CN104834923A - Fingerprint image registering method based on global information - Google Patents

Abstract

The invention discloses a fingerprint image registering method based on global information, and aims at solving the problem that registration is inaccurate and the time complexity is high in the present fingerprint registering technology. The method comprises the steps that 1) a logging fingerprint image and a template fingerprint image are input, and the direction field and frequency field of two images are calculated; 2) initial transformation registering parameters are set, whether fingerprint image registering can be carried out is determined, and if yes, six groups of new transformation registering parameters in neighborhood of the initial transformation registering parameters are selected to implement fingerprint image registering; and 3) whether one optimal group of new transformation registering parameters exists in the six groups of transformation registering parameters is determined, if not, the initial transformation registering parameters serve as the final fingerprint image registering parameters, registering is completed, and otherwise, the optimal group of new transformation registering parameters serve as initial transformation registering parameters, and the step 2) is returned to till registering is completed. According to the method of the invention, fingerprint image registering can be carried out accurately, and the method can be applied to an automatic fingerprint recognition system.

Description

Based on the fingerprint image method for registering of global information

Technical field

The invention belongs to digital image processing techniques field, relate to fingerprint image registration, can be used for Automated Fingerprint Identification System.

Background technology

In Automated Fingerprint Identification System, fingerprint image registration is the very crucial step of fingerprint matching stage one, if registration is inaccurate, will propagate in matching result, cause inaccurate coupling mark, cause the reduction of discrimination further.In the fingerprint recognition system of the overwhelming majority, the fingerprint minutiae extracted is used to carry out registration and coupling.This method deposits deficiency both ways: on the one hand, due to fingerprint image noise and scar interference, is difficult to therefrom extract accurately and reliable minutiae point; On the other hand, the similarity of minutiae point partial structurtes information is utilized to go to find registration with reference to minutiae point pair, owing to being local message, exist unmatched with reference to minutiae point pair completely, their partial structurtes information similarity is very high, this will be used so mistakenly to carry out registration to minutiae point, cause the matching judgment of terminal error.Existing fingerprint image singular point method for registering, although registration speed is very fast and reliable, the fingerprint image collected differs to establish a capital and there is singular point, so just cannot carry out registration operation.

Summary of the invention

The object of the invention is the deficiency for above-mentioned prior art, propose a kind of fingerprint image method for registering based on global information, to improve the degree of accuracy of fingerprint image registration.

Technical scheme of the present invention is achieved in that

One. know-why

Fingerprint diretion and frequency fields are two very important information of fingerprint image, show the global information of fingerprint, macroscopically describe the basic configuration of fingerprint, structure and crestal line direction, ridge density information.Global information even can extract reliably from low-quality fingerprint image, and non-linear deformation is very little for the register effects of global information, the present invention utilizes these two global informations to carry out fingerprint image registration exactly, by accurate with the global information deallocation of fingerprint image, effectively solve based on minutiae point method for registering and the deficiency based on singular point method for registering.

Two. performing step

Fingerprint image method for registering of the present invention comprises the steps:

(1) field of direction O of typing fingerprint image Q (x, y) is obtained respectively _qthe field of direction O of (x, y) and template fingerprint image T (x, y) _t(x, y), wherein (x, y) represents the pixel coordinate of image;

(2) the frequency fields F of typing fingerprint image Q (x, y) is obtained respectively based on step (1) _qthe frequency fields F of (x, y) and template fingerprint image T (x, y) _t(x, y);

(3) initial transformation registration parameter is chosen:

(3a) region contacted with fingerprint acquisition instrument is pointed in definition is the foreground area of fingerprint image, and the part removing foreground area in definition fingerprint image is the background area of fingerprint image; Calculate the barycentric coordinates (i of typing fingerprint image Q (x, y) foreground area _q, j _q) and the barycentric coordinates (i of template fingerprint image T (x, y) foreground area _t, j _t);

(3b) the principal direction angle [alpha] of typing fingerprint image Q (x, y) foreground area is calculated respectively _qwith the principal direction angle [alpha] of template fingerprint image T (x, y) foreground area _t;

(3c) initial transformation registration parameter is obtained by the result of (3a)-(3b): anglec of rotation da ₀=α _t-α _q, the row translational movement dx of fingerprint image ₀=i _t-i _q, the row translational movement dy of fingerprint image ₀=j _t-j _q, wherein, da ₀> 0 represents that fingerprint image is pressed and counterclockwise rotates | da ₀|, da ₀=0 represents fingerprint image non rotating, da ₀< 0 represents that fingerprint image rotates in the direction of the clock | da ₀|, || represent ABS function;

(4) obtain the rear fingerprint image of conversion, determine whether to carry out fingerprint image registration:

(4a) with the center of typing fingerprint image Q (x, y) for rotation center, this typing fingerprint image is rotated | da ₀| angle, da ₀> 0 represents that fingerprint image is pressed and counterclockwise rotates, da ₀=0 represents fingerprint image non rotating, da ₀< 0 represents that fingerprint image rotates in the direction of the clock, translation | dx ₀| row and | dy ₀| row, dx ₀> 0 represents that fingerprint image follows the direction translation of increase, dx ₀=0 represents that fingerprint image does not do row translation, dx ₀< 0 represents that fingerprint image follows the direction translation of minimizing, dy ₀> 0 represents the direction translation that fingerprint image increases along row, dy ₀=0 represents that fingerprint image does not do row translation, dy ₀< 0 represents the direction translation that fingerprint image reduces along row, obtains fingerprint image Q ' (x, y) after converting;

(4b) the pixel number N of the overlapping region A of rear fingerprint image Q ' (x, y) foreground area and template fingerprint image T (x, y) foreground area and setting threshold value T will be converted _m=(N ₁× N ₂)/4 compare, wherein N ₁represent the height of typing fingerprint image Q (x, y), N ₂represent the wide of typing fingerprint image Q (x, y): if N < is T _m, then can not carry out fingerprint image registration, otherwise, can fingerprint image registration be carried out, perform step (5);

(5) initial transformation registration parameter da is found ₀, dx ₀, dy ₀conversion registration parameter in neighborhood, six groups that obtain fingerprint image newly convert registration parameter:

First group: anglec of rotation da ₁=da ₀, row translational movement dx ₁=dx ₀-sh, row translational movement dy ₁=dy ₀;

Second group: anglec of rotation da ₂=da ₀, row translational movement dx ₂=dx ₀+ sh, row translational movement dy ₂=dy ₀;

3rd group: anglec of rotation da ₃=da ₀, row translational movement dx ₃=dx ₀, row translational movement dy ₃=dy ₀-sh;

4th group: anglec of rotation da ₄=da ₀, row translational movement dx ₄=dx ₀, row translational movement dy ₄=dy ₀+ sh;

5th group: anglec of rotation da ₅=da ₀-sf, row translational movement dx ₅=dx ₀, row translational movement dy ₅=dy ₀;

6th group: anglec of rotation da ₆=da ₀+ sf, row translational movement dx ₆=dx ₀, row translational movement dy ₆=dy ₀;

Wherein, sf is the increment step-length of the anglec of rotation, and sh is translation increment step-length;

(6) in (5) i-th ₀organize new conversion registration parameter, calculate typing fingerprint image Q (x, y) and converted the new fingerprint image obtained by the new registration parameter of this group foreground area and template fingerprint image T (x, y) foreground area between diversity factor i ₀=1,2 ..., 6:

(6a) two different directions angle θ are defined ₁and θ ₂the required minimum angles function rotated that overlaps is carried out by being rotated counterclockwise: wherein θ ₁∈ [0, π], θ ₂∈ [0, π], θ ₁>=θ ₂, min is minimum value function;

(6b) by the result of (6a), calculated direction field error with frequency fields error

${\mathrm{es}}_{i_0}=\frac{1}{b}\underset{(x,y)\&Element;B_{i_0}}{\&Sum;}|f_T(x,y)-{f_{i_0}}^{\&prime;\&prime;}(x,y)|$

Wherein, represent the new fingerprint image after conversion overlapping region between foreground area and template fingerprint image T (x, y) foreground area, (x, y) represents the pixel coordinate of image, ot _t(x, y) is template fingerprint field of direction O _t(x, y) at the orientation angle at pixel (x, y) place, for the new fingerprint image after conversion the field of direction at the orientation angle at pixel (x, y) place, b is overlapping region pixel number, f _t(x, y) is template fingerprint frequency fields F _t(x, y) at the frequency values at pixel (x, y) place, for new fingerprint image frequency fields at the frequency values at pixel (x, y) place;

(6c) new fingerprint image is obtained according to the result of (6b) foreground area and template fingerprint image T (x, y) foreground area between diversity factor i ₀=1,2 ..., 6:

$D_{i_0}=w_1\&times;{\mathrm{er}}_{i_0}+w_2\&times;{\mathrm{es}}_{i_0}$

Wherein, w ₁and w ₂be the different weights constant of two numerical value, w ₁> 0, w ₂> 0, w ₁> w ₂;

(7) according to the method calculating diversity factor in (6), initial transformation registration parameter da is obtained ₀, dx ₀, dy ₀diversity factor D between the foreground area of the fingerprint image Q ' (x, y) under corresponding and template fingerprint image T (x, y) foreground area ₀;

(8) to judge in (5) six groups of new conversion registration parameters according to the result of (6) and (7), whether have best one group and convert registration parameter:

(8a) error amount of six global informations is calculated wherein, i ₀=1,2 ..., 6;

(8b) an error threshold T is set _e=0.001, if all or wherein max is max function, then there is not one group of best conversion registration parameter, export initial transformation registration parameter da ₀, dx ₀, dy ₀, complete fingerprint image registration, otherwise, get i ₀=1,2 ..., 6, wherein arg max is the value function of the independent variable corresponding when obtaining maximal value of dependent variable, kth s group conversion registration parameter da _ks, dx _ks, dy _ksas the initial transformation registration parameter in (5), i.e. da ₀=da _ks, dx ₀=dx _ks, dy ₀=dy _ks, turn back in (5), order performs, until complete fingerprint image registration.

Advantage of the present invention

1. the present invention obtains the principal direction angle of fingerprint image foreground area by principal component analysis, find the barycentric coordinates of fingerprint image foreground area simultaneously, obtain initial transformation registration parameter, carry out fingerprint image registration operation with this, improve precision and the registration speed of fingerprint image registration.

2. the present invention utilizes global information and fingerprint image orientation field and fingerprint image frequency fields to carry out fingerprint image registration operation, further increases fingerprint image registration accuracy.

Accompanying drawing explanation

Fig. 1 is realization flow figure of the present invention;

Fig. 2 is the principal direction angle figure of the fingerprint image foreground area with the present invention's acquisition;

Fig. 3 is with the result figure of the present invention to fingerprint image registration.

Embodiment

Be further described below in conjunction with accompanying drawing 1 pair of embodiments of the invention and effect.

Step 1. obtains typing fingerprint image Q (x, y) and template fingerprint image T (x, y).

The fingerprint image J of two width from same finger is extracted from fingerprint database ₁(x, y) and J ₂(x, y), secondary for wherein fingerprint image J ₁(x, y) as typing fingerprint image Q (x, y), another width fingerprint image J ₂(x, y), as template fingerprint image T (x, y), wherein (x, y) represents the pixel coordinate of image.

Step 2. obtains the field of direction O of typing fingerprint image Q (x, y) respectively _qthe field of direction O of (x, y) and template fingerprint image T (x, y) _t(x, y).

What the method that existing fingerprint image orientation field is extracted was commonly used has AVERAGE GRADIENT METHOD WITH, least mean-square estimate method and PCA analytic approach, and this example PCA analytic approach carries out the extraction of fingerprint image orientation field, and its step is as follows:

(2.1) field of direction O of typing fingerprint image Q (x, y) is obtained _q(x, y):

(2.1a) the auto-covariance matrix C of typing fingerprint image Q (x, y) at pixel (i, j) place is calculated ₁:

$C_1=\underset{(i,j)\&Element;w}{\&Sum;}\left[\begin{array}{cc}{G_{x1}}^2(i,j)& G_{x1}(i,j)\&times;G_{y1}(i,j)\\ G_{x1}(i,j)\&times;G_{y1}(i,j)& {G_{y1}}^2(i,j)\end{array}\right]$

Wherein, G _x1(i, j) for typing fingerprint image Q (x, y) is in the line direction gradient at pixel (i, j) place, G _y1(i, j) for typing fingerprint image Q (x, y) is in the column direction gradient at pixel (i, j) place, the rectangular window that w is length is 8 pixels, width is 8 pixels;

(2.1b) by auto-covariance matrix C that step (2.1a) obtains ₁minimal eigenvalue λ ₁corresponding proper vector coordinate form is labeled as vector starting point O coordinate be (0,0), vector terminal A ₁coordinate is (r _xx, r _yy), calculate typing fingerprint image Q (x, y) the orientation angle o at pixel (i, j) place ₁(i, j):

o ₁(i,j)＝arctan(r _yy/r _xx)

Wherein, arctan represents arctan function, r _xxterminal A ₁row coordinate on typing fingerprint image Q (x, y), r _yyterminal A ₁row-coordinate on typing fingerprint image Q (x, y);

(2.1c) by pixel, step (2.1a) and step (2.1b) are used to typing fingerprint image Q (x, y), obtain the field of direction O of typing fingerprint image Q (x, y) _q(x, y);

(2.2) field of direction O of template fingerprint image T (x, y) is obtained _t(x, y):

(2.2a) calculation template fingerprint image T (x, y) is at the auto-covariance matrix C at pixel (i, j) place ₂:

$C_2=\underset{(i,j)\&Element;w}{\&Sum;}\left[\begin{array}{cc}{G_{x2}}^2(i,j)& G_{x2}(i,j)\&times;G_{y2}(i,j)\\ G_{x2}(i,j)\&times;G_{y2}(i,j)& {G_{y2}}^2(i,j)\end{array}\right]$

Wherein, G _x2(i, j) for template fingerprint image T (x, y) is in the line direction gradient at pixel (i, j) place, G _y2(i, j) for template fingerprint image T (x, y) is in the column direction gradient at pixel (i, j) place, the rectangular window that w is length is 8 pixels, width is 8 pixels;

(2.2b) by auto-covariance matrix C that (2.2a) obtains ₂minimal eigenvalue λ ₂corresponding proper vector coordinate form is labeled as vector starting point O coordinate be (0,0), vector terminal A ₂coordinate is (r _ss, r _pp), calculation template fingerprint image T (x, y) is at the orientation angle o at pixel (i, j) place ₂(i, j):

o ₂(i,j)＝arctan(r _pp/r _ss)

Wherein, arctan represents arctan function, r _ssterminal A ₂row coordinate on template fingerprint image T (x, y), r _ppterminal A ₂row-coordinate on template fingerprint image T (x, y);

(2.2c) by pixel, step (2.2a) and step (2.2b) are used to template fingerprint image T (x, y), obtain the field of direction O of template fingerprint image T (x, y) _t(x, y).

Step 3. obtains the frequency fields F of typing fingerprint image Q (x, y) based on step 2 _q(x, y).

Conventional the having Zymography and add up rectangular window method of method that existing fingerprint image frequency fields is extracted, this example adopts statistics rectangular window method to carry out the extraction of fingerprint image frequency fields, and its step is as follows:

(3.1) the average M of typing fingerprint image Q (x, y) is calculated ₁with variance V ₁:

$M_1=\frac{1}{N_1\&times;N_2}\underset{i=0}{\overset{N_1-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_2-1}{\mathrm{\&Sigma;}}}{I}_1(i,j)$

$V_1=\frac{1}{N_2\&times;N_2}\underset{i=0}{\overset{N_1-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_2-1}{\mathrm{\&Sigma;}}}{(I_1(i,j)-M_1)}^2$

Wherein, N ₁the height of typing fingerprint image Q (x, y), N ₂the wide of typing fingerprint image Q (x, y), I ₁(i, j) is the gray-scale value of typing fingerprint image Q (x, y) at pixel (i, j) place;

(3.2) according to the result of step (3.1), typing fingerprint image Q (x, y) the gray-scale value G after the normalization of pixel (i, j) place is calculated ₁(i, j):

If I ₁(i, j) > M ₁, get $G_1(i,j)=H_1+\sqrt{\frac{R_1\&times;{(I_1(i,j)-M_1)}^2}{V_1}}$

If I ₁(i, j)≤M ₁, get $G_1(i,j)=H_1-\sqrt{\frac{R_1\&times;{(I_1(i,j)-M_1)}^2}{V_1}}$

Wherein, H ₁the equal numerical constant relevant with typing fingerprint image Q (x, y) contrast, R ₁the variance constant relevant with typing fingerprint image Q (x, y) contrast;

(3.3) by pixel, step (3.1) and step (3.2) are used to typing fingerprint image Q (x, y), obtain normalized image G ₁(x, y);

(3.4) by normalized image G ₁(x, y) is divided into n block, and the length of every block is W pixel, width is W pixel, to each image block PP _k, computed image block PP _kfrequency values f _k, k=1,2 ..., n, n represent image block sum:

(3.4a) rectangular window that a length is 2W pixel, width is W pixel is constructed, image block PP _kcentral pixel point (mc _k, nc _k) as the center of this rectangular window;

(3.4b) using the minor face place straight line of this rectangular window as y-axis, typing fingerprint image Q (x, y) at pixel (mc _k, nc _k) the orientation angle O (mc at place _k, nc _k) as the positive dirction of y-axis, using the direction orthogonal with y-axis as x-axis, x-axis is chosen 2W coordinate points, calculate the normalized image G of typing fingerprint image Q (x, y) ₁(x, y) gray average S at each coordinate points place along y-axis, is labeled as S successively by gray average S corresponding for all coordinate points ₀, S ₁..., S _t..., S _2W-1, wherein S _tas follows by following formula:

$S_t=\frac{1}{W}\underset{d=0}{\overset{W-1}{\mathrm{\&Sigma;}}}{G}_1(u,v),t=\mathrm{0,1},...,(2W-1)$

$u={\mathrm{mc}}_k+(d-\frac{W}{2})\&times;\cos \left(O({\mathrm{mc}}_k,{\mathrm{nc}}_k)\right)+(t-\frac{1}{2})\&times;\sin \left(O({\mathrm{mc}}_k,{\mathrm{nc}}_k)\right)$

$v=n{c}_k+(d-\frac{W}{2})\&times;\sin \left(O(m{c}_k,{\mathrm{nc}}_k)\right)+(\frac{1}{2}-t)\&times;\cos \left(O({\mathrm{mc}}_k,{\mathrm{nc}}_k)\right);$

Wherein, sin represents sine function, and cos represents cosine function, and d represents normalized image G ₁(x, y) along the pixel sequence numbering of y-axis, O (mc _k, nc _k) be the field of direction O of typing fingerprint image Q (x, y) _q(x, y) is at pixel (mc _k, nc _k) orientation angle at place.

(3.4c) by gray average S all in (3.4b) ₀, S ₁..., S _t..., S _2W-1regard one group of discrete digital signal as, use Fourier transform, find fundamental frequency F _k, i.e. image block PP for this reason _kfrequency values f _k=F _k;

(3.5) n the image block frequency values f obtained in step (3.4) ₁, f ₂..., f _nthe original frequency field picture of composition typing fingerprint image Q (x, y), carries out bilinear interpolation to original frequency field picture, obtains the frequency fields F of typing fingerprint image Q (x, y) _q(x, y).

Step 4. obtains the frequency fields F of template fingerprint image T (x, y) based on step 2 _t(x, y).

Conventional the having Zymography and add up rectangular window method of method that existing fingerprint image frequency fields is extracted, this example adopts statistics rectangular window method to carry out the extraction of fingerprint image frequency fields, and its step is as follows:

(4.1) the average M of calculation template fingerprint image T (x, y) ₂with variance V ₂:

$M_2=\frac{1}{N_3\&times;N_4}\underset{i=0}{\overset{N_3-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_4-1}{\mathrm{\&Sigma;}}}{I}_2(i,j)$

$V_2=\frac{1}{N_3\&times;N_4}\underset{i=0}{\overset{N_3-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_4-1}{\mathrm{\&Sigma;}}}{(I_2(i,j)-M_2)}^2$

Wherein, N ₃the height of template fingerprint image T (x, y), N ₄the wide of template fingerprint image T (x, y), I ₂(i, j) is the gray-scale value of template fingerprint image T (x, y) at pixel (i, j) place;

(4.2) according to the result of step (4.1), the gray-scale value G of calculation template fingerprint image T (x, y) after the normalization of pixel (i, j) place ₂(i, j):

If I ₂(i, j) > M ₂, get $G_2(i,j)=H_2+\sqrt{\frac{R_2\&times;{(I_2(i,j)-M_2)}^2}{V_2}}$

If I ₂(i, j)≤M ₂, get $G_2(i,j)=H_2-\sqrt{\frac{R_2\&times;{(I_2(i,j)-M_2)}^2}{V_2}}$

Wherein, H ₂the equal numerical constant relevant with template fingerprint image T (x, y) contrast, R ₂the variance constant relevant with template fingerprint image T (x, y) contrast;

(4.3) by pixel, step (4.1) and step (4.2) are used to template fingerprint image T (x, y), obtain normalized image G ₂(x, y);

(4.4) the normalized image G will obtained in step (4.3) ₂(x, y) is divided into n block, and the length of every block is W pixel, width is W pixel, to each image block DD _k, computed image block DD _kfrequency values h _k, k=1,2 ..., n, here image block DD _kfrequency values h _kcomputing method and the middle image block P of step (3.4) _kfrequency values f _kcomputing method are identical;

(4.5) n the image block frequency values h obtained in step (4.4) ₁, h ₂..., h _nthe original frequency field picture of composition template fingerprint image T (x, y), carries out bilinear interpolation to original frequency field picture, obtains the frequency fields F of template fingerprint image T (x, y) _t(x, y).

Step 5. chooses initial transformation registration parameter, determines whether to carry out fingerprint image registration based on fingerprint image after conversion.

It is the foreground area of fingerprint image that the region contacted with fingerprint acquisition instrument is pointed in definition, and the part removing foreground area in definition fingerprint image is the background area of fingerprint image.

(5.1) barycentric coordinates (i of typing fingerprint image Q (x, y) foreground area is calculated as follows _q, j _q):

$i_Q=\frac{1}{\mathrm{Hs}}\underset{k=1}{\overset{\mathrm{Hs}}{\mathrm{\&Sigma;}}}{a}_k,j_Q=\frac{1}{\mathrm{Hs}}\underset{k=1}{\overset{\mathrm{Hs}}{\mathrm{\&Sigma;}}}{b}_k$

Wherein, (a _k, b _k) be a kth pixel coordinate in typing fingerprint image Q (x, y) foreground area, H _sit is the pixel number of typing fingerprint image Q (x, y) foreground area;

(5.2) barycentric coordinates (i of template fingerprint image T (x, y) foreground area is calculated as follows _t, j _t):

$i_T=\frac{1}{\mathrm{Ms}}\underset{k=1}{\overset{\mathrm{Ms}}{\mathrm{\&Sigma;}}}{c}_k,j_T=\frac{1}{\mathrm{Ms}}\underset{k=1}{\overset{\mathrm{Ms}}{\mathrm{\&Sigma;}}}{d}_k$

Wherein (c _k, d _k) be a kth pixel coordinate in template fingerprint image T (x, y) foreground area, Ms is the pixel number of template fingerprint image T (x, y) foreground area;

(5.3) the principal direction angle [alpha] of typing fingerprint image Q (x, y) foreground area is calculated _q:

(5.3a) all pixel coordinates of typing fingerprint image Q (x, y) foreground area are formed a point set X ₁={ x ₁, x ₂..., x _i..., x _lS, obtain mean vector m ₁:

m ₁＝(x ₁+x ₂+...+x _i+...+x _LS)/LS

Wherein x _i=(t _i, g _i) ^t, t _ithe row-coordinate of typing fingerprint image Q (x, y) foreground area i-th pixel, g _ibe the row coordinate of typing fingerprint image Q (x, y) foreground area i-th pixel, LS is the number of typing fingerprint image Q (x, y) foreground area pixel;

(5.3b) according to the result of step (5.3a), structure covariance matrix s ₁:

s ₁＝{(x ₁-m ₁)×(x ₁-m ₁) ^T+(x ₂-m ₁)×(x ₂-m ₁) ^T+...+(x _i-m ₁)×(x _i-m ₁) ^T+...+(x _LS-m ₁)×(x _LS-m ₁) ^T}/LS

Wherein, T represents and carries out matrix transpose operation to matrix;

(5.3c) covariance matrix s is determined ₁projection matrix Z ₁:

To the covariance matrix s in (5.3b) ₁carry out svd: wherein, L ₁covariance matrix s ₁eigenwert composition diagonal matrix, T ₁by covariance matrix s ₁eigenwert character pair vector composition eigenmatrix, represent T ₁inverse matrix, T ₁, L ₁with the product of three is used represent, according to covariance matrix s ₁eigenwert order from big to small, reconfigure eigenmatrix T ₁column vector, form projection matrix Z ₁;

(5.3d) according to principal component analysis (PCA) principle, the principal direction vector U of typing fingerprint image Q (x, y) foreground area is obtained ₁, the transform component u on typing fingerprint image Q (x, y) vertical direction _ccthrough projection matrix Z ₁be 1 after projection, the transform component u in typing fingerprint image Q (x, y) horizontal direction _ddthrough projection matrix Z ₁be 0 after projection, use equation Z ₁× U ₁=(1,0) ^trepresent; Calculate principal direction vector U ₁with the angle of typing fingerprint image Q (x, y) horizontal direction: α _q=arctan (u _dd/ u _cc), arctan represents arctan function, using the principal direction angle of this angle as typing fingerprint image Q (x, y) foreground area.

(5.4) the principal direction angle [alpha] of calculation template fingerprint image T (x, y) foreground area _t:

(5.4a) all pixel coordinates of template fingerprint image T (x, y) foreground area are formed a point set Y ₁={ y ₁, y ₂..., y _i..., y _lD, y _i=(p _i, q _i) ^t, obtain mean vector m ₂:

m ₂＝(y ₁+y ₂+...+y _i+...+y _LD)/LD

Wherein p _ithe row-coordinate of template fingerprint image T (x, y) foreground area i-th pixel, q _ibe the row coordinate of template fingerprint image T (x, y) foreground area i-th pixel, LD is the number of template fingerprint image T (x, y) foreground area pixel;

(5.4b) according to the result of step (5.4a), structure covariance matrix s ₂:

s ₂＝{(y ₁-m ₂)×(y ₁-m ₂) ^T+(y ₂-m ₂)×(y ₂-m ₂) ^T+...+(y _i-m ₂)×(y _i-m ₂) ^T+...+(y _LD-m ₂)×(y _LD-m ₂) ^T}/LD

Wherein, T represents and carries out matrix transpose operation to matrix;

(5.4c) covariance matrix s is determined ₂projection matrix Z ₂:

To the covariance matrix s in (5.4b) ₂carry out svd: wherein, L ₂covariance matrix s ₂eigenwert composition diagonal matrix, T ₂by covariance matrix s ₂eigenwert character pair vector composition eigenmatrix, represent T ₂inverse matrix, T ₂, L ₂with the product of three is used represent, according to covariance matrix s ₂eigenwert order from big to small, reconfigure eigenmatrix T ₂column vector, form projection matrix Z ₂;

(5.4d) according to principal component analysis (PCA) principle, the principal direction vector U of template fingerprint image T (x, y) foreground area is obtained ₂, the transform component u on template fingerprint image T (x, y) vertical direction _aathrough projection matrix Z ₂be 1 after projection, the transform component u in template fingerprint image T (x, y) horizontal direction _bbthrough projection matrix Z ₂be 0 after projection, use equation Z ₂× U ₂=(1,0) ^trepresent; Calculate principal direction vector U ₂with the angle of template fingerprint image T (x, y) horizontal direction: α _t=arctan (u _bb/ u _aa), arctan represents arctan function, using the principal direction angle of this angle as template fingerprint image T (x, y) foreground area.

(5.5) following initial transformation registration parameter is obtained by the result of (5.1)-(5.4):

The anglec of rotation: da ₀=α _t-α _q,

The row translational movement of fingerprint image: dx ₀=i _t-i _q,

The row translational movement of fingerprint image: dy ₀=j _t-j _q,

Wherein, da ₀> 0 represents that fingerprint image is pressed and counterclockwise rotates | da ₀|, da ₀=0 represents fingerprint image non rotating, da ₀< 0 represents that fingerprint image rotates in the direction of the clock | da ₀|;

(5.6) with the center of typing fingerprint image Q (x, y) for rotation center, this typing fingerprint image is rotated | da ₀| angle, da ₀> 0 represents that fingerprint image is pressed and counterclockwise rotates, da ₀=0 represents fingerprint image non rotating, da ₀< 0 represents that fingerprint image rotates in the direction of the clock, translation | dx ₀| row and | dy ₀| row, dx ₀> 0 represents that fingerprint image follows the direction translation of increase, dx ₀=0 represents that fingerprint image does not do row translation, dx ₀< 0 represents that fingerprint image follows the direction translation of minimizing, dy ₀> 0 represents the direction translation that fingerprint image increases along row, dy ₀=0 represents that fingerprint image does not do row translation, dy ₀< 0 represents the direction translation that fingerprint image reduces along row, obtains fingerprint image Q ' (x, y) after converting;

(5.7) the pixel number N of the overlapping region A of rear fingerprint image Q ' (x, y) foreground area and template fingerprint image T (x, y) foreground area and setting threshold value T will be converted _m=(N ₁× N ₂)/4 compare, wherein N ₁represent the height of typing fingerprint image Q (x, y), N ₂represent the wide of typing fingerprint image Q (x, y): if N < is T _m, then can not carry out fingerprint image registration, otherwise, can fingerprint image registration be carried out, perform step 6;

Step 6. obtains initial transformation registration parameter da ₀, dx ₀, dy ₀six groups in neighborhood new conversion registration parameters:

First group: anglec of rotation da ₁=da ₀, row translational movement dx ₁=dx ₀-sh, row translational movement dy ₁=dy ₀;

Second group: anglec of rotation da ₂=da ₀, row translational movement dx ₂=dx ₀+ sh, row translational movement dy ₂=dy ₀;

3rd group: anglec of rotation da ₃=da ₀, row translational movement dx ₃=dx ₀, row translational movement dy ₃=dy ₀-sh;

4th group: anglec of rotation da ₄=da ₀, row translational movement dx ₄=dx ₀, row translational movement dy ₄=dy ₀+ sh;

5th group: anglec of rotation da ₅=da ₀-sf, row translational movement dx ₅=dx ₀, row translational movement dy ₅=dy ₀;

6th group: anglec of rotation da ₆=da ₀+ sf, row translational movement dx ₆=dx ₀, row translational movement dy ₆=dy ₀;

Wherein, sf is the increment step-length of the anglec of rotation, and sh is translation increment step-length.

In step 7. determination step 6 in six groups of new conversion registration parameters, whether there is one group of best conversion registration parameter.

(7.1) in step 6 i-th ₀organize new conversion registration parameter, calculate typing fingerprint image Q (x, y) and converted the new fingerprint image obtained by the new registration parameter of this group foreground area and template fingerprint image T (x, y) foreground area between diversity factor i ₀=1,2 ..., 6:

(7.1a) two different directions angle θ are defined ₁and θ ₂the required minimum angles function rotated that overlaps is carried out by being rotated counterclockwise: wherein θ ₁∈ [0, π], θ ₂∈ [0, π], θ ₁>=θ ₂, min is minimum value function;

(7.1b) by the result of (7.1a), calculated direction field error with frequency fields error

${\mathrm{es}}_{i_0}=\frac{1}{b}\underset{(x,y)\&Element;B_{i_0}}{\mathrm{\&Sigma;}}|f_T(x,y)-{f_{i_0}}^{\&prime;\&prime;}(x,y)|$

Wherein, represent the new fingerprint image after conversion overlapping region between foreground area and template fingerprint image T (x, y) foreground area, (x, y) represents the pixel coordinate of image, ot _t(x, y) is template fingerprint field of direction O _t(x, y) at the orientation angle at pixel (x, y) place, for the new fingerprint image after conversion the field of direction at the orientation angle at pixel (x, y) place, b is overlapping region pixel number, f _t(x, y) is template fingerprint frequency fields F _t(x, y) at the frequency values at pixel (x, y) place, for new fingerprint image frequency fields at the frequency values at pixel (x, y) place;

(7.1c) new fingerprint image is obtained according to the result of (7.1b) foreground area and template fingerprint image T (x, y) foreground area between diversity factor i ₀=1,2 ..., 6:

$D_{i_0}=w_1\&times;{\mathrm{er}}_{i_0}+w_2\&times;{\mathrm{es}}_{i_0}$

Wherein, w ₁and w ₂be the different weights constant of two numerical value, w ₁> 0, w ₂> 0, w ₁> w ₂;

(7.2) according to the method calculating diversity factor in step (7.1), initial transformation registration parameter da is obtained ₀, dx ₀, dy ₀diversity factor D between the foreground area of the fingerprint image Q ' (x, y) under corresponding and template fingerprint image T (x, y) foreground area ₀;

(7.3) error amount of six global informations is calculated wherein, i ₀=1,2 ..., 6;

(7.4) an error threshold T is set _e=0.001, if all or wherein max is max function, then there is not one group of best conversion registration parameter, export initial transformation registration parameter da ₀, dx ₀, dy ₀, complete fingerprint image registration, otherwise, get i ₀=1,2 ..., 6, wherein arg max is the value function of the independent variable corresponding when obtaining maximal value of dependent variable, kth s group conversion registration parameter da _ks, dx _ks, dy _ksas the initial transformation registration parameter in step 6, i.e. da ₀=da _ks, dx ₀=dx _ks, dy ₀=dy _ks, turn back in step 6, order performs, until complete fingerprint image registration.

Effect of the present invention further illustrates by following emulation:

1 simulated conditions

Emulate under the MATLAB.R2013b environment of PC, PC configuration Core I7 processor, dominant frequency is 3.4-GHz.Emulation fingerprint image comes from FVC2002DB1 database, and fingerprint image size is 374 pixel × 388 pixels, and FVC2002DB1 database is one of internationally recognized fingerprint recognition database.

2. emulate content and interpretation of result

Emulation 1, the principal direction angle of fingerprint image foreground area is obtained by method of the present invention, as shown in Figure 2, wherein Fig. 2 (a) is fingerprint image 13_4.GIF, Fig. 2 (b) is the foreground area figure of fingerprint image 13_4.GIF, the foreground area principal direction angle schematic diagram of fingerprint image 13_4.GIF of Fig. 2 (c) for obtaining by method of the present invention, by the foreground area principal direction of arrow mark fingerprint image 13_4.GIF.

Emulation 2, by method of the present invention, registration emulation experiment is carried out to pair fingerprint image of three in FVC2002DB1 database, result as Fig. 3, wherein:

Fig. 3 (a) utilizes method of the present invention that typing fingerprint image 9_3.GIF and template fingerprint image 9_4.GIF is carried out the result figure after registration.

Fig. 3 (b) utilizes method of the present invention that typing fingerprint image 13_4.GIF and template fingerprint image 13_6.GIF is carried out the result figure after registration.

Fig. 3 (c) utilizes method of the present invention that typing fingerprint image 45_4.GIF and template fingerprint image 45_2.GIF is carried out the result figure after registration.

Fig. 3 shows, the present invention can carry out fingerprint image registration accurately.

Fingerprint image method for registering based on global information of the present invention is not limited in the description in instructions.Within the spirit and principles in the present invention all, any amendment made, equal replacement, improvement etc., be all included within right of the present invention.

Claims (9)

1., based on a fingerprint image method for registering for global information, comprise the following steps:

(1) field of direction O of typing fingerprint image Q (x, y) is obtained respectively _qthe field of direction O of (x, y) and template fingerprint image T (x, y) _t(x, y), wherein (x, y) represents the pixel coordinate of image;

(2) the frequency fields F of typing fingerprint image Q (x, y) is obtained respectively based on step (1) _qthe frequency fields F of (x, y) and template fingerprint image T (x, y) _t(x, y);

(3) initial transformation registration parameter is chosen:

(3a) region contacted with fingerprint acquisition instrument is pointed in definition is the foreground area of fingerprint image, and the part removing foreground area in definition fingerprint image is the background area of fingerprint image; Calculate the barycentric coordinates (i of typing fingerprint image Q (x, y) foreground area _q, j _q) and the barycentric coordinates (i of template fingerprint image T (x, y) foreground area _t, j _t);

(3b) the principal direction angle [alpha] of typing fingerprint image Q (x, y) foreground area is calculated respectively _qwith the principal direction angle [alpha] of template fingerprint image T (x, y) foreground area _t;

(3c) initial transformation registration parameter is obtained by the result of (3a)-(3b): anglec of rotation da ₀=α _t-α _q, the row translational movement dx of fingerprint image ₀=i _t-i _q, the row translational movement dy of fingerprint image ₀=j _t-j _q, wherein, da ₀> 0 represents that fingerprint image is pressed and counterclockwise rotates | da ₀|, da ₀=0 represents fingerprint image non rotating, da ₀< 0 represents that fingerprint image rotates in the direction of the clock | da ₀|, || represent ABS function;

(4) obtain the rear fingerprint image of conversion, determine whether to carry out fingerprint image registration:

(4a) with the center of typing fingerprint image Q (x, y) for rotation center, this typing fingerprint image is rotated | da ₀| angle, da ₀> 0 represents that fingerprint image is pressed and counterclockwise rotates, da ₀=0 represents fingerprint image non rotating, da ₀< 0 represents that fingerprint image rotates in the direction of the clock, translation | dx ₀| row and | dy ₀| row, dx ₀> 0 represents that fingerprint image follows the direction translation of increase, dx ₀=0 represents that fingerprint image does not do row translation, dx ₀< 0 represents that fingerprint image follows the direction translation of minimizing, dy ₀> 0 represents the direction translation that fingerprint image increases along row, dy ₀=0 represents that fingerprint image does not do row translation, dy ₀< 0 represents the direction translation that fingerprint image reduces along row, obtains fingerprint image Q ' (x, y) after converting;

(4b) the pixel number N of the overlapping region A of rear fingerprint image Q ' (x, y) foreground area and template fingerprint image T (x, y) foreground area and setting threshold value T will be converted _m=(N ₁× N ₂)/4 compare, wherein N ₁represent the height of typing fingerprint image Q (x, y), N ₂represent the wide of typing fingerprint image Q (x, y): if N < is T _m, then can not carry out fingerprint image registration, otherwise, can fingerprint image registration be carried out, perform step (5);

(5) initial transformation registration parameter da is found ₀, dx ₀, dy ₀conversion registration parameter in neighborhood, six groups that obtain fingerprint image newly convert registration parameter:

First group: anglec of rotation da ₁=da ₀, row translational movement dx ₁=dx ₀-sh, row translational movement dy ₁=dy ₀;

Second group: anglec of rotation da ₂=da ₀, row translational movement dx ₂=dx ₀+ sh, row translational movement dy ₂=dy ₀;

3rd group: anglec of rotation da ₃=da ₀, row translational movement dx ₃=dx ₀, row translational movement dy ₃=dy ₀-sh;

4th group: anglec of rotation da ₄=da ₀, row translational movement dx ₄=dx ₀, row translational movement dy ₄=dy ₀+ sh;

5th group: anglec of rotation da ₅=da ₀-sf, row translational movement dx ₅=dx ₀, row translational movement dy ₅=dy ₀;

6th group: anglec of rotation da ₆=da ₀+ sf, row translational movement dx ₆=dx ₀, row translational movement dy ₆=dy ₀;

Wherein, sf is the increment step-length of the anglec of rotation, and sh is translation increment step-length;

(6) in (5) i-th ₀organize new conversion registration parameter, calculate typing fingerprint image Q (x, y) and converted the new fingerprint image obtained by the new registration parameter of this group foreground area and template fingerprint image T (x, y) foreground area between diversity factor i ₀=1,2 ..., 6:

(6a) two different directions angle θ are defined ₁and θ ₂the required minimum angles function rotated that overlaps is carried out by being rotated counterclockwise: wherein θ ₁∈ [0, π], θ ₂∈ [0, π], θ ₁>=θ ₂, min is minimum value function;

(6b) by the result of (6a), calculated direction field error with frequency fields error

${\mathrm{es}}_{i_0}=\frac{1}{b}\underset{(x,y)\&Element;B_{i_0}}{\mathrm{\&Sigma;}}|f_T(x,y)-{f_{i_0}}^{\&prime;\&prime;}(x,y)|$

Wherein, represent the new fingerprint image after conversion overlapping region between foreground area and template fingerprint image T (x, y) foreground area, (x, y) represents the pixel coordinate of image, ot _t(x, y) is template fingerprint field of direction O _t(x, y) at the orientation angle at pixel (x, y) place, for the new fingerprint image after conversion the field of direction at the orientation angle at pixel (x, y) place, b is overlapping region pixel number, f _t(x, y) is template fingerprint frequency fields F _t(x, y) at the frequency values at pixel (x, y) place, for new fingerprint image frequency fields at the frequency values at pixel (x, y) place;

(6c) new fingerprint image is obtained according to the result of (6b) foreground area and template fingerprint image T (x, y) foreground area between diversity factor i ₀=1,2 ..., 6:

$D_{i_0}=w_1\&times;{\mathrm{er}}_{i_0}+w_2\&times;{\mathrm{es}}_{i_0}$

Wherein, w ₁and w ₂be the different weights constant of two numerical value, w ₁> 0, w ₂> 0, w ₁> w ₂;

(7) according to the method calculating diversity factor in (6), initial transformation registration parameter da is obtained ₀, dx ₀, dy ₀diversity factor D between the foreground area of the fingerprint image Q ' (x, y) under corresponding and template fingerprint image T (x, y) foreground area ₀;

(8) to judge in (5) six groups of new conversion registration parameters according to the result of (6) and (7), whether have best one group and convert registration parameter:

(8a) error amount of six global informations is calculated wherein, i ₀=1,2 ..., 6;

(8b) an error threshold T is set _e=0.001, if all or wherein max is max function, then there is not one group of best conversion registration parameter, export initial transformation registration parameter da ₀, dx ₀, dy ₀, complete fingerprint image registration, otherwise, get i ₀=1,2 ..., 6, wherein arg max is the value function of the independent variable corresponding when obtaining maximal value of dependent variable, kth s group conversion registration parameter da _ks, dx _ks, dy _ksas the initial transformation registration parameter in (5), i.e. da ₀=da _ks, dx ₀=dx _ks, dy ₀=dy _ks, turn back in (5), order performs, until complete fingerprint image registration.

2. the fingerprint image method for registering based on global information according to claim 1, obtains the field of direction O of typing fingerprint image Q (x, y) in wherein said step (1) _q(x, y), carry out as follows:

(1.1) typing fingerprint image Q (x, y) the auto-covariance matrix C at pixel (i, j) place is calculated ₁:

$G_1=\underset{(i,j)\&Element;w}{\mathrm{\&Sigma;}}\left[\begin{array}{cc}{G_{x1}}^2(i,j)& G_{x1}(i,j)\&times;G_{y1}(i,j)\\ G_{x1}(i,j)\&times;G_{y1}(i,j)& {G_{y1}}^2(i,j)\end{array}\right]$

Wherein, G _x1(i, j) for typing fingerprint image Q (x, y) is in the line direction gradient at pixel (i, j) place, G _y1(i, j) for typing fingerprint image Q (x, y) is in the column direction gradient at pixel (i, j) place, the rectangular window that w is length is 8 pixels, width is 8 pixels;

(1.2) by auto-covariance matrix C that (1.1) obtain ₁minimal eigenvalue λ ₁corresponding proper vector coordinate form is labeled as vector starting point O coordinate be (0,0), vector terminal A ₁coordinate is (r _xx, r _yy), calculate typing fingerprint image Q (x, y) the orientation angle o at pixel (i, j) place ₁(i, j):

o ₁(i,j)＝arctan(r _yy/r _xx)

Wherein, arctan represents arctan function, r _xxterminal A ₁row coordinate on typing fingerprint image Q (x, y), r _yyterminal A ₁row-coordinate on typing fingerprint image Q (x, y);

(1.3) by pixel, step (1.1) and step (1.2) are used to typing fingerprint image Q (x, y), obtain the field of direction O of typing fingerprint image Q (x, y) _q(x, y).

3. the fingerprint image method for registering based on global information according to claim 1, obtains the field of direction O of template fingerprint image T (x, y) in wherein said step (1) _t(x, y), carry out as follows:

(1.4) calculation template fingerprint image T (x, y) is at the auto-covariance matrix C at pixel (i, j) place ₂:

$G_2=\underset{(i,j)\&Element;w}{\mathrm{\&Sigma;}}\left[\begin{array}{cc}{G_{x2}}^2(i,j)& G_{x2}(i,j)\&times;G_{y2}(i,j)\\ G_{x2}(i,j)\&times;G_{y2}(i,j)& {G_{y2}}^2(i,j)\end{array}\right]$

Wherein, G _x2(i, j) for template fingerprint image T (x, y) is in the line direction gradient at pixel (i, j) place, G _y2(i, j) for template fingerprint image T (x, y) is in the column direction gradient at pixel (i, j) place, the rectangular window that w is length is 8 pixels, width is 8 pixels;

(1.5) by auto-covariance matrix C that (1.4) obtain ₂minimal eigenvalue λ ₂corresponding proper vector coordinate form is labeled as vector starting point O coordinate be (0,0), vector terminal A ₂coordinate is (r _ss, r _pp), calculation template fingerprint image T (x, y) is at the orientation angle o at pixel (i, j) place ₂(i, j):

o ₂(i,j)＝arctan(r _pp/r _ss)

Wherein, arctan represents arctan function, r _ssterminal A ₂row coordinate on template fingerprint image T (x, y), r _ppterminal A ₂row-coordinate on template fingerprint image T (x, y);

(1.6) by pixel, step (1.4) and step (1.5) are used to template fingerprint image T (x, y), obtain the field of direction O of template fingerprint image T (x, y) _t(x, y).

4. the fingerprint image method for registering based on global information according to claim 1, obtains the frequency fields F of typing fingerprint image Q (x, y) in wherein said step (2) _q(x, y), carry out as follows:

(2.1) the average M of typing fingerprint image Q (x, y) is calculated ₁with variance V ₁:

$M_1=\frac{1}{N_1\&times;N_2}\underset{i=0}{\overset{N_1-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_2-1}{\mathrm{\&Sigma;}}}{I}_1(i,j)$

$V_1=\frac{1}{N_1\&times;N_2}\underset{i=0}{\overset{N_1-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_2-1}{\mathrm{\&Sigma;}}}{(I_1(i,j)-M_1)}^2$

Wherein, N ₁the height of typing fingerprint image Q (x, y), N ₂the wide of typing fingerprint image Q (x, y), I ₁(i, j) is the gray-scale value of typing fingerprint image Q (x, y) at pixel (i, j) place;

(2.2) according to the result of step (2.1), typing fingerprint image Q (x, y) the gray-scale value G after the normalization of pixel (i, j) place is calculated ₁(i, j):

If I ₁(i, j) > M ₁, get $G_1(i,j)=H_1+\sqrt{\frac{R_1\&times;{(I_1(i,j)-M_1)}^2}{V_1}}$

If I ₁(i, j)≤M ₁, get $G_1(i,j)=H_1-\sqrt{\frac{R_1\&times;{(I_1(i,j)-M_1)}^2}{V_1}}$

Wherein, H ₁the equal numerical constant relevant with typing fingerprint image Q (x, y) contrast, R ₁the variance constant relevant with typing fingerprint image Q (x, y) contrast;

(2.3) by pixel, step (2.1) and step (2.2) are used to typing fingerprint image Q (x, y), obtain normalized image G ₁(x, y);

(2.4) the normalized image G will obtained in step (2.3) ₁(x, y) is divided into n block, and the length of every block is W pixel, width is W pixel, to each image block PP _k, computed image block PP _kfrequency values f _k, k=1,2 ..., n, n represent image block sum;

(2.5) n the image block frequency values f obtained in step (2.4) ₁, f ₂..., f _nthe original frequency field picture of composition typing fingerprint image Q (x, y), carries out bilinear interpolation to original frequency field picture, obtains the frequency fields F of typing fingerprint image Q (x, y) _q(x, y).

5. the fingerprint image method for registering based on global information according to claim 1, obtains the frequency fields F of template fingerprint image T (x, y) in wherein said step (2) _t(x, y), carry out as follows:

(2.6) the average M of calculation template fingerprint image T (x, y) ₂with variance V ₂:

$M_2=\frac{1}{N_3\&times;N_4}\underset{i=0}{\overset{N_3-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_4-1}{\mathrm{\&Sigma;}}}{I}_2(i,j)$

$V_2=\frac{1}{N_3\&times;N_4}\underset{i=0}{\overset{N_3-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_4-1}{\mathrm{\&Sigma;}}}{(I_2(i,j)-M_2)}^2$

Wherein, N ₃the height of template fingerprint image T (x, y), N ₄the wide of template fingerprint image T (x, y), I ₂(i, j) is the gray-scale value of template fingerprint image T (x, y) at pixel (i, j) place;

(2.7) according to the result of step (2.6), the gray-scale value G of calculation template fingerprint image T (x, y) after the normalization of pixel (i, j) place ₂(i, j):

If I ₂(i, j) > M ₂, get $G_2(i,j)=H_2+\sqrt{\frac{R_2\&times;{(I_2(i,j)-M_2)}^2}{V_2}}$

If I ₂(i, j)≤M ₂, get $G_2(i,j)=H_2-\sqrt{\frac{R_2\&times;{(I_2(i,j)-M_2)}^2}{V_2}}$

Wherein, H ₂the equal numerical constant relevant with template fingerprint image T (x, y) contrast, R ₂the variance constant relevant with template fingerprint image T (x, y) contrast;

(2.8) by pixel, step (2.6) and step (2.7) are used to template fingerprint image T (x, y), obtain normalized image G ₂(x, y);

(2.9) the normalized image G will obtained in step (2.8) ₂(x, y) is divided into n block, and the length of every block is W pixel, width is W pixel, to each image block DD _k, computed image block DD _kfrequency values h _k, k=1,2 ..., n;

(2.10) n the image block frequency values h obtained in step (2.9) ₁, h ₂..., h _nthe original frequency field picture of composition template fingerprint image T (x, y), carries out bilinear interpolation to original frequency field picture, obtains the frequency fields F of template fingerprint image T (x, y) _t(x, y).

6. method according to claim 4, computed image block PP in wherein said step (2.4) _kfrequency values f _k, carry out as follows:

(2.4a) rectangular window that a length is 2W pixel, width is W pixel is constructed, image block PP _kcentral pixel point (mc _k, nc _k) as the center of this rectangular window;

(2.4b) using the minor face place straight line of this rectangular window as y-axis, typing fingerprint image Q (x, y) at pixel (mc _k, nc _k) the orientation angle O (mc at place _k, nc _k) as the positive dirction of y-axis, using the direction orthogonal with y-axis as x-axis, x-axis is chosen 2W coordinate points, calculate the normalized image G of typing fingerprint image Q (x, y) ₁(x, y) gray average S at each coordinate points place along y-axis, is labeled as S successively by gray average S corresponding for all coordinate points ₀, S ₁..., S _t..., S _2W-1, wherein S _tas follows by following formula:

$S_t=\frac{1}{W}\underset{d=0}{\overset{W-1}{\mathrm{\&Sigma;}}}{G}_1(u,v),t=\mathrm{0,1},...,(2W-1)$

$u={\mathrm{mc}}_k+(d-\frac{W}{2})\&times;\cos \left(O({\mathrm{mc}}_k,{\mathrm{nc}}_k)\right)+(t-\frac{1}{2})\&times;\sin \left(O({\mathrm{mc}}_k,{\mathrm{nc}}_k)\right)$

$v={\mathrm{nc}}_k+(d-\frac{W}{2})\&times;\sin \left(O({\mathrm{mc}}_k,{\mathrm{nc}}_k)\right)+(\frac{1}{2}-t)\&times;\cos \left(O({\mathrm{mc}}_k,{\mathrm{nc}}_k)\right);$

Wherein, sin represents sine function, and cos represents cosine function, and d represents normalized image G ₁(x, y) along the pixel sequence numbering of y-axis, O (mc _k, nc _k) be the field of direction O of typing fingerprint image Q (x, y) _q(x, y) is at pixel (mc _k, nc _k) orientation angle at place.

(2.4c) by gray average S all in (2.4b) ₀, S ₁..., S _t..., S _2W-1regard one group of discrete digital signal as, use Fourier transform, find fundamental frequency F _k, i.e. image block PP for this reason _kfrequency values f _k=F _k.

7. the fingerprint image method for registering based on global information according to claim 1, obtains the barycentric coordinates (i of typing fingerprint image Q (x, y) foreground area in wherein said step (3a) _q, j _q) and the barycentric coordinates (i of template fingerprint image T (x, y) foreground area _t, j _t), undertaken by following formula:

$i_Q=\frac{1}{\mathrm{Hs}}\underset{k=1}{\overset{\mathrm{Hs}}{\mathrm{\&Sigma;}}}{a}_k,j_Q=\frac{1}{\mathrm{Hs}}\underset{k=1}{\overset{\mathrm{Hs}}{\mathrm{\&Sigma;}}}{b}_k$

$i_T=\frac{1}{\mathrm{Ms}}\underset{k=1}{\overset{\mathrm{Ms}}{\mathrm{\&Sigma;}}}{c}_k,j_T=\frac{1}{\mathrm{Ms}}\underset{k=1}{\overset{\mathrm{Ms}}{\mathrm{\&Sigma;}}}{d}_k$

Wherein, (a _k, b _k) be a kth pixel coordinate in typing fingerprint image Q (x, y) foreground area, Hs is the pixel number of typing fingerprint image Q (x, y) foreground area, (c _k, d _k) be a kth pixel coordinate in template fingerprint image T (x, y) foreground area, Ms is the pixel number of template fingerprint image T (x, y) foreground area.

8. the fingerprint image method for registering based on global information according to claim 1, obtains the principal direction angle [alpha] of typing fingerprint image Q (x, y) foreground area in wherein said step (3b) _q, calculate as follows:

(3b1) all pixel coordinates of typing fingerprint image Q (x, y) foreground area are formed a point set X ₁={ x ₁, x ₂..., x _i..., x _lS, obtain mean vector m ₁:

m ₁＝(x ₁+x ₂+...+x _i+...+x _LS)/LS

Wherein x _i=(t _i, g _i) ^t, t _ithe row-coordinate of typing fingerprint image Q (x, y) foreground area i-th pixel, g _ibe the row coordinate of typing fingerprint image Q (x, y) foreground area i-th pixel, LS is the number of typing fingerprint image Q (x, y) foreground area pixel;

(3b2) according to the result of step (3b1), structure covariance matrix s ₁:

s ₁＝{(x ₁-m ₁)×(x ₁-m ₁) ^T+(x ₂-m ₁)×(x ₂-m ₁) ^T+...+(x _i-m ₁)×(x _i-m ₁) ^T+...+(x _LS-m ₁)×(x _LS-m ₁) ^T}/LS

Wherein, T represents and carries out matrix transpose operation to matrix;

(3b3) covariance matrix s is determined ₁projection matrix Z ₁:

To the covariance matrix s in (3b2) ₁carry out svd: s ₁=T ₁l ₁t ₁ ^-1, wherein, L ₁it is covariance matrix _s1eigenwert composition diagonal matrix, T ₁by covariance matrix s ₁eigenwert character pair vector composition eigenmatrix, T ₁ ^-1represent T ₁inverse matrix, T ₁, L ₁and T ₁ ^-1the product T of three ₁l ₁t ₁ ^-1represent, according to covariance matrix s ₁eigenwert order from big to small, reconfigure eigenmatrix T ₁column vector, form projection matrix Z ₁;

(3b4) according to principal component analysis (PCA) principle, the principal direction vector U of typing fingerprint image Q (x, y) foreground area is obtained ₁transform component u on typing fingerprint image Q (x, y) vertical direction _ccthrough projection matrix Z ₁be 1 after projection, the transform component u in typing fingerprint image Q (x, y) horizontal direction _ddthrough projection matrix Z ₁be 0 after projection, use equation Z ₁× U ₁=(1,0) ^trepresent; Calculate principal direction vector U ₁with the angle of typing fingerprint image Q (x, y) horizontal direction: α _q=arctan (u _dd/ u _cc), arctan represents arctan function, using the principal direction angle of this angle as typing fingerprint image Q (x, y) foreground area.

9. the fingerprint image method for registering based on global information according to claim 1, obtains the principal direction angle [alpha] of template fingerprint image T (x, y) foreground area in wherein said step (3b) _t, calculate as follows:

(3b5) all pixel coordinates of template fingerprint image T (x, y) foreground area are formed a point set Y ₁={ y ₁, y ₂..., y _i..., y _lD, y _i=(p _i, q _i) ^t, obtain mean vector m ₂:

m ₂＝(y ₁+y ₂+...+y _i+...+y _LD)/LD

Wherein p _ithe row-coordinate of template fingerprint image T (x, y) foreground area i-th pixel, q _ibe the row coordinate of template fingerprint image T (x, y) foreground area i-th pixel, LD is the number of template fingerprint image T (x, y) foreground area pixel;

(3b6) according to the result of step (3b4), structure covariance matrix s ₂:

s ₂＝{(y ₁-m ₂)×(y ₁-m ₂) ^T+(y ₂-m ₂)×(y ₂-m ₂) ^T+...+(y _i-m ₂)×(y _i-m ₂) ^T+...+(y _LD-m ₂)×(y _LD-m ₂) ^T}/LD

Wherein, T represents and carries out matrix transpose operation to matrix;

(3b7) covariance matrix s is determined ₂projection matrix Z ₂:

To the covariance matrix s in (3b6) ₂carry out svd: s ₂=T ₂l ₂t ₂ ^-1, wherein, L ₂covariance matrix s ₂eigenwert composition diagonal matrix, T ₂by covariance matrix s ₂eigenwert character pair vector composition eigenmatrix, T ₂ ^-1represent T ₂inverse matrix, T ₂, L ₂and T ₂ ^-1the product T of three ₂l ₂t ₂ ^-1represent, according to covariance matrix s ₂eigenwert order from big to small, reconfigure eigenmatrix T ₂column vector, form projection matrix Z ₂;

(3b8) according to principal component analysis (PCA) principle, the principal direction vector U of template fingerprint image T (x, y) foreground area is obtained ₂, the transform component u on template fingerprint image T (x, y) vertical direction _aathrough projection matrix Z ₂be 1 after projection, the transform component u in template fingerprint image T (x, y) horizontal direction _bbthrough projection matrix Z ₂be 0 after projection, use equation Z ₂× U ₂=(1,0) ^trepresent; Calculate principal direction vector U ₂with the angle of template fingerprint image T (x, y) horizontal direction: α _t=arctan (u _bb/ u _aa), arctan represents arctan function, using the principal direction angle of this angle as template fingerprint image T (x, y) foreground area.