WO2003085591A1

WO2003085591A1 - Method for synthesizing a minutiae template and increasing reliability of fingerprint matching using a fingerprint sensor with small sensor area

Info

Publication number: WO2003085591A1
Application number: PCT/SG2002/000046
Authority: WO
Inventors: Kar Ann Toh; Wei Yun Yau; Xudong Jiang; Eyung Lim
Original assignee: Nanyang Technological University, Centre For Signal Processing
Priority date: 2002-03-20
Filing date: 2002-03-20
Publication date: 2003-10-16
Also published as: CN100447809C; CN1623166A

Abstract

A method for synthesizing a minutiae template from a first minutiae template and a second minutiae template, which at least partically overlap each other by selecting a minutia from a plurality of minutiae in the first minutiae template, selecting a minutia from a plurality of minutiae in the second minutiae template, matching the minutia in the first minutiae template with the minutia in the second minutiae template by defining an inage region around the minutia template by defining an image region around the minutia in the first minutiae template, defining an inage region around the minutia in the second minutiae template, determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, determining the image pixel value from a plurality of corresponding image pixels in the image region around the minutia in the second$iminutiae template, comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corrxesponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, determining the minutia in the first minutiae template to be matched with the minutia in the second minutia template to the minutiae pair if the result of the comparison fulfils a prescribed criterion, repeating all the above steps for all minutiae template, and transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined when there is at least one matched minutiae, thereby forming a synthesized minutiae template and incorporating codeword into the matching to improve the accuracy.

Description

METHOD FOR SYNTHESIZING A MINUTIAE TEMPLATE AND INCREASING RELIABILITY OF FINGERPRINT MATCHING USING A FINGERPRINT SENSOR

WITH SMALL SENSOR AREA

BACKGROUND OF THE INVENTION

This invention relates to a method, a device, a computer readable medium and a computer program for synthesizing a minutiae template and increasing reliability of fingerprint matching, in particular, using fingerprint images acquired from a fingerprint sensor with a small sensor area.

The word "fingerprint" is herein used as a representative of a fingerprint or a like pattern or figure. More particularly, the fingerprint may be an actual finger, a palm print, a toe print, a soleprint, and a streaked pattern composed of streaks. The fingerprint may also be a diagram drawn by a skilled person to represent a faint fingerprint which is, for example, left at the scene of a crime.

A person who needs to use a fingerprint verification or identification system for authentication of his identity is required to register his fingerprint with the system. During this registration phase, the fingerprint images of the user are acquired from a fingerprint sensor, and characteristic features of the fingerprint images are extracted and stored in a storage media of the system. The characteristic features of the fingerprint images are subsequently used to match with the characteristic features of a fingerprint image which is presented to the fingerprint sensor during the verification phase .

A fingerprint image acquired from a fingerprint sensor is called a fingerprint template, and the characteristic features extracted from the fingerprint template is called the minutiae template. A minutiae template collected during the registration phase of the fingerprint verification/identification system is called a registered minutiae template.

Methods to determine minutiae templates are disclosed in [1] and [2] . The basic concept of these two methods is to determine all the minutiae present in the fingerprint template. From these minutiae determined, the required parameters are subsequently determined.

Matching techniques for matching minutiae templates of two fingerprint templates are disclosed in [3], [4], [5] and [6].

Known optical fingerprint sensors usually have an imaging area of about 25.4 mm by 25.4 mm, which is usually sufficient to cover the entire surface of the finger when the finger is pressed on the imaging area of the optical fingerprint sensor.

However, a typical solid-state fingerprint sensor, such as the FPS 110 sensor from Veridicom, Inc. or the FingerTip™ sensor from Infineon, Inc. has a smaller sensor area of 15 mm by 15 mm and 11.1 mm by 14.3 mm, respectively, in order to ensure that the costs of the sensor are acceptably low. Thus, with such a small sensor area, it is usually not possible to detect the entire fingerprint with one detection iteration, i.e. with one fingerprint image, since the sensor area is not sufficient to cover the entire surface of the finger when the finger is pressed on the sensor area of such fingerprint sensor. A method disclosed in [7] thus solves the problem of using a small fingerprint sensor by providing a method for determining a total minutiae template from a^' plurality of partial minutiae templates acquired from a fingerprint sensor with a small sensor area. A partial minutiae template is chosen as the reference minutiae template from a plurality of partial minutiae templates acquired from the different portions of the fingerprint image of a same finger during the registration phase. A first partial minutiae template with an overlapping region with the reference minutiae template is chosen and matching process as disclosed in [6] between the two templates is performed to locate corresponding minutiae pairs. A transformation function is determined based on the minutiae pairs, and the first partial minutiae template is combined with the reference minutiae template by aligning onto the reference minutiae template using the transformation function. Therefore, a new reference minutiae template is formed. The process is iterated until all the partial minutiae templates are combined with the reference minutiae template to form a total minutiae template representing the total fingerprint template of the user.

The number of corresponding minutiae pairs between the reference minutiae template and the minutiae template to be aligned is to be equal or higher than a pre-determined number, in particular, 3. Otherwise, the accuracy may not be good.

However, with a fingerprint sensor having a small sensor area, this pre-determined number of corresponding minutiae pairs may not be easily obtained. An encoding scheme for fingerprint recognition and retrieval is disclosed in [8], which provides additional information to the minutiae present in the minutiae template of the fingerprint template. However, this additional information is only used to compare between two fingerprint templates. The encoding scheme does not cover aspects of comparison for partial fingerprints with reference to a large total fingerprint . In [9], [10] and [11], the security level of a fingerprint verification/identification system is increased by combining the identification or verification of other body parts or fingerprints of other fingers. However the input of different body parts or fingers are visible to other users and hence the sequence of presenting the other body parts or fingerprints is not kept secret. In [12], a method and apparatus are described for determining whether a user of a system is an authorized user by examining the keystroke characteristics of the user. It is noted that this method for keystroke sequence encoding does not apply to fingerprint verification/identification system.

In [13], a method using a constantly changing key depending on the user's body part is described.- The invention uses time stamps to encode the transmitted biometric data to improve transmission security. However this method is used for providing a secure data transmission and not a secure matching of fingerprints.

In [14], a method for combining and synchronizing biometric and/or non-biometric features to improve matching accuracy is described. In particular, in a speaker verification system, the utterance and the corresponding lips shape are sampled by the system. The utterance and the lips shape must match with the stored utterance and the corresponding lips shape for authentication to be successful. However this invention does not describe how a fingerprint can be used in such a synchronization system.

In [15], a method for matching two images is described. The matching described is a general matching technique of two images using correlation, and does not suggest matching of two minutiae for alignment of two fingerprint templates.

Therefore, methods to overcome the above problems and increase the security of a fingerprint verification/identification system using information of a plurality of partial fingerprint templates acquired from a fingerprint sensor with a small sensor area are desired.

SUMMARY OF INVENTION

It is thus an object of the invention to overcome at least one of the above problems, and also to increase the reliability of a fingerprint verification/identification system using fingerprint images acquired from a fingerprint sensor with a small sensor area. A method for synthesizing a minutiae template from a plurality of partial fingerprint templates, in^' particular, using a plurality of partial fingerprint templates acquired from a fingerprint sensor with a small sensor area is proposed.

A partial fingerprint template refers to a fingerprint image covering a portion of the entire fingerprint of a finger, such as fingerprint template acquired from a fingerprint sensor with a small sensor area. A partial minutiae template refers to the minutiae template extracted from a partial fingerprint template .

According to the invention, a minutia from a first minutiae template and a minutia from a second minutiae template is selected. An image region around the minutia of the first • minutiae template is compared with a image region around the minutia of the second minutiae template. The image pixel values of the image pixels of the two image regions are computed and compared using a correlation function.

In the preferred embodiment of the invention, an error function is used as the correlation function to determine the correlation of the two minutiae:

£(/«ι,m2).22||^{/ , '}.J^')- 2"²(0^'.))||, ( )e≡5

wherein ml is a first minutia surrounded by image region S , m2 is a second minutia surrounded by image region S , I'"^x(i, j) is an image pixel value at coordinate (i,j) in the image region S around ml from a first fingerprint image, I_ι ²(i' j) is ^an image pixel value at coordinate (i,j) in the second image region S around m2 from a second fingerprint image,

S is a set containing coordinate (i,j) defining a region around a minutia, preferably a circular region and /(.)is a transformation function relating the location, orientation and deformation between I^™l(i, j) and I"²(i, j) ■

When E(m\,m2) falls below a certain prescribed threshold value, the two minutiae ml and m2 are considered to be matched, and is called a matched minutiae pair.

The process described above is repeated to locate all matched minutiae pairs from all possible combination of minutiae pairs from the first minutiae template and the second minutiae template . The search process described above only searches within a small pre-defined image region around a minutia, and not the entire fingerprint template. This pre-defined image region is preferably a circular region centred around the minutia with radius of a few pixels length. Therefore, the computation of the search process is low, resulting in a fast means of searching for a plurality of matched minutiae pairs.

The transformation function f (. ) can be represented by a linear transformation matrix for the case of a linear transformation: XI T. X2

Wherein T is the linear transformation matrix, and XI is the transformed coordinate column vector of X2. X2 is the coordinate column vector of m2.

The transformation matrix in the above equation can be resolved into a rotation matrix R and a displacement vector d: cos ,(θ) - Sin(θ)

R = sin (θ) cos(0)

^'dx d = ^dy. whe rei n θ denotes the rotation angle, and dx and dy are the displacements in the x and y directions.

Therefore, the transformation for alignment can be written as: I = R . X2 + d

The transformation function /(.) relating to a matched minutiae pair from the first minutiae template and the second minutiae template is then used to align the second minutiae template onto the first minutiae template, thereby forming a synthesized minutiae template.

It is to be noted that a third minutiae template may be aligned onto the synthesized minutiae template according to the method according to the invention to form a new synthesized minutiae template. The synthesis of the new minutiae template is repeated until all the partial minutiae templates acquired from the fingerprint sensor are used to form a total synthesized minutiae template.

The described embodiments may be incorporated into a registration phase of an existing fingerprint verification/identification system to increase the reliability of synthesis of a minutiae template from a plurality of partial minutiae templates using a fingerprint sensor with a small sensor area. In this embodiment of the invention, a first minutiae template is chosen from a plurality of partial minutiae templates from a same user, and is matched, preferably using the matching method disclosed in [6], with a second minutiae template, having preferably an intersecting region with the first minutiae template. A list displaying all the matched minutiae pairs from the two templates is generated. If the number of matched minutiae pairs is above a pre-determined value, the second minutiae template is aligned onto the first minutiae template, preferably using the method disclosed in [7], thus forming a new synthesized minutiae template.

If the number of matched minutiae pairs falls below the predetermined value, the method of minutiae template synthesis according to the invention is used as the search process to search for any matched minutiae pairs which are not located during the matching process described above.

If at least one new minutiae pair is located to be matched from the search process, and the total number of matched minutiae pairs is above the pre-determined value, the second minutiae template is aligned onto the first minutiae template, preferably using the method disclosed in [7].

If there is at least one matched minutiae pair located, but the total number of matched minutiae pairs is below the predetermined value, the transformation function determined from the search process according to the invention, is used to align the second minutiae template onto the first minutiae template .

When there is no matched minutiae pair found during the matching and the search process described above, it is assumed there is no intersecting region between the two minutiae templates, and therefore, no alignment of the minutiae templates is performed.

When the second minutiae template is aligned onto the first minutiae template, a new synthesized minutiae template is formed. The whole synthesizing process is repeated to align a third minutiae template onto the synthesized minutiae template to synthesize a new minutiae template. When all the minutiae templates are synthesized into a single synthesized minutiae template, a total minutiae template is formed. The total minutiae template is stored in the storage media of the system and all the minutiae templates used for the synthesis process are discarded. The stored total minutiae template is called the registered total minutiae template, and is used in the verification phase for the verification of a user. In the minutiae template synthesis process of a fingerprint verification/identification system using a fingerprint sensor with a small sensor area, wherein the method of minutiae template synthesis according to the invention is incorporated, only a minimum of one matched minutiae pair between the two minutiae templates is needed to form a synthesized minutiae template from the two templates.

During the verification phase, a partial minutiae template acquired from a fingerprint sensor with a small sensor area is used to match with the registered total minutiae template.

The partial minutiae template of a fingerprint image entered by a user for verification during the verification phase is called the search partial minutiae template. There is usually a significant difference in the minutiae counts between the search partial minutiae templates and the registered total minutiae template, and a match score corresponding to the matched minutiae pairs in the intersecting region of the two templates may not accurately indicate if the minutiae pairs of the two templates are matched. Therefore, a normalization of the match score by a weighted average is performed using the following formula:

ms

wherein ms is the match score, nl is the minutiae count in the Area Of Interest of a search partial fingerprint template, n2 is the minutiae count in the Area Of Interest of the registered fingerprint template area which intersects with the Area Of Interest of the search partial fingerprint template. The Area of Interest (AOI) of a fingerprint template is defined as the non-empty image area of the template. The AOI may be represented by a continuous boundary or joined line segments, and is determined according to the disclosure in The reliability of the fingerprint verification/identification system can be enhanced by incorporating a fingerprint codeword system.

The spatial information of the registered total minutiae template is partitioned into different portions with reference to the centroid position of the AOI of the fingerprint template and a positional code is assigned to each of the partitioned portions. The assigned code may be a numeric or an alphabetic code. A positional codeword, consisting of a combination of the positional codes, is pre-registered by a user during registration, and the user enters the pre- registered positional codeword during verification by presenting the different portions of his fingerprint to the sensor in a sequence corresponding to the pre-registered positional codeword.

It is to be noted that the fingerprint codeword system according to the invention is in contra-distmction with a

Personal Identification Number (PIN) system. The fingerprint codeword system enters a codeword using different portions of a fingerprint, and the codeword m a PIN system is entered using a keypad.

To determine the position of the search partial minutiae template of the search partial fingerprint template entered, the centroid position of the search partial minutiae template is computed and the corresponding position of the centroid of the search partial minutiae template on the total minutiae template is determined based on the transformation function determined from the matching method disclosed in [6].

The centroid position of a minutiae template is computed using the following formula:

A _r MX wherein

My is the moment of the AOI of the fingerprint template about the y axis,

Mx is the moment of the AOI of the fingerprint template about the x axis, and A is the area of the AOI of the fingerprint template.

With the fingerprint codeword system incorporated into the fingerprint verification/identification system, a user does not only need to have the search partial minutiae template of his search partial fingerprint template matched with the registered total minutiae template stored in the storage media of the fingerprint verification/identification system for verification, but also to make sure that he presents the correct portions of his fingerprint to the fingerprint sensor in a unique sequence corresponding to the pre-registered positional codeword.

In another embodiment of the invention, a unique time duration the finger is presented to the sensor is incorporated into the verification phase of the fingerprint verification/identification system to increase the reliability of the system. A pre-determined time duration, or a time- based code is assigned to each of the positional code. A user needs to enter the time-based codeword corresponding to the positional codeword by presenting his finger to the fingerprint sensor for the pre-determined time duration for successful authentication. In another embodiment of the invention, a unique positional codeword corresponding to a specific day and/or time of the day is incorporated. A user not only need to have the search partial minutiae template of his fingerprint template matched with the registered total minutiae template, but also to enter the unique positional codeword using the fingerprint codeword system for the specific time and day.

In another embodiment of the invention, a pre-registered orientation angle of the registered partial fingerprint template during registration is used, whereby a user not only need to have the search partial minutiae template of his fingerprint template matched with the registered total minutiae template during verification, but also to present his fingerprint to the sensor such that the orientation angle of the search partial fingerprint template is within a predefined angle of the orientation angle of the registered partial fingerprint template.

In another embodiment of the invention, a user is required to present his fingerprint to the sensor a plurality of times so that a plurality of search partial minutiae templates are acquired by the fingerprint sensor to form a search total minutiae template. This search total minutiae template is used to match with the total registered minutiae template for verification.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 shows a fingerprint verification/identification system. Figure 2 shows two image blocks around a matched minutiae pair from two partial fingerprint templates.

Figure 3 shows the flow chart of synthesis of a minutiae template from two minutiae templates according to the invention.

Figure 4 shows an example of four partial fingerprint templates to be used in template synthesis to form a total minutiae template.

Figure 5a shows a search partial fingerprint ^• template and the minutiae template found. Figure 5b shows a total fingerprint template with the Area of Interest and the minutiae template found.

Figure 5c shows the minutiae location and the AOI of the partial fingerprint template which is aligned onto the AOI of the total fingerprint template.

Figure 6a shows an example of how the spatial information of a total minutiae template can be partitioned into different areas, and having each area assigned a numeric code.

Figure 6b shows another example of how the spatial information of a total minutiae template can be partitioned into different areas, and having each area assigned a numeric code. Figure 6c shows an example of how the spatial information of a total minutiae template can be partitioned into different areas, and having each area assigned an alphabetic code.

Figure 6d shows another example of how the spatial information of a total minutiae template can be partitioned into different areas, and having each area assigned an alphabetic code.

Figure 7 shows a region R defined on x and y axes. Figure 8 shows an example of a correlation table showing the PIN for a specific day in a week, and a time zone of a day.

DETAILED DESCRIPTION OF THE INVENTION Preferred embodiments of the invention will now be described with reference to the accompanying drawings.

Fig.l shows an overview of a fingerprint verification/identification system comprising a fingerprint sensor 102, a central processing unit (CPU) 103, a buffer 104, a storage media containing stored fingerprint templates 105, a comparator 106, a feedback unit 107, and a access control unit 108.

A fingerprint template of a finger 101 is acquired by a fingerprint sensor 102, and is stored in a buffer 104. During the synthesis of a minutiae template, a plurality of fingerprint templates corresponding to different portion of the same finger 101 are acquired, and their corresponding minutiae templates are extracted by the CPU 103. The plurality of minutiae templates are synthesized into a total minutiae template, and the synthesized total minutiae template is then stored in a storage media 105 of the fingerprint system, which is subsequently used for verification of a user during the verification phase. The matching of a search partial minutiae template with the total registered minutiae template is done in the comparator unit 106, and the result of the matching in the comparator unit 106 is fed to the feedback unit 107. The feedback unit 107 examines the output from the comparator unit 106. If the output of the comparator unit 106 indicates that the fingerprint template of the user matches with the registered template, a signal is sent to the access control unit 108 to allow access to a device or system controlled by the fingerprint verification/identification system. Otherwise, access to a device or system is not granted.

According to the invention, a minutia from a minutiae template of a first fingerprint template and a minutia from a minutiae template of a second fingerprint template are selected, wherein the minutiae template of the first fingerprint template and the second fingerprint template are to form a synthesized minutiae template using the method of alignment which will be described in detail.

Fig.2 shows the first fingerprint template 201 and the second fingerprint template 202 which are to be used to form the synthesized minutiae template. The image regions corresponding to the minutia in the two fingerprint templates

201, 202 are shown in the same Fig.2. The image regions 203, 204 corresponds to minutia ml 205 and m2 206 of the first fingerprint template 201 and the second fingerprint template

202, respectively.

The image pixel value is represented by I^m (i, j) and the coordinates (i ,j ) are contained in a region S around a minutia point m . In Fig.2, the region S corresponds to the image regions 203, 204 around the minutia ml 205 of the first fingerprint template 201 and around the minutia ^"m2 206 of the second fingerprint template 202, wherein ml 205 and m2 206 are considered to be a potentially matched minutiae pair. According to the preferred embodiment of the invention, a circular image region around the minutia is used for ease of rotation. A correlation between the image pixels of the two image regions is achieved by comparing the image pixel values of each image pixel within the two pre-defined circular image regions using a correlation function. When the result from the correlation function of the two image regions of the minutiae belonging to the first and second minutiae templates fulfils a prescribed criterion, the two minutiae are considered to be a matched minutiae pair.

An error function is used as the correlation function to determine the correlation of the minutiae pair based on the image pixel value of the image pixels around the minutiae pair:

wherein

H denotes the absolute distance.

To cater for any misalignment of the fingerprint images 201, 202 due to additional transformations like rotation, translation and deformation, the following formula is used:

E( i,m2) = 22|| r ^"./)- 2^m2(/ .y))| ₃ ( )<≡S ₍2) i i wherein

/(•)is a transformation function that relates the location, orientation and deformation between/,^ml(i, ;^') and I ²(i, j) .

When E(m\,m2) falls below a certain prescribed threshold value, the two minutiae ml and m2 are considered to be a matched minutiae pair.

The process described above is repeated for all the possible combinations of minutiae pairs from the first and the second minutiae templates to locate all the matched minutiae pairs. A list of all matched minutiae pairs is then generated.

In the preferred embodiment of the invention, the following linear transformation matrix is used as the transformation function f ( . ) :

XI = T. X2 (3) Wherein T is the linear transformation matrix, and XI is the transformed coordinate column vector of X2. X2 is the coordinate column vector of m2.

In a further preferred embodiment of the invention, the transformation matrix T in equation (3) is resolved into a rotation matrix R and a displacement vector d: cos <θ) -sin(6>)

R = (4 sin (θ) cos(θ)

dx - ; 5 ) ^d>'.

θ denotes the rotation angle, and dx and dy are the displacements in the x and y directions.

Therefore, the transformation for alignment can be written as:

XI R. X2 + d (6)

The computation of the error function (2) is performed for all combination of values of θ , dx and dy, or until the result of equation (2) falls below a pre-determined threshold value, indicating a matched minutiae pair.

The absolute orientation difference between the first minutiae template 201 and the second minutiae template 202 is assumed to be less than 30 degrees. To avoid heavy computation, the values of dx and dy of the search process is performed within a small pre-defined image region around the minutia, and the value of θ is performed in steps of 5 degrees. The predefined image region is preferably a circular region centred around the minutia with radius of a few pixels length.

The transformation function f ( . ) determined from a matched minutiae pair is used to align the first minutiae template onto the second minutiae template, forming a synthesized minutiae template.

After the synthesized minutiae template is formed, the whole synthesis process according to the invention is repeated for all the partial minutiae templates, until a total minutiae template is synthesized.

The method of minutiae template synthesis according to the invention may be incorporated into an existing fingerprint verification/identification system to increase the reliability of minutiae template synthesis from a plurality of partial minutiae templates acquired from a fingerprint sensor with a small sensor area. Fig.3 shows the flow chart of synthesis of a minutiae template of a fingerprint verification/identification system for a first and a second minutiae templates acquired from a fingerprint sensor with a small sensor area during the registration phase, wherein the synthesis of minutiae template according to the invention is incorporated.

A plurality of partial fingerprint templates are acquired from a user before the synthesis of the minutiae template. Fig.4 shows four partial fingerprint templates acquired from the user, which is to be used to form a total minutiae template. A corresponding minutiae template is extracted from each of the partial fingerprint templates acquired, preferably using method disclosed in^' [1]. In a first step of the synthesis process, a first minutiae template 301 derived from a first partial fingerprint template, preferably covering a portion of the center of the fingerprint of the user is selected. A second minutiae template 302 derived from a second fingerprint template, preferably having an area covering a region of the area of the first minutiae template 301, is selected.

The second minutiae template 302 is matched with the first minutiae template 301 in a matching unit 303 to determine at least one matched minutiae pair. The matching may be achieved using the methods disclosed in [3], [4], [5] or [6]. In the preferred embodiment of the invention, the matching is achieved using the method disclosed in [6]. In a further step (step 304), the number of matched minutiae pairs found between the first minutiae template 301 and the second minutiae template 302 from the matching unit 303 is compared with a pre-determined value. If the number of matched minutiae pair is below the pre-determined value, a searching unit 305 then searches for at least one matched minutiae pair from the first minutiae template 301 and the second minutiae template 302 which are not determined from the matching process in the matching unit 303, using the search process according to the invention. In the preferred embodiment, the pre-determined value is chosen to be 3.

A matched or corresponding minutiae pair means two minutiae having a match score as disclosed in [6] above a prescribed value, or a value determined from the error function defined in (2) below the prescribed value, indicating a high measure of similarity between the two minutiae. After the search process according to the invention in the searching unit 305 is completed, the total number of matched minutiae pairs determined in both the matching unit 303 and the searching unit 305 is compared with the pre-determined value (step 306) . If the total number of matched minutiae pairs is at least one but smaller than the pre-determined value, the transformation function determined in the searching unit 305 is used to synthesize a new minutiae template by aligning the second minutiae template 302 onto the first minutiae template 301 in the first formation unit 307.

If the total number of matched minutiae pairs is above the pre-determined value, a new minutiae template is synthesized using the method disclosed in [7] in the second formation unit 308.

If no matched minutiae pair is found between the first minutiae template 301 and the second minutiae template 302 in both the matching unit 303 and the searching unit 305, the two templates are considered unmatched, and no synthesis of a new minutiae template is performed.

The synthesis process described is repeated for a third minutiae template acquired from the user to synthesize a new minutiae template with the synthesized minutiae template which is the result of synthesizing the first and the second minutiae template. The process is repeated until all the minutiae templates acquired from the user are synthesized into a total minutiae template. This total minutiae template is called the registered total minutiae template, and is to be used to compare with a search fingerprint template of a user during the verification phase to verify the user.

During the verification phase, a partial minutiae template of a partial fingerprint template acquired from a sensor with a small sensor area is matched against a registered total minutiae template stored in the storage media of the fingerprint verification/identification system. A minutiae template acquired by the sensor during the verification phase is called a search minutiae template. The matching is done according to the method disclosed in [6] and the minutiae pairs determined between the search partial minutiae template and the registered total minutiae template with a degree of match or match score higher than a pre-determined value disclosed in [6] are considered to be matched. However, the minutiae count contained in the search partial minutiae template is usually significantly less than the minutiae count contained in the registered total minutiae template. Therefore, a match score for the minutiae pairs in the intersecting region of the two templates may not accurately indicate if a match is found for the minutiae pairs between the two templates. Therefore, a normalization of the match score obtained in the intersecting region of the two templates is used.

The normalization of the match score is performed by taking a weighted average using the following formula:

(7)

(/-l)²+(«2)²+(«l*Ai2)

wherein ms is the match score or the degree of match, nl is the minutiae count in the Area Of Interest of the search partial fingerprint template, and n2 is the minutiae count in the .Area Of Interest of the registered total fingerprint template area which intersects with the Area Of Interest of the search partial fingerprint template .

The Area of Interest (AOI) of a fingerprint template is represented by a continuous boundary or joined line segments covering the non-empty image area of the template. In the preferred embodiment of the invention, the AOI of a template is represented by joined line segments, and is defined according to the method disclosed in [7],

A search partial fingerprint template 501 in Fig.5a, is aligned onto the AOI of a registered total fingerprint template 502 in Fig.5b using a transformation function disclosed in [7]. The area of intersection 503 of the AOI belonging to the search partial fingerprint template 501 and the registered total fingerprint template 502 is shown in Fig.5c.

In a further preferred embodiment of the invention, the value of nl has to exceed a pre-determined threshold value to avoid a high value of the normalization of the match score when the information for matching is not sufficient.

The reliability of a fingerprint verification/identification system using a fingerprint sensor with a small sensor area can be increased by incorporating a fingerprint codeword system. According to the invention, the spatial information of the registered total minutiae template is partitioned into different portions, and each partitioned portions of the registered total minutiae template is assigned a numeric or an alphabetic positional code. A few examples of partitioning the spatial information of the registered total minutiae template are shown in Fig.6a, Fig.6b, Fig.6c and Fig.βd. A positional codeword consisting of numeric and/or alphabetic positional codes can be entered into the fingerprint verification/identification system by presenting the portion of the fingerprint corresponding to the positional code assigned to the partitioned portion of the registered total minutiae template.

With reference to Fig.7, the centroid of an AOI, as represented by region R 701 is computed according to the mathematical method disclosed in [16] using the formulae:

Cx = *l ₍₈₎

A

Cy . (9,

A wherein

M is the moment of the region R 701 about the y axis 702, Mx is the moment of the region R 701 about the x axis 703, A is the area of region R 701, and

R 701 shown is defined within x = a 708, x = b 709 on the x axis 702, and y = c 710, y = d 111 on the y axis 703.

The moments of region R 701 about the x and y axis 702, 703 are computed using the following formulae:

Mx = f y h(y0)dy (10)

wherein v (xθ) 704 is a segment along the y axis 703 in region R 701 at x = xO 706, h (yθ) 705 is a segment along the x axis 702 in region R 701 at y = y0 707. According to the preferred embodiment of the invention, the coordinates of the centroid is approximated using the following formulae:

Cx = — 1 Y Vx, (12)

wherein (Vxi , Vyi) , i=l , 2 ,..., N denotes the vertices defining the region R 701.

According to the invention, the centroid position of an AOI of a fingerprint template is computed using equations (12) and (13) to determine the positional codeword entered by the fingerprint of a user during verification. The coordinates of the centroid position of the AOI of the search partial fingerprint template is computed, and the corresponding coordinates of the centroid position of the AOI of the search fingerprint template on the registered total minutiae template is determined using the transformation function according to the method disclosed in [7]. In this way, the partitioned portion of the registered total minutiae template which the search partial minutiae template corresponds to can be determined and the positional code assigned to that portion of the registered total minutiae template is said to have been entered. It is to be noted that the fingerprint codeword system according to the invention is in contra-distinction with a Personal Identification Number (PIN) system. The fingerprint codeword system enters a codeword using different portions of a fingerprint, and the codeword in a PIN system is entered using a keypad.

With the fingerprint codeword system incorporated according to the invention, a user does not only need to have at least one portion of his fingerprint of the same finger matched with the registered total minutiae template during the verification process, but also to present the portion of his finger to the sensor which corresponds to the registered positional codeword in order for the whole authentication process to be successful .

For example, a user pre-registered a positional codeword '13' according to Fig.6b during the registration phase. During the verification phase, the user needs to present his fingerprint two times to the fingerprint sensor. Both the search partial minutiae templates of the fingerprint templates must match with the registered total minutiae template of the registered user. In addition, the centroid position of the first search minutiae template must correspond to the top center position on the portion of the registered total minutiae template which is assigned a positional code of ^Λl'. Similarly, the centroid position of the second search partial minutiae template must also correspond to the lower center position on the portion of the same registered total minutiae template which is assigned a positional code of ^Λ3' . Only if all the above conditions are satisfied, the authentication of the identify of the user is successful. In another embodiment of the invention, a plurality of devices are controlled by the fingerprint verification/identification system, and each device is assigned an unique positional codeword. The plurality of devices are activated by entering the pre-assigned positional codeword of the device using the fingerprint codeword system according to the invention. For example a codeword of ^Λll' is assigned to access a particular building, ^Λ22' to turn off the power of an equipment, or '99' to activate an alarm. The user is able to activate the alarm by presenting his fingerprint to the sensor which corresponds to a positional codeword of '99' in a seemingly normal verification process.

In another embodiment of the invention, the time duration a finger is presented on a fingerprint sensor, or a time-based codeword is used as an additional feature to increase the reliability of a fingerprint verification/identification system during the verification phase. When the finger is presented to the fingerprint sensor, the time duration of the finger presented to the sensor is recorded. A different time- based code is set for each positional code. Therefore, even if the correct positional codeword is entered using the fingerprint codeword system, the corresponding time-based codeword must also be entered correctly for the authentication process to be successful.

For example, a time-based code ^Λ l ' corresponding to a time duration of 1 second the finger of a user is presented to the fingerprint sensor and a time-based code of ^λ3' corresponding to a time duration of 3 seconds is set. For a pre-registered positional codeword of '13', a user not only have to enter a positional codeword of ^λ13' using the fingerprint codeword system, but also to present his finger to the sensor for a time duration of 1 second for the first fingerprint template and 3 seconds for the second fingerprint template corresponding to a time-based codeword of 13' for a successful authentication process.

In another embodiment of the invention, a unique positional codeword is assigned for a different day and/or a different time zone of the day. A user needs to know the positional codeword corresponding to the day and/or time zone of the day, and enters the positional codeword using the fingerprint codeword system for the verification process. In this embodiment, the security of the fingerprint system is enhanced, not only by matching the search partial minutiae templates of the search partial fingerprint templates with the registered total minutiae template, but also by entering a unique positional codeword which changes with a different day and time zone. For example, a correlation table showing the positional codeword for a day in a week and the time zone of the day is shown in Fig. 8. Tuesday at 3 pm will correspond to a positional codeword of 22' . A different set of positional codeword corresponding to a different day and/or may be used.

In another embodiment of the invention, the orientation angle in equation (4) of each of the partial fingerprint templates acquired by the fingerprint sensor during the registration phase is computed. During the verification phase, a matching process is performed to match the search partial minutiae template of the search partial fingerprint template with the registered total minutiae template. After a match is successful, the orientation angle of the AOI of the search partial fingerprint template with reference to the registered partial fingerprint template of the registered minutiae template is found for equation (.4). using the matched information. For the authentication process to be successful, the search partial minutiae template must match with a registered total minutiae template, and the orientation angle of the search partial fingerprint template must be within a pre-determined region with respect to the orientation angle of the registered partial fingerprint template. In this case, the reliability of the verification phase of the fingerprint verification/identification system is increased.

In another embodiment of the invention, a search total minutiae template is obtained from a plurality of search partial minutiae templates, and the search total minutiae template is matched with the registered total minutiae template during the verification phase using the method described in [7] . Further, a match score normalization may also be performed between the two total templates as the minutiae counts between the two templates may be significantly different. For higher security setting, a minimum number of minutiae count for each search total minutiae template may be pre-determined and the minimum pre-determined minutiae count should preferably cover sufficient information as that obtained from a rolled fingerprint as disclosed in [17].

It should be noted that the described embodiments of the invention may be implemented separately, or may be implemented in combination with the other embodiments in a fingerprint verification/identification system.

The described embodiments of the invention apply not only to the method but also to the device, the computer readable medium and the computer program. In this document, the following publications are cited:

[1] Xudong Jiang, Wei-Yun Yau and Wee Ser, Minutiae Extraction by Adaptive Tracing the Gray Level Ridge of the Fingerprint Image, International Conference on Image Processing, October 1999;

[2] Dario Maio and Davide Maltoni, Direct Gray-Scale Minutiae Detection in Fingerprints, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 19, No. 1, pp. 27 - 39, January 1997;

[3] US 4,135,147; [4] US 4,646,352:

[5] US 5,493,621; . .

[6] XD Jiang, WY Yau, "Fingerprint Minutiae Matching Based on the Local And Global Structures," 15^th International Conference on Pattern Recognition, Proc. ICPR 2000, Barcelona, Spain, Sep 2000.

[7] Wei Yun Yau, Kar Ann Toh and Xudong Jiang, "A method and a device for computer based determination of a total minutiae template from a plurality of partial minutiae templates and a computer readable medium"; 13 April 2000; Ref no: PCT/SG00/00048. [8] US 5,631,971; ("Vector Based Topological Fingerprint Matching", 1997)

[9] US 5,933,515; ("User Identification Through Sequential Input Of Fingerprints", 1999)

[10] US 6,229,906; ("Biometric Sequence Codes", 2001)

[11] US 2001/0007592A1, Patent Application Publication ("Biometric Combination Lock", 2001)

[12] US 5,557,686; ("Method And Apparatus For Verification Of A Computer User' s Identification, Based On Keystroke Characteristics", 1996) [13] US 2001/0004231 Al, Patent Application Publication

("Secure System Using Images Of Only Part Of A Body As The Key Where The Parts Has Continuously-Changing Features")

[14] US 6,219,639; ("Method And Apparatus For Recognizing Identity Of Individuals Employing Synchronized Biometrics", 2001) [15] R.C. Gonzalez & R.E. Woods, "Digital Image Processing", Addison-Wesley 1992.

[16] A. W. Goodman, "Analytic Geometry and the Calculus", Macmillian Publishing Co. Inc., Newyork, 1980.

[17] Nalini K. Ratha, Jonathan H. Connell and Ruud M. Bolle, "Image Mosaicing for Rolled Fingerprint Construction", 14th International Conference on Pattern Recognition, pp. 1651-1653, Vol. 2, 1998.

Claims

1. A method for synthesizing a minutiae template from a first minutiae template and a second minutiae template, which at least partially overlap each other, comprising the following steps: a) selecting a minutia from a plurality of minutiae in the first minutiae template, b) selecting a minutia from a plurality of minutiae in the second minutiae template, c) matching the minutia in the first minutiae template with the minutia in the second minutiae template by: i) defining an image region around the minutia in the first minutiae template, ii) defining an image region around the minutia in the second minutiae template, iii) determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, iv) determining the image pixel value from a plurality of corresponding image pixels in the -image region around the minutia in the second minutiae template, v) comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corresponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, and vi) determining the minutia in the first minutiae template to be matched with the minutia in the second minutiae template to the minutiae pair if the result of the comparison fulfils a prescribed criterion, d) repeating steps a) to c) for all minutiae in the first minutiae template and the second minutiae template, e) transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined when there is at least one matched minutiae, thereby forming a synthesized minutiae template.

2. The method according to claim 1, wherein an error function is used as the correlation function given by the following formula:

wherein ml is a minutia in the first minutiae template, m2 is a minutia in the second minutiae template,

/,""'(, j) is the image pixel value at coordinate (i,j) in the image region around ml,

^₂ ²(^ J) i-^{s tne} image pixel value at coordinate (i,j) in the image region around m2,

S is a set containing coordinate (i,j) defining the image region around minutia ml and m.2, respectively, and

/(.) is a transformation function relating the location, orientation and deformation between I™^l(i, j) and /™²(ι, j) .

The method according to claims 1 and 2, wherein a circular image region around the minutiae in the first minutiae template and the second minutiae template is used.

The method according to claim 2, wherein the transformation function /(.) is represented by a linear transformation function given by the formula:

XI = T X2 wherein

T is a linear transformation matrix,

X2 is the coordinate column vector of m2, and

XI is the transformed coordinate column vectors of X2. The method according to claim 4, wherein the transformation matrix T is resolved into a rotation matrix R and a displacement matrix given by the following formulae: cos(θ) -sin(θ)^' dx

R = sin(0) cos(0) dy wherein θ is the rotation angle, and dx and dy are the displacements in the x and y directions.

The method according to claims 4 and 5, wherein the transformation function f(.) obtained from a matched minutiae pair, is used to align the first minutiae template with the second minutiae template.

A method for synthesizing a minutiae template from a first minutiae template and a second minutiae template, which at least partially overlap each other, comprising the following steps: a) matching the minutiae from the first minutiae template with the minutiae of the second minutiae template, b) determining the number of matched minutiae pairs of a minutia from the first minutiae template and a minutia from the second minutiae template, c) searching for at least one further matched minutiae pair of a minutia from the first minutiae template and a minutia from the second minutiae template when the determined number of matched minutiae pairs is smaller than a pre-determined value, comprising the following steps :

I) selecting a minutia from a plurality of minutiae m the first minutiae template, ii) selecting a minutia from a plurality of minutiae in the second minutiae template, m) matching the minutia in the first minutiae template with th'e minutia in the second minutiae template by: a. defining an image region around the minutia in the first minutiae template, b. defining an image region around the minutia in the second minutiae template, c. determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, d. determining the image pixel value from a plurality of corresponding image pixels in the image region around the minutia in the second minutiae template, e. comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corresponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, and f . determining the minutiae in the first minutiae template to be matched with the minutiae in the second minutiae template to the minutiae pair if the result of the comparison fulfils a prescribed criterion,

IV) repeating steps l) to m) for all minutiae in the first minutiae template and the second minutiae template. d) determining a total number of matched minutiae pairs and further matched minutiae pairs, e) forming the synthesized minutiae template by transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined from the search process in c) , when there is at least one matched minutiae pair but the total number of matched minutiae pairs located by the matching process and the search process is less than- the pre-determined value, and f) forming the synthesized minutiae template from the first minutiae template and the second minutiae template using a transformation function determined from the matching process in a) , when the number of matched minutiae pairs located from the matching process, or the total number of matched minutiae pairs located by the matching process and the search process at least equals the pre-determined value.

8. The method according to claim 7, wherein the synthesized . minutiae template is used as a registered total minutiae template for verification of a user.

9. The method according to claim 8, wherein matching of a search partial minutiae template with the registered total minutiae template is performed by a normalization of a match score using the following formula: ms

(nl)² + (n2)² + (nl * n2)

wherein ms is the match score, nl is the minutia count in the AOI of a search partial fingerprint template, n2 is the minutia count in the AOI of the registered total fingerprint template which intersects with the AOI of the search partial fingerprint template.

10. The method according to claims 8 and 9, wherein the verification of the user further comprises a matching of a registered positional codeword with a positional codeword entered by at least one search partial minutiae template, comprising the following steps: a) partitioning the spatial information of the registered total minutiae template into at least one area, b) assigning a code to all the partitioned areas of the registered total minutiae template, c) computing the centroid position of a first search partial minutiae template when a match between the registered total minutiae template and the first search partial minutiae template is found, d) determining the coordinates of the centroid position of the first search partial minutiae template on the registered total minutiae template, e) determining the code assigned to the partitioned area on the registered total minutiae template which the coordinates of the centroid position of the first search partial minutiae template corresponds to, f) repeating steps c) to e) with a second search partial minutiae template replacing the first search partial minutiae template, until the codeword from all the search partial minutiae templates is formed, and g) comparing the codeword formed from the search partial minutiae templates with the registered codeword. h) determining the codeword entered by at least one of the search partial minutiae template to be matched with the registered codeword if a prescribed criterion is fulfilled.

11. The method according to claim 10, wherein the verification of the user further comprises the following steps: a) recording a time duration a finger is presented to a fingerprint sensor during verification of the user, b) comparing the recorded time duration with a predetermined time duration corresponding to the predetermined positional code the search partial minutiae template of the fingerprint of the finger corresponds to, and c) determining if the recorded time duration fulfils a prescribed criterion for verification of the user.

12. The method according to claim 10, wherein at least one unique positional codeword is used as the registered positional codeword for a specific day or time of the day.

13. The method according to claim 8, wherein the verification of the user comprises the following steps: a) computing the orientation angle of at least one search partial fingerprint template with reference to the registered partial fingerprint template, b) determining the user to be verified if the orientation angle of the search partial fingerprint template is within a pre-defined angle with respect to the registered fingerprint template.

14. The method according to claim 8, wherein a total minutiae template is synthesized from a plurality of search partial minutiae templates, and the synthesized total minutiae template is matched with the registered total minutiae template .

15. A device for synthesizing a minutiae template from a first minutiae template and a second minutiae template, which at least partially overlap each other, comprising: a) a first selection unit for selecting a minutia. from a plurality of minutiae in the first minutiae template, b) a second selection unit for selecting a minutia from a plurality of minutiae in the second minutiae template, c) a matching unit for matching the minutia in the first minutiae template with the minutia in the second minutiae template comprising: i) a first definition unit for defining an image region around the minutia in the first minutiae template, ii) a second definition unit for defining an image region around the minutia in the second minutiae template, iii) a first determination unit for determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, iv) a second determination unit for determining the image pixel value from a plurality of corresponding image pixels in the image region around the minutia in the second minutiae template, v) a comparison unit for comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corresponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, and vi) a third determination unit for determining the minutia in the first minutiae template to be matched with the minutia in the second minutiae template to the minutiae pair if the result of the comparison fulfils a prescribed criterion, d) a repetition unit for repeating steps a) to c) for all minutiae in the first minutiae template and the second minutiae template, e) a transformation unit for transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined when there is at least one matched minutiae, thereby forming a synthesized minutiae template.

16. A computer readable medium, having a program recorded thereon, where the program is to make the computer execute a procedure, comprising the following steps for synthesizing a minutiae template from a first minutiae template and a second minutiae template, which at least partially overlap each other, comprising the following steps : a) selecting a minutia from a plurality of minutiae in the first minutiae template, b) selecting a minutia from a plurality of minutiae in the second minutiae template, c) matching the minutia in the first minutiae template with the minutia in the second minutiae template by: i) defining an image region around the minutia in the first minutiae template, ii) defining an image region around the minutia in the second minutiae template, iii) determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, iv) determining the image pixel value from a plurality of corresponding image pixels in the image region around the minutia in the second minutiae template, v) comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corresponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, and vi) determining the minutia in the first minutiae template to be matched with the minutia in the second minutiae template to the minutiae pair if the result of the comparison fulfils a prescribed criterion, d) repeating steps a) to c) for all minutiae in the first minutiae template and the second minutiae template, e) transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined when there is at least one matched minutiae, thereby forming a synthesized minutiae template.

17. A computer program element which is to make the computer execute a procedure, comprising the following steps for synthesizing a minutiae template from a first minutiae template and a second minutiae template, which at least partially overlap each other, comprising the following steps : a) selecting a minutia from a plurality of minutiae in the first minutiae template, b) selecting a minutia from a plurality of minutiae in the second minutiae template, c) matching the minutia in the first minutiae template with the minutia in the second minutiae template by: i) defining an image region around the minutia in the first minutiae template, ii) defining an image region around the minutia in the second minutiae template, iii) determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, iv) determining the image pixel value from a plurality of corresponding image pixels in the image region around the minutia in the second minutiae template, v) comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corresponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, and vi) determining the minutia in the first minutiae template to be matched with the minutia in the second minutiae template to the minutiae pair if the result of the comparison fulfils a prescribed criterion, d) repeating steps a) to c) for all minutiae in the first minutiae template and the second minutiae template, e) transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined when there is at least one matched minutiae, thereby forming a synthesized minutiae template.

18. A device for synthesis of a minutiae template from a first minutiae template and a second minutiae template, which at least partially overlap each other, comprising: a) a first matching unit for matching the minutiae from the first minutiae template with the minutiae of the second minutiae template, b) a first determination unit for determining the number of matched minutiae pairs of a minutia from the first minutiae template and a minutia from the second minutiae template, c) a searching unit for searching at least one further matched minutiae pair of a minutia from the first minutiae template and a minutia from the second minutiae template when the determined number of matched minutiae pairs is smaller than a pre-determined value, comprising: i) a first selection unit for selecting a minutia from a plurality of minutiae in the first minutiae template, ii) a second selection unit for selecting a minutia from a plurality of minutiae in the second minutiae template, iii) a second matching unit for matching the minutia in the first minutiae template with the minutia in the second minutiae template comprising: a. a first defining unit for defining an image region around the minutia in the first minutiae template, b. a second definition unit for defining an image region around the minutia in the second minutiae template, c. a second determination unit for determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, d. a third determination unit for determining the image pixel value from a plurality of corresponding image pixels in the image region around the minutia in the second minutiae template, e. a comparison unit for comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corresponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, and f. a fourth determination unit for determining the minutia in the first minutiae template to be matched with the minutia in the second minutiae template to the minutiae pair if the result of the comparison fulfils a prescribed criterion, iv) a repetition unit for repeating steps i) to iii) for all minutiae in the first minutiae template and the second minutiae template. d) a fifth determination unit for determining a total number of matched minutiae pairs and further matched minutiae pairs, e) a first formation unit ror rormmg tne syntnesizeu minutiae template by transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined from -the search process in c) , when there is at least one matched minutiae pair but the total number of matched minutiae pairs located by the matching process and the search process is less than the pre-determined value, and f) a second formation unit for forming the synthesized minutiae template from the first minutiae template and the second minutiae template using a transformation function determined from the matching process in a) , when the number of matched minutiae pairs located from the matching process, or the total number of matched minutiae pairs located by the matching process and the search process at least equals the pre-determined value . 19. A computer readable medium, having a program recorded thereon, where the program is to make the computer execute a procedure, comprising the following steps for synthesis of a synthesized minutiae template from a first minutiae template and a second minutiae template, which at least partially overlap each other: a) matching the minutiae from the first minutiae template with the minutiae of the second minutiae template, b) determining the number of matched minutiae pairs of a minutia from the first minutiae template and a minutia from the second minutiae template, c) searching for at least one further matched minutiae pair of a minutia from the first minutiae template and a minutia from the second minutiae template when the determined number of matched minutiae pairs is smaller than a pre-determined value, comprising the following steps : i) selecting a minutia from a plurality of minutiae in the first minutiae template, ii) selecting a minutia from a plurality of minutiae in the second minutiae template, iii) matching the minutia in the first minutiae template with the minutia in the second minutiae template by: a. defining an image region around the minutia in the first minutiae template, b. defining an image region around the minutia in the second minutiae template, c. determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, d. determining the image pixel value from a plurality of corresponding image pixels in the image region around the minutia in the second minutiae template, e. comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corresponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, and f. determining the minutia in the first minutiae template to be matched with the minutia in the second minutiae template to the minutiae pair if the result of the comparison fulfils a prescribed criterion, iv) repeating steps i) to iii) for all minutiae in the first minutiae template and the second minutiae template . d) determining a total number of matched minutiae pairs and further matched minutiae pairs, e) forming the synthesized minutiae template by transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined from the search process in c) , when there is at least one matched minutiae pair but the total number of matched minutiae pairs located by the matching process and the search process is less than the pre-determined value, and f) forming the synthesized minutiae template from the first minutiae template and the second minutiae template using a transformation function determined from the matching process in a) , when the number of matched minutiae pairs located from the matching process, or the total number of matched minutiae pairs located by the matching process and the search process at least equals the pre-determined value.

20. A computer program element which is to make the computer execute a procedure, comprising the following steps for synthesis of a synthesized minutiae template from a first minutiae template and a second minutiae template, which at least partially overlap each other: a) matching the minutiae from the first minutiae template with the minutiae of the second minutiae template, b) determining the number of matched minutiae pairs of a minutia from the first minutiae template and a minutia from the second minutiae template, c) searching tor at le c uπe lurtner maucneu minuLia- pair of a minutia from the first minutiae template and a minutia from the second minutiae template when the determined number of matched minutiae pairs is smaller than a pre-determined value, comprising the following steps : i) selecting a minutia from a plurality of minutiae in the first minutiae template, ii) selecting a minutia from a plurality of minutiae in the second minutiae template, iii) matching the minutia in the first minutiae template with the minutia in the second minutiae template by: a. defining an image region around the minutia in the first minutiae template, b. defining an image region around the minutia in the second minutiae template, c. determining the image pixel value from a plurality of image pixels in the image region around the minutia in the first minutiae template, d. determining the image pixel value from a plurality of corresponding image pixels in the image region around the minutia in the second minutiae template, e. comparing the image pixel value of the image pixels around the minutia in the first minutiae template with the corresponding image pixel value of the image pixels around the minutia in the second minutiae template using a correlation function, and f . determining the minutia in the first minutiae template to be matched with the minutia in the second minutiae template to the minutiae pair if the result of the comparison fulfils a prescribed criterion, iv) repeating steps i) to iii) for all minutiae in the first minutiae template and the second minutiae template. d) determining a total number of matched minutiae pairs and further matched minutiae pairs, e) forming the synthesized minutiae template by transforming the coordinates of the first minutiae template onto the second minutiae template using a transformation function determined from the search process in c) , when there ^'is at least one matched minutiae pair but the total number of matched minutiae pairs located by the matching process and the search process is less than the pre-determined value, and forming the synthesized minutiae template from the first minutiae template and the second minutiae template using a transformation function determined from the matching process in a) , when the number of matched minutiae pairs located from the matching process, or the total number of matched minutiae pairs located by the matching process and the search process at least equals the pre-determined value.