Master Thesis Presentation, 14Dec07 Pair Wise Distance Histogram Based Fingerprint Minutiae Matching Algorithm Developed By: Neeraj Sharma M.S. student, Dongseo University, Pusan South Korea.
2Master Thesis Presentation: 14DEC07 Presented Contents Introduction Why Fingerprints: some facts Essential Preprocessing (Feature Extraction etc.) Abstract Previous Work Problem Simulation Steps of Algorithm Flow Chart Local Matching Global Matching Results Comparison with Reference method (Wamelen et al) Future work Publications
3Master Thesis Presentation: 14DEC07 Presented Introduction Fingerprints are most useful biometric feature in our body. Due to their durability, stability and uniqueness fingerprints are considered the best passwords. In places of access security, high degree authentication, and restricted entry, fingerprints suggests easy and cheap solutions.
4Master Thesis Presentation: 14DEC07 Presented Biometric Modalities
5Master Thesis Presentation: 14DEC07 Presented Market Capture by different Biometric modalities
6Master Thesis Presentation: 14DEC07 Presented
7Master Thesis Presentation: 14DEC07 Presented Different fingerprints of two fingers
8Master Thesis Presentation: 14DEC07 Presented Different Features in a Fingerprint Ridge Ending Enclosure Bifurcation Island Texture Singular points
9Master Thesis Presentation: 14DEC07 Presented Feature points extracted
10Master Thesis Presentation: 14DEC07 Presented Extraction of minutiae Image skeleton Gray scale image Minutia features Feature Extraction with CUBS-2005 algorithm ( Developed by SHARAT et al)
11Master Thesis Presentation: 14DEC07 Presented Feature points pattern of same finger
12Master Thesis Presentation: 14DEC07 Presented High level description of algorithms in FVC (Fingerprint Verification Competition)2004
13Master Thesis Presentation: 14DEC07 Presented Abstract Thesis proposes a novel approach for matching of minutiae points in fingerprint patterns. The key concept used in the approach is the neighborhood properties for each of the minutiae points. One of those characteristics is pair wise distance histogram, that remains consistent after the addition of noise and changes too.
14Master Thesis Presentation: 14DEC07 Presented Previous Work Fingerprint Identification is quite mature area of research. Its almost impossible to describe all the previous approaches in a short time here. The previous methods closely related to this approach and also taken in reference are by Park et al.[2005] and Wamelen et al.[2004]. Park et al. used pair wise distances first ever to match fingerprints in their approach before two years. Wamelen et al. gave the concept of matching in two steps, Local match and Global match.
15Master Thesis Presentation: 14DEC07 Presented Problem Simulation The input fingerprint of the same finger seems to be different while taken on different times. There may be some translational, rotational or scaling changes, depending upon situation. Our aim is to calculate these changes as a composite transformation parameter “T”. The verification is done after transforming the input with these parameters, new transformed pattern should satisfy desired degree closeness with template pattern. Template Input Pattern
16Master Thesis Presentation: 14DEC07 Presented Minutiae matching-Aligning two point sets Input Template
17Master Thesis Presentation: 14DEC07 Presented Algorithm Steps The algorithm runs in two main steps:- (i) Local matching (ii) Global matching In local matching stepwise calculations are there: 1.Calculate “k” nearest neighbors for each and every point in both patterns. 2.Calculate histogram of pair wise distances in the neighborhood of every point. 3.Find out the average histogram difference between all the possible cases. 4.Set the threshold level of average histogram difference 5.Compare the average histogram differences with the threshold level.
18Master Thesis Presentation: 14DEC07 Presented Flow Chart No Point pattern “P” stored in database pattern “Q” is taken that is to be matched with “P” Select a local point and it’s “k” nearest neighbors in patterns P Select a local point and it’s “k” nearest neighbors in patterns Q Make pair wise distance Histogram Average histogram difference < Threshold level Calculate and store transformation parameter Iteration algorithm to calculate final Transformation Parameter Start All point’s in pattern p is examined End
19Master Thesis Presentation: 14DEC07 Presented Calculation of “k” Nearest neighbors (Local match) For the given input fingerprint pattern and the template pattern, calculate “k” nearest neighbors in order to distances. Here k is a constant can be calculated with the formula given by wamelen et al.(2004)
20Master Thesis Presentation: 14DEC07 Presented Histogram Calculation (Local Match) Histogram of pair wise distances in their neighborhood for each and every point is calculated here. It describe the variety of distances of particular point in its neighborhood. Here for one point “P 1 ”; P 1 n 1, P 1 n 2, P 1 n 3, P 1 n 4, P 1 n 5 are five nearest neighbors. Note: step size is 0.04unit, here. P1n2P1n2 P1P1 P1n1P1n1 P1n2P1n2 P1n3P1n3 P1n4P1n4 P1n5P1n5 P1n1P1n1 P1n3P1n3 P1n4P1n4 P1n5P1n5 P1n2P1n2 P1n3P1n3 P1n4P1n4 P1n5P1n5 P1n3P1n3 P1n4P1n4 P1n5P1n5 P1n4P1n4 P1n5P1n5
21Master Thesis Presentation: 14DEC07 Presented Average Histogram Difference and Threshold Setting (Local Match) To calculate average histogram differences for two points, first subtract the their histograms. It comes in a form of matrix. To calculate average, just normalize it on corresponding scale. H 1 =[ ] H 2 =[ ] H 1 - H 2 =[ ] Average histogram diff.(ΔH avg )=(1/10)*Σ(| H 1 - H 2 | i ) Setting of threshold depends on the size of point pattern. Larger the number of points, smaller the threshold count. Every matching pair is related with a transformation function. That transformation parameter is calculated mathematically.
22Master Thesis Presentation: 14DEC07 Presented Threshold check for a 20 points pattern input
23Master Thesis Presentation: 14DEC07 Presented Transformation Parameter calculation On the basis of histogram differences, we can make decision on local matching pairs. Then the transformation parameter is calculated in the following way by least squire method. Here “r” represents the corresponding Transformation Parameter.
24Master Thesis Presentation: 14DEC07 Presented Axial Representation of all Transformation Parameters after Local match Three axes represent the translational (in both x& y direction), rotaional and scale changes. The most dense part in graph represents the correct transformation parameters only. We need to conclude our results to that part.
25Master Thesis Presentation: 14DEC07 Presented Mean and Standard Deviation Mean and standard deviation is calculated with the following mathematical equations.
26Master Thesis Presentation: 14DEC07 Presented Global Matching (Iteration Algorithm) Iteration method is used to converge the result towards dense part of the graph. For applying this method we need to calculate mean and standard deviation of the distribution. In the graph all transformation parameters are present, calculated after local matching step. The mean for this distribution is shown by the “triangle” in centre.
27Master Thesis Presentation: 14DEC07 Presented Result after first iteration In this graph, black triangle is describing the mean for the distribution. After one iteration step some of the transformation parameters, due to false local match got removed.
28Master Thesis Presentation: 14DEC07 Presented Result after second iteration After second iteration, mean converges more towards the dense area. Black triangle is the mean point for this distribution shown here.
29Master Thesis Presentation: 14DEC07 Presented After third iteration Performing iterations to converge the result, gives the distribution having least standard deviation. Black star in this graph is the desired transformation parameter i.e. “r”
30Master Thesis Presentation: 14DEC07 Presented Verification by transforming the template with calculated parameters Template Pattern in Database Transformed version with Parameter ”r”Verification by Overlapping with original input pattern Transformed With parameter “r”
31Master Thesis Presentation: 14DEC07 Presented Matching Result for Ideal point sets No missing point N=60 Exactly matching
32Master Thesis Presentation: 14DEC07 Presented Random and Normalized Noise pattern (a) (b) (c) (d)
33Master Thesis Presentation: 14DEC07 Presented Results With Randomly Missing Points Missing points =20 Total points=60 Matching points=36 Matching factor= 0.76 Noise factor=0.031
34Master Thesis Presentation: 14DEC07 Presented Matching results after missing points Half no. of points (30) missing from pattern Missing points =30 Total points=60 Matching points=16 Matching factor= 0.91 Noise factor=0.021
35Master Thesis Presentation: 14DEC07 Presented Definitions t= λ *r/(2*√n) t is distance of closeness which can be some fraction of minimum pair-wise distance λ is called “ matching factor ”, depends on point pattern r=maximum pair-wise distance/2 N is no. of points in the pattern η is “ noise factor ” shows the extent of noise added to point pattern η= added average error/mean pair-wise distance
Master Thesis Presentation, 14Dec07 Results for different λ and η Total pointsMatching factor(λ) Noise factor(η) Time of match(sec) Accuracy(% )
37Master Thesis Presentation: 14DEC07 Presented Performance with missing points regionally
38Master Thesis Presentation: 14DEC07 Presented Performance with missing points regionally
39Master Thesis Presentation: 14DEC07 Presented Performance with missing points regionally
40Master Thesis Presentation: 14DEC07 Presented Performance with a real fingerprint
41Master Thesis Presentation: 14DEC07 Presented Results with real and random fingerprints The algorithm was tested on both randomly generated point pattern and real data base. The results shows correct identification in more than 93.73% cases out of 500 tests, with randomly generated data. For real fingerprint data, method was tested on some FVC (Fingerprint Verification Competition) 2004 samples. In most of the cases performance was found satisfactory.
42Master Thesis Presentation: 14DEC07 Presented Comparative Performances of two methods over randomly data Total points to match Missing PointsExternal noise added (%) Translation [x y] Time of match with wamelen’s method (sec.) Time of match with histogram method (sec.)
43Master Thesis Presentation: 14DEC07 Presented Comparison of performances
44Master Thesis Presentation: 14DEC07 Presented Advantages of the Method over others Proposed earlier This algorithm undergoes two steps, so accuracy is good and false acceptance rate is low. Calculation is less complex with comparison to other methods proposed yet. Here, histogram is a basis to select the local matching pairs, while in other randomize algorithms are lacking in any basic attribute to compare. Performance is better in case with missing points from a specific region.
45Master Thesis Presentation: 14DEC07 Presented Limitations This algorithm is dependent on accuracy of feature extraction method used for minutiae extraction. Method performs well if the number of missing points in the pattern is less than 50% of total minutiae points.
46Master Thesis Presentation: 14DEC07 Presented Future Work To enhance the performance of algorithm on real fingerprint data is also a big challenge. To calculate the computational complexity in big “O” notation. One important task is to develop an independent method for feature points extraction from fingerprints.
47Master Thesis Presentation: 14DEC07 Presented Publications Journal: 1.Sharma Neeraj, Lee Joon Jae “Fingerprint Minutiae Matching Algorithm Using Distance Histogram Of Neighborhood”, Journal of KMMS. (To be published in Dec edition) International Conferences: 1.Sharma Neeraj, Choi Nam Seok, Lee Joon Jae, “Fingerprint Minutiae Matching Algorithm Using Distance Distribution Of Neighborhood”, MITA (2007), Lye Wei Shi, Sharma Neeraj, Choi Nam Seok, Lee Joon Jae, Lee Byung Gook, “Matching Of Point Patterns By Unit Circle”, APIS(2007),
48Master Thesis Presentation: 14DEC07 Presented References 1. Wamelen P. B. Van, Li Z., and Iyengar S. S.: “A fast expected time algorithm for the 2-D point pattern matching problem. Pattern Recognition” 37, Elsevier Ltd, (2004), Park Chul-Hyun, Smith Mark J.T., Boutin Mireille, Lee J.J.: “Fingerprint Matching Using the Distribution of the Pairwise Distance Between Minutiae”, AVBPA (2005), LNCS 3546 (2005), Sakata Koji, Maeda Takuji, Matsushita Masahito, Sasakawa Koichi, Tamaki Hishashi: “Fingerprint Authentication based on matching scores with other data”, ICB, LNCS 3832, (2006), Maltoni D., Maio D., Jain A.K., Prabhakar S. :”Handbook of Fingerprint Recognition”, Springer Chang S.H., Cheng F. H., Hsu Wen-Hsing, Wu Guo-Zua: “Fast algorithm for point pattern matching: Invariant to translation, rotations and scale changes.” Pattern Recognition, Elsevier Ltd., Vol-30, No.-2, (1997), Irani S., Raghavan P.:” Combinatorial and Experimental Result on randomized point matching algorithms”, Proceeding of the 12th Annual ACM symposium on computational geometry, Philadelphia, PA, (1996), Adjeroh D.A., Nwosu K.C.: ”Multimedia Database Management – Requirements and Issues”, IEEE Multimedia. Vol. 4, No. 3, 1997, pp
49Master Thesis Presentation: 14DEC07 Presented Thanks for your kind attention.
50Master Thesis Presentation: 14DEC07 Presented
51Master Thesis Presentation: 14DEC07 Presented (a) (b) (c) (d)