Face Recognition Using Face-ARG Matching

1. Face Recognition Using
Face-ARG Matching
Bo-Gun Park, Kyoung-Mu Lee, Member, IEEE,
and Sang-Uk Lee, Member, IEEE
Zheng-Wen Shen
2006/04/11
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE
INTELLIGENCE, VOL. 27, NO. 12, DECEMBER 2005
1

2. Outline
1. Introduction
2. The description of the Face-ARG
3. The Correspondence Graph
4. Similarity between two Face-ARGs
and Face Recognition
5. Experimental Results
6. Conclusion
2

3. 1. Introduction
 A set of simple lines characterizing the
structure of an object are sufficient to
identify its shape and recognizable as
gray-level images
 A face is represented by the Face-ARG
model
 All the geometric quantities and the structural
information are encoded in an Attributed
Relational Graph
 A set of nodes (line features) and binary
relations between them
3

4. Two-phases
1. The correspondence graph of the
reference Face-ARG and the test Face-
ARG is constructed using the partial ARG
matching algorithm through the
stochastic analysis of the feature
correspondence in the relation vector
space.
2. The stochastic distance between the
corresponding relation vector spaces of
the extracted subgraphs is evaluated
and compared for the identification of a
face.
4

5. An example of the Face-ARG
representation and partial matching
5

6. 2. The description of the Face-ARG
 An ordered triple Face-ARG model
defined as:
 V = {V1,…,VN}
 R ={rij}
 F ={Ri|i=1,…,N}
 Ri = {rij|j=1,..,N, j<>i}
6

7. 3. The Correspondence Graph
 Assume that two Face-ARGs, G1 and
G2 are given by
7

8. A Correspondence Graph
 A correspondence graph between G1
and G2:
8

9. A correspondence graph between G1
and G2
9

10. Block diagram and data flows for the
partial ARG matching process.
10

11. 4. Similarity between two Face-ARGs
and Face Recognition
 Similarity between Two Face-ARGs
11

12.  Face Recognition
 Select the best matched identity
among the database which gives the
highest similarity value above some
prespecified threshold as the final
recognition.
12

13. 5. Experimental Results
 For the evaluation of the proposed
algorithm’s performance, it was tested on
the AR face DB, which is composed of
color images of 135 people (76 men and
59 women).
 The AR face DB includes frontal view
images with different facial expressions,
illumination conditions, and occlusion by
sunglasses and scarf.
 All images used for experiments were
normalized to face images of 120 by 170
pixels.
13

14. The ARG face database
14

15. Analysis on the Effect of imprecise line
extraction
 Imprecise line extraction or noise
can affect the overall performance
of the line feature-based face
recognition algorithms
 Position error
 Loss of line segments
 broken line segments
15

16. Examples of feature variations due to
the imprecise line extraction. Position error
Missing lines broken lines
16

17. Position error
17

18. Missing lines
18

19. Broken lines
19

20. Recognition Results on the Real Face DB
20

21. 6. Conclusion
21