CN1716274A - Finger print image splitting method based on direction information - Google Patents

Abstract

The fingerprint image splitting method based on direction information includes the following steps: determining the minimum selected pixel number Mmin each block based on the block size of the fingerprint image; calculating the directional variance of block based on the pixel numbers in eight directions; splitting fingerprint image; determining one splitting threshold Tr based on the selected pixel number inside each block; and determining the foreground area of the whole fingerprint image by means of adaptive image splitting threshold Tr. The fingerprint image splitting method has reduced fingerprint directional variance calculating time, and the developer can determine the image splitting threshold based on different fingerprint image blocks, resulting in greatly raised adaptability of fingerprint image splitting algorithm.

Description

A kind of fingerprint image dividing method based on directional information

Technical field

The invention belongs to technical field of image processing, particularly the fingerprint image treatment technology in the fingerprint identification technology.

Background technology

Biological identification technology is to carry out identity authentication according to a people's physiological characteristic.Owing to biological characteristic both can't pass into silence, also can not shared easily or falsely use to have, so biological identification technology more and more comes into one's own.As maturation and member the most easily in the biological identification technology, fingerprint identification technology has been successfully applied to fields of society.As: gate inhibition, attendance checking system, ecommerce, ATM Automatic Teller Machine and criminal's identity authentication system etc.In entry personnel's identity authentication system that the U.S. is formulating at present, also related to fingerprint identification technology in a large number.The automatic system of fingerprint recognition that relies on fingerprint identification technology foundation is as a kind of safe and reliable personal identification method, along with the downward modulation significantly of fingerprint acquisition instrument price and continuing to optimize of algorithm for recognizing fingerprint performance, will replace common personal identification methods such as password and instruction, the main recognition method that becomes personal identification is seen document: A.K.Jain and S.Pankanti, " Fingerprint Classification and Recognition ", The Imageand video Processing Handbook, A.Bovik (ed), Academic Press, April 2000; And document: Min Wu, Trappe, W., Wang.Z.J., Liu, K.J.R., " Collusion-Resistant Fingerprinting for Multimedia ", and Signal Processing Magazine, IEEE, Vol.21, pp.15-27,2004 is described.

In automatic system of fingerprint recognition, it is the technology of a key that fingerprint image is cut apart, and the purpose that fingerprint image is cut apart is to find out the foreground area that contains finger print information in fingerprint image, and discarded packets contains the background parts in territory, noise range.One of weakness of fingerprint identification technology is can find a large amount of false minutiae point in the background area of fingerprint image to algorithm for recognizing fingerprint, thereby has seriously influenced the quality of fingerprint recognition.With regard to present fingerprint image cutting techniques, accurately cutting apart fingerprint image according to the overall situation of fingerprint image or local gray level variance is the comparison difficulty.Referring to document Qinzhi Zhang, Kai Huang and HongYan, " Fingerprint Classification Based On Extraction and Analysis Of Singularities and Pseudoridges " Conferencein Research and Practice in Information Technology, Vol.11, pp 83-87,2002; . and document Asker M.Bazen and Sabih H.Gerez, " Segmentation of Fingerprint Images ", Proceedings of ProRISC 2001, Veldhoven, The Netherlands, November 2001.

And, might address this problem, because the streakline in the fingerprint image self just has tangible directivity based on the dividing method of fingerprint image orientation information.

The method of use fingerprint image orientation information commonly used has:

1. the method that adopts anisotropic filter to combine with the experience threshold values.Referring to document Lin Hong, Yi Fei Wan and Anil Jain " Fingerprint Image Enhancement:Algorithm and Performance Evaluation " IEEE Transactions on PAMI, Vol.20, No.8, pp.777-789, August 1998

2. adopt the method for the direction variance of fingerprint image gray-scale value.Referring to document A.K.Jain, L.Hong and R.Bolle, " On-lineFingerprint Verification ", IEEE Transactions on PAMI, Vol.19, No.4, pp.302-314,1997.

Above-mentioned two kinds of methods based on fingerprint image orientation information when fingerprint image is cut apart possess a common characteristic: the size of fingerprint image piecemeal is certain, and the image segmentation threshold values only is adapted to the fingerprint image piecemeal of fixed size.And this method that needs all picture element information in the statistics block, except image segmentation threshold values adaptability is not strong, also lacked the dirigibility on operation time.

Summary of the invention

Task of the present invention provides a kind of fingerprint image dividing method based on directional information, adopt method of the present invention, can reduce computing time widely, simultaneously can also be according to the size of fingerprint image piecemeal, generate suitable image segmentation threshold values, thereby can obtain segmentation effect preferably by the characteristic of fingerprint image orientation variance.

According to a kind of fingerprint image dividing method based on directional information provided by the invention, it comprises the following step:

Step 1, determine eight directions (as shown in Figure 1) of fingerprint image

The concrete angle of each direction is θ _i, θ _i=22.5 ° of * i (i=0,1,2,3,4,5,6,7)

Step 2, determine the picture element direction

Because fingerprint ridge line (being streakline part recessed among the human finger) has directivity clearly, and the fingerprint grey scale change on the fingerprint ridge line direction is very mild, just can determine the picture element direction according to the least error principle; Concrete grammar is:

On eight directions of fingerprint image, respectively get N adjacent image point point, calculate the average of the N that gets a picture element gray scale on each direction, then on each direction each picture element gray scale of getting all with its direction on gray average subtract each other and get the absolute value of difference, then the absolute value addition on each direction is sued for peace, judge that at last the direction that has minimum absolute value sum is the direction of this picture element, can be formulated as:

$K(i,j)=\mathrm{Min}\underset{l=1}{\overset{N}{\mathrm{\Σ}}}|C(i_m,j_m)-C^{\′}(i_m,j_m)|,$

(i j) is picture element p (i, direction j), the C (i that the capable j of i is listed as to K _m, j _m) (m=0,1,2...7) wherein be a N continuous picture element on the direction m, C ' (i _m, j _m) be each C (i on the m direction _m, j _m) (m=0,1, average 2...7).The implication of Min then be calculate m (m=0,1,2...7) in 8 of eight direction gained absolute value sums, select minimum absolute value sum; The pairing direction m of least absolute value sum is picture element p (i, direction j).Fig. 3 shown picture element p (i, j) with and continuous 5 picture elements on direction 3;

Step 3, fingerprint image is divided into a plurality of nonoverlapping (as shown in Figure 4), size is W * W, and W is the length of side of overlapping piecemeal not, divide nonoverlapping number to require to determine according to realistic accuracy;

It is characterized in that it also comprises following step:

The number of pixels that minimum should be selected in any one in step 4, the determining step 3 and the directional information of the enough picture elements of selected mesopodium

At first, according to the length of side W of the not overlapping piecemeal that is divided in the step 3, determine the number of pixels M that minimum should be selected _MinM _MinCan obtain by following formula

$M_{\min }=\left\{\begin{array}{cc}{(W/2)}^2& W=2k(k=\mathrm{4,5},...)\\ {[(W+1)/2)]}^2& W=2k+1(k=\mathrm{3,4},...)\end{array}\right.$

W is the length of side of overlapping piecemeal not; The number of pixels M that should select in the minimum of having determined to select in the piece _MinAfter, by getting a picture element,, get the mode of next picture element again and choose picture element then every a picture element; The picture element of choosing covers whole piecemeal (as shown in Figure 5) as much as possible, guarantees that the picture element of choosing is M _MinAnd can cover whole piecemeal as much as possible; In actual mechanical process, the number of pixels Num that selects in the piece should be more than or equal to M _MinThen, determine that according to the method described in the step 2 all choose the direction of picture element in the piece;

Direction variance in step 5, the determining step 3 in any middle piece

At first, according to all choose the direction of picture element in the piece of determining in the step 4, determine selected picture element on all fingerprint image orientations;

Then, according to the number of pixels on all (totally eight) directions, by the direction variance in the following formula calculation procedure 3 any:

$\stackrel{\‾}{P}(i,j)=(1/S)\underset{k=0}{\overset{7}{\mathrm{\Σ}}}{P}_k(i,j)$

$\mathrm{Var}(i,j)=(1/S)\underset{k=0}{\overset{7}{\mathrm{\Σ}}}{\left[P_k(i,j)\stackrel{\‾}{P}(i,j)\right]}^2$

Wherein S is eight direction sums of fingerprint image, P _k(i, j) be in the piece direction be k (k=0,1, the number of pixels 2...7), (i j) is the average of all the number of pixels on eight directions to P; (i j) is exactly direction variance according to the calculating gained of the number of pixels on all directions in the image block to Var;

Step 6, determine adapting to image partition threshold T _r:

By adapting to image partition threshold T _rDetermine display foreground zone and image background regions; Keep the display foreground zone that has than the general orientation variance, abandon the image background regions that has less direction variance;

Adapting to image partition threshold T _rCan obtain by following formula

T _r＝(7/64)*Num ²*β

Wherein Num is the number of pixels of selecting in the fingerprint image piecemeal, and β is an experience factor 0.008;

Step 7, determine region template R;

If the direction variance of image block is less than the adapting to image partition threshold T in the step 6 _r, the region template of then setting this piecemeal is 0, promptly abandons; If the direction variance of image block is greater than the adapting to image partition threshold T in the step 6 _r, the region template of then setting this piecemeal is 1, promptly keeps; Region template R can be obtained by following formula

Step 8, determine the foreground area of whole fingerprint image

Repeating step 4～step 7, multiplicity are the piece number that is divided in the step 3, just can obtain the foreground area of whole fingerprint image;

By above step, the fingerprint image after we just can obtain cutting apart.

Need to prove,

At the fringe region of fingerprint image, because the confidence level of picture element is not high, no matter (whether i is 0 j) to its region template R, all is regarded as the background area that abandon.All foreground area and background parts just can be by well separated in the fingerprint image like this;

In step 4, only need the directional information of picture element abundant in the statistics block, just can obtain to represent the directional information of whole fingerprint image piecemeal;

In step 5, in conjunction with minimum the number of pixels M that should select _MinAnd based on the direction variance Var of the number of pixels on all directions in the piece (i j), can reduce the calculated amount on asking for aspect the image block directional information;

In step 6, because image segmentation threshold values T _rCalculate by formula and to obtain, the number of pixels Num and the image block length of side W that choose in its size and the piece are relevant, so image segmentation threshold values T _rFor the number of pixels Num that chooses in the different pieces and the image block length of side W excellent adaptability is arranged all;

In step 7, according to the region template R of image block (i, j), if R (i, value j) is 0, thinks that then this image block is should abandoned background area in the fingerprint image, and the gray-scale value of all picture elements in the piece all is changed to 0; If (i, value j) is 1 to R, thinks that then this image block is the prospect part that should be retained in the fingerprint image, does not change the gray-scale value of any picture element in the piece.

Essence of the present invention: it is by finding out through the minimum the number of pixels that should select in any behind the piecemeal and the directional information of the enough picture elements of selected mesopodium; Determine that all choose direction, the direction variance of picture element in the piece, by adapting to image partition threshold T _rDetermine the foreground area of whole fingerprint image, the foreground area of described whole fingerprint image is exactly the fingerprint image after cutting apart.

Innovation part of the present invention is:

1. minimum the number of pixels M that should select has been proposed _Min, its size is relevant with the length of side W of fingerprint image piecemeal.Only need directional information, just can obtain to represent the directional information of whole fingerprint image piecemeal by picture element abundant in the statistics fingerprint image piecemeal.

2. proposed the direction variance based on the number of pixels on all directions in the piece, this is and the different notion of the traditional direction variance based on the fingerprint image gray-scale value (specifying referring to the list of references in 2.).In conjunction with minimum the number of pixels M that should select _MinWith direction variance based on the number of pixels on all directions in the piece, can reduce the calculated amount on asking for aspect the fingerprint image piecemeal directional information, thereby the operation time that has guaranteed this algorithm is than in the past based on the fingerprint image dividing method much less of directional information.

3. because image segmentation threshold values T _rCalculate by formula and to obtain, the number of pixels Num and the image block length of side W that choose in its size and the piece are relevant, so image segmentation threshold values T _rHas good adaptivity.

Existing fingerprint cutting techniques all can only be applicable to the image block of fixed size, has also lacked the dirigibility on operation time simultaneously.And adopt fingerprint image dividing method of the present invention, not only can reduce the required time of calculated fingerprint direction variance, and the developer can determine corresponding image segmentation threshold values according to different fingerprint image piecemeals, improved the adaptivity of fingerprint image partitioning algorithm greatly.

Description of drawings

Fig. 1 is employed eight directions in the algorithm

Wherein, the concrete angle of each direction is θ _i, θ _i=22.5 ° of * i (i=0,1,2,3,4,5,6,7), the 0th, angle is 0 direction, the 1st, angle is 22.5 ° a direction, the 2nd, and angle is 45 ° a direction, the 3rd, angle is 67.5 ° a direction, the 4th, angle is 90 ° a direction, the 5th, angle is 112.5 ° a direction, the 6th, and angle is 135 ° a direction, the 7th, angle is 157.5 ° a direction.

Fig. 2 is the fingerprint original image

Fig. 3 selects continuous 5 picture element synoptic diagram in any one piece

Wherein, picture element p (i, j) with and continuous 5 picture elements on direction 3;

Fig. 4 is the synoptic diagram of any one piecemeal of fingerprint image,

Wherein the length of side of piecemeal is W;

Fig. 5 is the picture element synoptic diagram of choosing in the image block

Wherein, W is the length of side of image block, and stain is the picture element of choosing, and number is M _Min

Fig. 6 is the fingerprint image synoptic diagram after the cutting

Fig. 7 is a FB(flow block) of the present invention

Embodiment:

Below to provide a concrete realization example of the present invention.

Need to prove: the parameter in the following example does not influence the generality of this patent.

1. determine eight directions of fingerprint image, the concrete angle of each direction is θ _i

θ _i＝22.5°*i(i＝0，1，2，3，4，5，6，7)；

2. fingerprint image is divided into 900 nonoverlapping, W is 10;

3. according to the size of piece, determine that the number of pixels that minimum should be selected is 25; '

4. by formula

$K(i,j)=\mathrm{Min}\underset{i=1}{\overset{L}{\mathrm{\Σ}}}|C(i_m,j_m)-C^{\′}(i_m,j_m)|$ Calculate arbitrary picture element of choosing in each image block

P (i, direction j).

5. the direction variance of each piecemeal of fingerprint image can be obtained by following formula

$\stackrel{\‾}{P}(i,j)=(1/S)\underset{k=0}{\overset{7}{\mathrm{\Σ}}}{P}_k(i,j)$

$\mathrm{Var}(i,j)=(1/S)\underset{k=0}{\overset{7}{\mathrm{\Σ}}}{[P_k(i,j)-\stackrel{\‾}{P}(i,j)]}^2;$

6. calculate adapting to image partition threshold T _rBe 8.2;

8. at last according to region template R, be partitioned into the prospect and the background area of fingerprint.

Claims (1)

1, a kind of fingerprint image dividing method based on directional information, it comprises the following step:

Step 1, determine eight directions of fingerprint image

The concrete angle of each direction is θ _i, θ _i=22.5 ° of * i (i=0,1,2,3,4,5,6,7)

Step 2, determine the picture element direction

Because fingerprint ridge line (being streakline part recessed among the human finger) has directivity clearly, and the fingerprint grey scale change on the fingerprint ridge line direction is very mild, just can determine the picture element direction according to the least error principle;

Concrete grammar is:

On eight directions of fingerprint image, respectively get N adjacent image point point, calculate the average of the N that gets a picture element gray scale on each direction, then on each direction each picture element gray scale of getting all with its direction on gray average subtract each other and get the absolute value of difference, then the absolute value addition on each direction is sued for peace, judge that at last the direction that has minimum absolute value sum is the direction of this picture element, can be formulated as:

$K(i,j)=\mathrm{Min}\underset{l=1}{\overset{N}{\mathrm{\Σ}}}|C(i_m,j_m)-C^{\′}(i_m,j_m)|,$

(i j) is picture element p (i, direction j), the C (i that the capable j of i is listed as to K _m, j _m) (m=0,1,2...7) wherein be a N continuous picture element on the direction m, C ' (i _m, j _m) be each C (i on the m direction _m, j _m) (m=0,1, average 2...7).The implication of Min then be calculate m (m=0,1,2...7) in 8 of eight direction gained absolute value sums, select minimum absolute value sum; The pairing direction m of least absolute value sum is picture element p (i, direction j).Fig. 3 shown picture element p (i, j) with and continuous 5 picture elements on direction 3;

Step 3, fingerprint image is divided into a plurality of nonoverlapping, size is W * W, and W is the length of side of overlapping piecemeal not, divide nonoverlapping number to require to determine according to realistic accuracy;

It is characterized in that it also comprises following step:

The number of pixels that minimum should be selected in any one in step 4, the determining step 3 and the directional information of the enough picture elements of selected mesopodium

At first, according to the length of side W of the not overlapping piecemeal that is divided in the step 3, determine the number of pixels M that minimum should be selected _MinM _MinCan obtain by following formula

$M_{\min }=\left\{\begin{array}{cc}{(W/2)}^2& W=2k(k=\mathrm{4,5},...)\\ {[(W+1)/2]}^2& W=2k+1(k=\mathrm{3,4},...)\end{array}\right.$

W is the length of side of overlapping piecemeal not; The number of pixels M that should select in the minimum of having determined to select in the piece _MinAfter, by getting a picture element,, get the mode of next picture element again and choose picture element then every a picture element; The picture element of choosing covers whole piecemeal as much as possible, guarantees that the picture element of choosing is M _MinAnd can cover whole piecemeal as much as possible; In actual mechanical process, the number of pixels Num that selects in the piece should be more than or equal to M _MinThen, determine that according to the method described in the step 2 all choose the direction of picture element in the piece;

Direction variance in step 5, the determining step 3 in any middle piece

At first, according to all choose the direction of picture element in the piece of determining in the step 4, determine selected picture element on all fingerprint image orientations;

Then, according to the number of pixels on all (totally eight) directions, by the direction variance in the following formula calculation procedure 3 any:

$\stackrel{\‾}{P}(i,j)=(1/S)\underset{k=0}{\overset{7}{\mathrm{\Σ}}}{P}_k(i,j)$

$\mathrm{Var}(i,j)=(1/S)\underset{k=0}{\overset{7}{\mathrm{\Σ}}}{[P_k(i,j)-\stackrel{\‾}{P}(i,j)]}^2$

Wherein S is eight direction sums of fingerprint image, P _k(i, j) be in the piece direction be k (k=0,1, the number of pixels 2...7), (i j) is the average of all the number of pixels on eight directions to P; (i j) is exactly direction variance according to the calculating gained of the number of pixels on all directions in the image block to Var;

Step 6, determine adapting to image partition threshold T _r:

By adapting to image partition threshold T _rDetermine display foreground zone and image background regions; Keep the display foreground zone that has than the general orientation variance, abandon the image background regions that has less direction variance;

Adapting to image partition threshold T _rCan obtain by following formula

T _r＝(7/64)*Num ²*β

Wherein Num is the number of pixels of selecting in the fingerprint image piecemeal, and β is an experience factor 0.008;

Step 7, determine region template R;

If the direction variance of image block is less than the adapting to image partition threshold T in the step 6 _r, the region template of then setting this piecemeal is 0, promptly abandons; If the direction variance of image block is greater than the adapting to image partition threshold T in the step 6 _r, the region template of then setting this piecemeal is 1, promptly keeps; Region template R can be obtained by following formula

Step 8, determine the foreground area of whole fingerprint image

Repeating step 4～step 7, multiplicity are the piece number that is divided in the step 3, just can obtain the foreground area of whole fingerprint image;

By above step, the fingerprint image after we just can obtain cutting apart.

Images