eye sockets, the areas surrounding the cheekbones, the sides of the mouth, and the location of the nose and eyes. Upper outlines of the eye sockets, the areas surrounding the cheekbones, the sides of the mouth, and the location of the nose and eyes. The distance between the eyes, the length of the nose, and the angle of the jaw. The distance between the eyes, the length of the nose, and the angle of the jaw.

Slide 48

48 Face Features Some technologies do not utilize areas of the face located near the hairline, so they are somewhat resistant to moderate changes in hairstyle. Some technologies do not utilize areas of the face located near the hairline, so they are somewhat resistant to moderate changes in hairstyle. When used in identification mode, facial recognition technology generally returns candidate lists of close matches as opposed to returning a single definitive match as does fingerprint and iris-scan. When used in identification mode, facial recognition technology generally returns candidate lists of close matches as opposed to returning a single definitive match as does fingerprint and iris-scan. The file containing facial micro features is called a "template." The file containing facial micro features is called a "template." Using templates, the software then compares that image with another image and produces a score that measures how similar the images are to each other. Using templates, the software then compares that image with another image and produces a score that measures how similar the images are to each other.

Slide 49

49 Typical sources of images for use in facial recognition include video camera signals and pre-existing photos such as those in driver's license databases. including: Typical sources of images for use in facial recognition include video camera signals and pre-existing photos such as those in driver's license databases. including: Distance between the micro elements Distance between the micro elements A reference feature A reference feature Size of the micro element Size of the micro element Amount of head radiated from the face (unseen by human eye). Heat can be measured using an infrared camera. Amount of head radiated from the face (unseen by human eye). Heat can be measured using an infrared camera. Face Features

Slide 50

50 A face recognition based on local feature analysis A face is represented as a graph, whose nodes, positioned in correspondence to the facial fiducial points. A fiducial point is a point or line on a scale used for reference or comparison purposes. A face recognition system uses an automatic approach to localize the facial fiducial points. It then determines the head pose and compares the face with the gallery images. This approach is invariant to rotation, light and scale.

Slide 51

51 A template for the 34 fiducial points on a face image:

Slide 52

52 EBGM Elastic Bunch-Graph Matching (EBGM) algorithm locates landmarks on an image, such as the eyes, nose, and mouth. Gabor jets are extracted from each landmark and are used to form a face graph for each image. A face graph serves the same function as the projected vectors in the PCA or LDA algorithm; they represent the image in a low dimensional space. After a face graph has been created for each test image, the algorithm measures the similarity of the face graphs. Paper:http://www.snl.salk.edu/~fellous/posters/Bu97post er/BUPoster.pdf Paper:http://www.snl.salk.edu/~fellous/posters/Bu97post er/BUPoster.pdfhttp://www.snl.salk.edu/~fellous/posters/Bu97post er/BUPoster.pdfhttp://www.snl.salk.edu/~fellous/posters/Bu97post er/BUPoster.pdf

Slide 53

53 Summary Three algorithms have been introduced Three algorithms have been introduced Eigenfaces Eigenfaces Reduce the dimension of the data from N 2 to M Reduce the dimension of the data from N 2 to M Verify if the image is a face at all Verify if the image is a face at all Allow online training Allow online training Fast recognition of faces Fast recognition of faces Problems with illumination, head pose etc Problems with illumination, head pose etc

Slide 54

54 Summary Fisherfaces Reduce dimension of the data from N 2 to P-1 Reduce dimension of the data from N 2 to P-1 Can outperform eigenfaces on a representative DB Can outperform eigenfaces on a representative DB Works also with various illuminations Works also with various illuminations Can only classify a face which is known to DB Can only classify a face which is known to DB Elastic Bunch-Graph Matching Reduce the dimension of the data from N 2 to M Reduce the dimension of the data from N 2 to M Recognize face with different poses Recognize face with different poses Recognize face with different expressions Recognize face with different expressions

Slide 55

55 References [1] M. Turk, A. Pentland, Face Recognition Using Eigenfaces [2] J. Ashbourn, Avanti, V. Bruce, A. Young, Face Recognition Based on Symmetrization and Eigenfaces [3] http://www.markus-hofmann.de/eigen.html [4] Eigenfaces vs Fisherfaces: Recognition using Class Specific Linear Projection [4] P. Belhumeur, J. Hespanha, D. Kriegman, Eigenfaces vs Fisherfaces: Recognition using Class Specific Linear Projection [5] R. Duda, P. Hart, D. Stork, Pattern Classification, ISBN 0-471- 05669-3, pp. 121-124 [6] F.Deformable Face Mapping For Person Identification, ICIP 2003, Barcelona [6] F. Perronin, J.-L. Dugelay, Deformable Face Mapping For Person Identification, ICIP 2003, Barcelona [7] B. Moghaddam, C. Nastar, and A. Pentland. A bayesian similarity measure for direct image matching. ICPR, B:350358, 1996. http://www.face-rec.org/interesting-papers/

Slide 56

56 Hands-on Lab of Face Biometrics Present one of the following algorithms Elastic Bunch-Graph Matching (EBGM) algorithm Bayesian Intrapersonal/Extrapersonal Classifier, or http://www.face-rec.org/interesting- papers/ One from http://www.face-rec.org/interesting- papers/ http://www.face-rec.org/interesting- papers/http://www.face-rec.org/interesting- papers/ Hands-on Lab of Face Biometrics Hands-on Lab of Face Biometrics http://www.cs.colostate.edu/evalfacerec/ http://www.cs.colostate.edu/evalfacerec/ http://www.cs.colostate.edu/evalfacerec/ User Guide User Guide