CN101901331A - Broken grain splicing method - Google Patents

Abstract

The invention discloses a broken grain splicing method which is used for processing a fingerprint image with broken grains and comprises the following steps of: 1. obtaining a directional diagram of the fingerprint image; 2. computing a broken grain ridgeline period and a broken grain curvature according to the fingerprint image and the directional diagram thereof; 3. drawing the effective area of a broken grain ridgeline according to the broken grain ridgeline period and the broken grain curvature; and 4. fitting the broken grain ridgeline in the effective area. By adopting the broken grain splicing method, the broken grains can be effectively spliced, false characteristic points caused by the broken grains are greatly decreased, the validity of fingerprint characteristic point matching is further improved, and meanwhile, because the broken grains are effectively spliced, the passing rate of ridgeline comparison is improved.

Description

Broken grain splicing method

Technical field

The present invention relates to the fingerprint recognition field, particularly relate to a kind of broken grain splicing method.

Background technology

Along with the progress of society, the security of identification obtains people's attention day by day.Modes such as certificate, password are often adopted in traditional identification.Yet certificate may be lost or be replicated; And password is forgotten about or is produced obscuring easily.Especially along with the arriving of cybertimes, increasing password setting is perplexing people: startup password, mailbox password, bank cipher, forum's password ... for these if identical password is set, can increase potential safety hazard; If different passwords is set, brought puzzlement for Password Management again.For this reason, with biological characteristic (for example, fingerprint, people's face, iris etc.) for to distinguish that the identity recognizing technology of foundation obtains people's attention day by day.Wherein, the discrimination height of fingerprint recognition, and application is the most universal, is acknowledged as " first of the material evidence ".

Existing fingerprint identification technology mainly is based on minutia, promptly extracts minutiae point and characterizes fingerprint image as feature, discerns by comparing these features.Yet,, quite exist the phenomenon of rupture of texture in the fingerprint image of vast scale owing to the breakage of fingerprint, reason such as cast off a skin.This is a very big trouble for the fingerprint recognition Processing Algorithm, because it has brought a lot of pseudo-characteristic points.

At this problem, existing disposal route mainly comprises following mode:

1. on the basis of detecting disconnected line, get rid of the pseudo-characteristic point, promptly at first detect disconnected line region, then will get rid of at the pseudo-characteristic point in this zone.As publication number is a kind of disconnected line disposal route the Chinese patent " fingerprint identification method that detects based on disconnected line " of CN 1588424A just discloses, but this method more complicated needs determine optimal filter by a large amount of statistical experiments.

2. adopt the method for trend pass filtering that fingerprint image is carried out enhancement process, promptly utilize the directivity of fingerprint ridge, make its tangential level and smooth, in the normal direction sharpening, this splices disconnected line to a certain extent.But the spacing between the fingerprint ridge is often smaller, when disconnected line than broad the time, the situation of wrong splicing usually occurs.As seen, there is certain limitation in this method, and it can only handle very narrow disconnected line, and for the disconnected line of broad, this method will lose effectiveness.

Summary of the invention

Technical matters to be solved by this invention is to overcome the deficiency of existing algorithm for recognizing fingerprint on broken grain splicing, makes that the disconnected line of broad also can well be spliced, and then reduces the generation of pseudo-characteristic point.

For solving above technical matters, the invention provides a kind of broken grain splicing method, have the fingerprint image of disconnected line in order to processing, it comprises: (1) asks for the directional diagram of described fingerprint image; (2), calculate described disconnected wrinkle ridge line cycle and disconnected line curvature according to described fingerprint image and directional diagram thereof; (3), describe the effective coverage of disconnected wrinkle ridge line according to described disconnected wrinkle ridge line cycle and disconnected line curvature; (4) in described effective coverage, simulate disconnected wrinkle ridge line.

Further, described step (1) comprising:

(11) the some direction of each pixel of the described fingerprint image of calculating;

(12) the described fingerprint point field of direction is carried out two-dimentional low-pass filtering;

(13) described fingerprint image is divided into the piece of f*f, wherein said f is not less than 1 times of average crestal line cycle of described fingerprint;

(14) utilize median filtering method to count every piece direction;

(15) for the difference of piece direction and its neighborhood piece, adopt the method for three rank convolution interpolation to carry out interpolation, and write down corresponding piece greater than the first default number of degrees;

(16) the described fingerprint-block field of direction is carried out two-dimentional low-pass filtering;

(17) described fingerprint image is divided into the piece of 2f*2f, and repeating step (14) is to (16);

(18) be recorded the pixel of the difference of piece direction for a direction and its, replace the some direction with its piece direction greater than the second default number of degrees;

(19) the described fingerprint point field of direction is carried out two-dimentional low-pass filtering.

Further, the default number of degrees of first in the described step (15) are 20 ± 5 degree.

Further, the default number of degrees of second in the described step (18) are 30 ± 5 degree.

Further, in described step (2), the described disconnected wrinkle ridge line cycle is to obtain by three rank convolution method of interpolation, it comprises the steps: to calculate described disconnected line crestal line cycle of the neighbour's piece on the four direction up and down, and the size of described four neighbour's pieces is k*k, and wherein k is not less than 3 times of average crestal line cycles of described fingerprint; Judge the quality of described four neighbour's pieces, wherein the crestal line cycle is ropy greater than 1.4 times of average crestal line cycles or less than the piece in 0.7 times of average crestal line cycle; In described four neighbour's pieces, exist when being less than 2 ropy, utilize the crestal line cycle of described four neighbour's pieces to carry out three rank convolution interpolation.

Further, in described step (2), the described disconnected wrinkle ridge line cycle is to obtain by three rank convolution method of interpolation, it also comprises the steps: in described four neighbour's pieces, there are 2 or during:, outwards search time neighbour's piece according to the direction of described ropy neighbour's piece more than 2 ropy; Calculate the crestal line cycle of described neighbour's piece and judge its quality; If non-ropy of described neighbour's piece then replaces crestal line cycle of equidirectional neighbour's piece with crestal line cycle of this time neighbour piece, carry out three rank convolution interpolation; If described time neighbour's piece is ropy, then replace the crestal line cycle of equidirectional neighbour's piece with the average crestal line of the fingerprint cycle, carry out three rank convolution interpolation.

Further, in described step (2), described disconnected line curvature is to obtain by three rank convolution method of interpolation, comprises the steps: to search on the vertical direction of two breakpoints of described disconnected line two nearest neighbour's crestal lines; Keep two neighbour's crestal lines between described two breakpoints, obtain two vallate line segments; Adopt sampling a little the method for getting to calculate the curvature of described two vallate line segments; Calculate the curvature of described two breakpoint place crestal lines; Above resulting four curvature are carried out three rank convolution interpolation, obtain described disconnected line curvature.

Further, in described step (2), described disconnected line curvature is to obtain by three rank convolution method of interpolation, and it also comprises the steps: to get the curvature of this crestal line section place crestal line when on the described crestal line section disconnected line being arranged, and is the curvature of this crestal line section.

Further, in described step (3), the effective coverage of described disconnected wrinkle ridge line is made of two rough disconnected wrinkle ridge line segments, and wherein said rough disconnected wrinkle ridge line segment obtains by the following method:

(31) described disconnected line two breakpoints become 90 degree and 270 degree, two vertical direction with its place piece direction on, extend, obtain four initial points by the certain proportion in described disconnected wrinkle ridge line cycle;

(32) calculate the mid point between two initial points on the same vertical direction by described disconnected line curvature, as initial point;

(33) on same vertical direction, calculate two mid points between the adjacent initial point by described disconnected line curvature, as initial point;

(34) repeating step (33) is up to described two the rough disconnected wrinkle ridge line segments that obtain being connected into by initial point.

Further, in described step (4),, simulate described disconnected wrinkle ridge line to the centre respectively from described disconnected line two breakpoints according to described fingerprint image and directional diagram thereof.

The invention has the beneficial effects as follows: adopt above broken grain splicing method, can effectively splice disconnected line, significantly reduce because the pseudo-characteristic point that disconnected line causes has further improved the validity that fingerprint feature point mates, because disconnected line has been carried out effective splicing, improve the percent of pass of crestal line comparison simultaneously.

Description of drawings

Fig. 1 is the schematic flow sheet of the broken grain splicing method that one embodiment of the invention provided;

Fig. 2 is multiple dimensioned schematic diagram that one embodiment of the invention provided;

Fig. 3 is multiple dimensioned direction correction figure that one embodiment of the invention provided;

Fig. 4 utilizes the schematic flow sheet of the direction that this multiple dimensioned method corrects a mistake for one embodiment of the invention provided;

Fig. 5 is the schematic flow sheet that obtains in disconnected wrinkle ridge line cycle that one embodiment of the invention provided;

Fig. 6 to Figure 10 is the splicing synoptic diagram of the present invention's one disconnected line example;

The design sketch of broken grain splicing is provided for the method for utilizing one embodiment of the invention and providing Figure 11 to Figure 13.

Embodiment

For technical characterictic of the present invention is become apparent, below in conjunction with specific embodiment, the present invention will be further described.

Please refer to Fig. 1, it is the schematic flow sheet of the broken grain splicing method that one embodiment of the invention provided.This method has the fingerprint image of disconnected line in order to processing, as shown in the figure, may further comprise the steps:

S1: the directional diagram of asking for fingerprint image;

S2:, calculate disconnected wrinkle ridge line cycle and disconnected line curvature according to fingerprint image and directional diagram thereof;

S3:, describe the effective coverage of disconnected wrinkle ridge line according to disconnected wrinkle ridge line cycle and disconnected line curvature;

S4: in described effective coverage, simulate disconnected wrinkle ridge line.

Above method is different from prior art, and it utilizes the directional diagram of fingerprint image, disconnected wrinkle ridge line cycle and disconnected line curvature, obtains the effective coverage of a disconnected wrinkle ridge line, and simulates disconnected wrinkle ridge line in this effective coverage, and reaches the purpose of broken grain splicing.It is with respect to the method that detects disconnected line and then get rid of pseudo-characteristic point, not to get rid of the pseudo-characteristic point, but, reduce the generation of pseudo-characteristic point by the disconnected line of splicing by detecting, and then need not the devise optimum wave filter, effectively reduced the complicacy that pseudo-characteristic point is removed.In addition, for the method that trend pass filtering carries out the figure image intensifying, the present invention fully takes into account fingerprint ridge and is mostly smoothly, has these characteristics of continuous streakline certain curvature, gradual change, mode by match is spliced, make it not be subject to the width of disconnected line, have better splicing and render a service, and use more extensive.

Describe the implementation procedure of above each step in detail below in conjunction with Fig. 6 to Figure 10: wherein Fig. 6 is disconnected line exemplary plot; Fig. 7 is disconnected line Curvature Interpolation schematic diagram: promptly by the break interpolation of line curvature of neighborhood curvature information, as figure, obtain disconnected line curvature r by curvature r1, r2, r3, the r4 of neighborhood according to three rank convolution interpolation formulas (4), in like manner obtain the wrinkle ridge line cycle of breaking; Fig. 8 obtains the effective coverage figure of disconnected wrinkle ridge line for by disconnected line curvature and disconnected wrinkle ridge line cycle.Fig. 9 obtains disconnected line direction, the disconnected wrinkle ridge line chart that simulates for referring to by neighbour's directional information interpolation in the effective coverage.Figure 10 is final broken grain splicing design sketch.

One, the asking for of the directional diagram of fingerprint image (step S1):

In step S1,, methods multiple dimensioned and three rank convolution interpolation have been introduced in a preferred embodiment of the present invention, the direction of correcting a mistake in order to guarantee the accuracy of direction of fingerprint.Specifically please refer to Fig. 2 and Fig. 3, it is respectively multiple dimensioned schematic diagram and multiple dimensioned direction correction figure that one embodiment of the invention provides.The height of yardstick is observed in 1,2,3,4 expressions among the figure, and numeral is big more, and it is high more that yardstick is observed in representative.For example, observing yardstick 1 is the view window of 2*2, and observing yardstick 2 is the view window of 4*4, and observing yardstick 3 is the view window of 8*8, and observing yardstick 4 is the view window of 16*16.The present invention adopts from observing yardstick 1 beginning upwards travel direction correction step by step.As can be seen from Figure 3, it is high more to observe yardstick, and direction changes mild more, helps direction of fingerprint more.Below in conjunction with Fig. 4, specifically describe direction how to utilize this multiple dimensioned method to correct a mistake.It comprises the steps:

S11: the some direction of each pixel of calculated fingerprint image;

S12: the fingerprint point field of direction is carried out two-dimentional low-pass filtering;

S13: fingerprint image is divided into the piece of f*f, and wherein f is not less than 1 times of average crestal line cycle of fingerprint;

S14: utilize median filtering method to count every piece direction;

S15:, adopt the method for three rank convolution interpolation to carry out interpolation, and write down corresponding piece for the difference of piece direction and its neighborhood piece greater than the first default number of degrees;

S16: the fingerprint-block field of direction is carried out two-dimentional low-pass filtering;

S17: fingerprint image is divided into the piece of 2f*2f and repeating step S14 to S16;

S18: be recorded the pixel of the difference of piece direction for a direction and its, replace the some direction with its piece direction greater than the second default number of degrees;

S19: the fingerprint point field of direction is carried out two-dimentional low-pass filtering.

Usually, after obtaining fingerprint image, need fingerprint image is carried out pre-service, for example filtering and noise reduction, picture specificationization etc. obtain binary image (or refined image).It is well known to those skilled in the art, and does not repeat them here.

For the fingerprint image that obtains, at first calculate the some direction (step S11) of each pixel with general Sobel operator gradient formula; And carry out two-dimentional low-pass filtering, this filtering is equivalent to following formula (1):

$g(u,v)=\underset{x}{\mathrm{\Σ}}\underset{y}{\mathrm{\Σ}}f(x,y)H(u-x+1,v-y+1)---\left(1\right)$

Wherein, wherein, (x y) is the fingerprint image picture point field of direction to f, and (x y) is the impulse Response Function of wave filter to H, and (u v) is the low-pass filtering result to g.

After finishing two-dimentional low-pass filtering, can obtain putting comparatively accurately the field of direction.But because view window is too little, the direction of a lot of regional areas can't obtain correcting, and need more correct under the large observation window.For this reason, the piece (step S13) that image is divided into f*f (f is not less than 1 times of this fingerprint average crestal line cycle) size; And utilize median filtering method to count every direction (step S14); For the difference of piece direction and its neighborhood piece, adopt the method for three rank convolution interpolation to carry out interpolation, and write down corresponding piece (step S15) greater than the first default number of degrees; Find that through statistical research when the difference of piece direction and its neighborhood was spent greater than 20, the possibility that field of direction mistake occurs was bigger, so these first default number of degrees are set in about 20 degree, for example 20 ± 5 spend; And three rank convolution interpolation formulas such as formula (2), (3) and (4):

$\mathrm{Sinc}\left(x\right)=\left\{\begin{array}{c}\frac{\sin x}{x},x\≠0\\ 1,x=0\end{array}\right.---\left(2\right)$

$\mathrm{Cbs}\left(x\right)=\left\{\begin{array}{c}(x-2)x^2+1,(x\&le;1)\\ ((-x+5)x-8)x+4,(1<x\&le;2)\\ 0,(2<x)\end{array}\right.---\left(3\right)$

$f(x_0,y_0)=\underset{k}{\mathrm{\&Sigma;}}\underset{l}{\mathrm{\&Sigma;}}f(x_k,y_l)\mathrm{Cbs}(x_k-x_0)\mathrm{Cbs}(y_l-y_0)---\left(4\right)$

Wherein, Cbs (x) is the approximate match of Sinc (x), f (x ₀, y ₀) be from peripheral neighborhood f (x _k, y _l) interpolation that obtains.

By after above method corrected all pieces, do once two-dimentional low-pass filtering (step S16).The piece that again image is divided into 2f*2f is corrected once more and filtering, and is write down corresponding piece (step S17).Usually, two-stage is observed yardstick and can be corrected the field of direction preferably.But for the relatively poor zone of some picture quality, direction of fingerprint need just can be judged under higher scale, observes the correction of yardstick travel direction so can continue to increase, and can be increased to more than 4 grades or 4 grades in case of necessity.

At last, be recorded for a direction and its place and correct back piece direction and differ bigger pixel, with its piece direction replacement point direction (step S18), two-dimentional once more low-pass filtering is with the level and smooth field of direction (step S19).

Through the correction of above direction, the accuracy of asking for of direction of fingerprint is higher, helps the match of back teasel root wrinkle ridge line more.

Two, calculate disconnected wrinkle ridge line cycle and disconnected line curvature;

In the present embodiment, introduce the method for three rank convolution interpolation and calculate disconnected wrinkle ridge line cycle and disconnected line curvature, concrete grammar is: at first find two breakpoints of disconnected line, adopt three rank convolution interpolation methods to calculate disconnected wrinkle ridge line cycle and disconnected line curvature again in its peripheral higher scale scope.

The acquisition process in disconnected wrinkle ridge line cycle (one), at first, is described:

Equally, adopt three rank convolution interpolation methods as formula (2), (3) and (4) shown in, the neighborhood interpolation by the line place of breaking obtains the wrinkle ridge line cycle of breaking.Described neighborhood is disconnected line neighbour's piece (wherein k is not less than 3 times of average crestal line cycles of described fingerprint) of k*k size on the four direction up and down.Detailed process such as Fig. 5 may further comprise the steps:

S21: the crestal line cycle of calculating four neighbour's pieces;

S22: judge the quality of four neighbour's pieces, particularly, basis for estimation is: the crestal line cycle is ropy greater than 1.4 times of average crestal line cycles or less than the piece in 0.7 times of average crestal line cycle;

In four neighbour's pieces, exist when being less than 2 ropy, carry out step S23:

S23: utilize the crestal line cycle of four neighbour's pieces to carry out three rank convolution interpolation.

In four neighbour's pieces, there are 2 or during more than 2 ropy, carry out following steps;

S24:, outwards search time neighbour's piece according to the direction of ropy neighbour's piece;

S25: calculate the crestal line cycle of time neighbour's piece and judge its quality;

If non-ropy of this time neighbour piece then carries out step S26,

S26: replace crestal line cycle of equidirectional neighbour's piece with crestal line cycle of this time neighbour piece, carry out three rank convolution interpolation;

If this time neighbour piece is ropy, then carry out step S27:

S27: replace crestal line cycle of equidirectional neighbour's piece with the average crestal line of the fingerprint cycle, carry out three rank convolution interpolation.

More than relate to the calculating in the crestal line cycle of piece, provide its computing method and formula (5) and (6) below:

$X\left[k\right]=\frac{1}{w}\underset{d=0}{\overset{w-1}{\mathrm{\&Sigma;}}}G(u,v),k=\mathrm{0,1},...,L-1---\left(5\right)$

$u=i+(d-\frac{w}{2}).\cos \mathrm{\&theta;}(i,j)+(k-\frac{L}{2}).\sin \mathrm{\&theta;}(i,j)$

$v=j+(d-\frac{w}{2}).\sin \mathrm{\&theta;}(i,j)+(k-\frac{L}{2}).\cos \mathrm{\&theta;}(i,j)---\left(6\right)$

To each so that (i j) is the piece at center, calculates so that (i j) for the size at center is the rectangular window (L is parallel to the gradient direction of fingerprint, and W is parallel to the fingerprint ridge line direction, and W is generally 1 times of crestal line cycle, and L is 4 times of crestal line cycles) of L*W; To the wicket of each L*W, calculate L and go up each point in projection average along the W direction.Wherein G is a fingerprint image, and θ is the directional diagram of fingerprint image; X[k] form a similar sinusoidal curve, (i, j), this is the crestal line cycle of this piece to calculate the mean distance T of a plurality of adjacent summits of this sinusoidal curve.

(2), the following acquisition process (as Fig. 6 and 7) of describing disconnected line curvature:

Below mention, mostly fingerprint ridge is smoothly, has continuous streakline certain curvature, gradual change.At disconnected line place, according to the change information of neighborhood streakline, the change information of especially disconnected line vertical direction neighbour's streakline just can be estimated the curvature of disconnected line to be spliced roughly.Specific as follows:

At first on the vertical direction of two breakpoints 11 of disconnected line 10 and 12, search nearest two neighbour's crestal lines 20,30, block this two neighbour's crestal lines 20,30, keep the part between two breakpoints, obtain two vallate line segments; Adopt sampling a little the method for getting to calculate the curvature r1 of this two vallates line segment and r4 (if on the crestal line section that obtains disconnected line is arranged, then get the curvature of this crestal line section place crestal line), be the center with disconnected line 10 again, two vallate curvature of a curve r2 and r3 (i.e. the curvature of two breakpoint place crestal lines) about calculating on the crestal line of disconnected line place respectively, get k=2, l=2 to these four curvature r1 to r4 by formula (4) carry out three rank convolution interpolation and obtain disconnected line curvature r to be spliced.

Three, describe the effective coverage (as Fig. 8) of disconnected wrinkle ridge line;

In above step, obtained to resolve wrinkle ridge line cycle and disconnected line curvature, so, just can depict the effective coverage A that constitutes by two rough disconnected wrinkle ridge line segments 40,50.Specific as follows:

At first, become 90 degree and 270 to spend two vertical direction on 12 with its place piece direction at described disconnected line two breakpoints 11, certain proportion by the disconnected wrinkle ridge line cycle extends, obtain four initial points 41,42,51 and 52, these four initial points 41,42,51 and 52 just constitute the end points of two rough disconnected wrinkle ridge line segments 40,50; The ratio of wherein extending the selected disconnected wrinkle ridge line cycle can be 15% to 20%.Then, calculate the mid point between two initial points 43 and 53 on the same vertical direction by disconnected line curvature, as initial point.Then, on same vertical direction, calculate two mid points (for example, on 90 degree vertical direction, calculate the mid point between initial point 41 and 43, and the mid point between initial point 42 and 43) between the adjacent initial point by disconnected line curvature, as initial point.Repeat above step,,, thus, obtain the effective coverage A of disconnected wrinkle ridge line to depict two rough disconnected wrinkle ridge line segments 40 and 50 that connect into by initial point up to obtaining abundant initial point.

Four, simulate disconnected wrinkle ridge line (as Fig. 9).

Behind the effective coverage A that obtains disconnected wrinkle ridge line, the just directional information that can slowly change according to fingerprint, from two breakpoints 11 and 12 respectively to middle match.At first,, calculate the position of next match point respectively, obtain two match points according to directional information since two breakpoints 11 and 12; In like manner, calculate following two match points respectively with these 2 again, so repeatedly, for the point that exceeds two rough disconnected wrinkle ridge line segments 40 and 50, with the point on the disconnected roughly wrinkle ridge line segment 40 and 50 of arest neighbors as match point, with guarantee all match points two rough disconnected wrinkle ridge line segments 40 and 50 and they between.

Please refer to Figure 11 to Figure 13, it has provided and has utilized method provided by the present invention to carry out the instantiation of broken grain splicing, and wherein, Figure 11 one has the fingerprint exemplary plot of disconnected line, and Figure 12 is the refined image of this fingerprint image, and Figure 13 is the design sketch through broken grain splicing.As can be seen from the figure, original-gray image has very dark disconnected line; From refined image, as can be seen, in not having the algorithm of broken grain splicing, there are a lot of disconnected lines; And after the broken grain splicing method that utilizes above embodiment to provide handled, disconnected line phenomenon disappeared.

Adopt broken grain splicing method provided by the present invention, can effectively splice disconnected line, significantly reduce because the pseudo-characteristic point that disconnected line causes has further improved the validity that fingerprint feature point mates, because disconnected line has been carried out effective splicing, improve the percent of pass of crestal line comparison simultaneously.

, be not that protection scope of the present invention should be as the criterion with the scope that claims are contained in order to qualification the present invention below only for for example.

Claims (10)

1. broken grain splicing method has the fingerprint image of disconnected line in order to processing, it is characterized in that, comprising:

(1) asks for the directional diagram of described fingerprint image;

(2), calculate described disconnected wrinkle ridge line cycle and disconnected line curvature according to described fingerprint image and directional diagram thereof;

(3), describe the effective coverage of disconnected wrinkle ridge line according to described disconnected wrinkle ridge line cycle and disconnected line curvature;

(4) in described effective coverage, simulate disconnected wrinkle ridge line.

2. broken grain splicing method according to claim 1 is characterized in that, described step (1) comprising:

(11) the some direction of each pixel of the described fingerprint image of calculating;

(12) the described fingerprint point field of direction is carried out two-dimentional low-pass filtering;

(13) described fingerprint image is divided into the piece of f*f, wherein said f is not less than 1 times of average crestal line cycle of described fingerprint;

(14) utilize median filtering method to count every piece direction;

(15) for the difference of piece direction and its neighborhood piece, adopt the method for three rank convolution interpolation to carry out interpolation, and write down corresponding piece greater than the first default number of degrees;

(16) the described fingerprint image piece field of direction is carried out two-dimentional low-pass filtering;

(17) described fingerprint image is divided into the piece of 2f*2f, and repeating step (14) is to (16);

(18) be recorded the pixel of the difference of piece direction for a direction and its, replace the some direction with its piece direction greater than the second default number of degrees;

(19) the described fingerprint point field of direction is carried out two-dimentional low-pass filtering.

3. broken grain splicing method according to claim 2 is characterized in that, the default number of degrees of first in the described step (15) are 20 ± 5 degree.

4. broken grain splicing method according to claim 2 is characterized in that, the default number of degrees of second in the described step (18) are 30 ± 5 degree.

5. broken grain splicing method according to claim 1 is characterized in that, in described step (2), the described disconnected wrinkle ridge line cycle is to obtain by three rank convolution method of interpolation, and it comprises the steps:

Calculate described disconnected line crestal line cycle of the neighbour's piece on the four direction up and down, and the size of described four neighbour's pieces is k*k, wherein k is not less than 3 times of average crestal line cycles of described fingerprint;

Judge the quality of described four neighbour's pieces, wherein the crestal line cycle is ropy greater than 1.4 times of average crestal line cycles or less than the piece in 0.7 times of average crestal line cycle;

In described four neighbour's pieces, exist when being less than 2 ropy, utilize the crestal line cycle of described four neighbour's pieces to carry out three rank convolution interpolation.

6. broken grain splicing method according to claim 5 is characterized in that, in described step (2), the described disconnected wrinkle ridge line cycle is to obtain by three rank convolution method of interpolation, and it also comprises the steps:

In described four neighbour's pieces, there are 2 or during more than 2 ropy:

According to the direction of described ropy neighbour's piece, outwards search time neighbour's piece;

Calculate the crestal line cycle of described neighbour's piece and judge its quality;

If non-ropy of described neighbour's piece then replaces crestal line cycle of equidirectional neighbour's piece with crestal line cycle of this time neighbour piece, carry out three rank convolution interpolation;

If described time neighbour's piece is ropy, then replace the crestal line cycle of equidirectional neighbour's piece with the average crestal line of the fingerprint cycle, carry out three rank convolution interpolation.

7. broken grain splicing method according to claim 1 is characterized in that, in described step (2), described disconnected line curvature is to obtain by three rank convolution method of interpolation, comprises the steps:

On the vertical direction of two breakpoints of described disconnected line, search two nearest neighbour's crestal lines;

Keep two neighbour's crestal lines between described two breakpoints, obtain two vallate line segments;

Adopt sampling a little the method for getting to calculate the curvature of described two vallate line segments;

Calculate the curvature of described two breakpoint place crestal lines;

Above resulting four curvature are carried out three rank convolution interpolation, obtain described disconnected line curvature.

8. broken grain splicing method according to claim 7 is characterized in that, in described step (2), described disconnected line curvature is to obtain by three rank convolution method of interpolation, and it also comprises the steps:

When on the described crestal line section disconnected line being arranged, get the curvature of this crestal line section place crestal line, be the curvature of this crestal line section.

9. broken grain splicing method according to claim 1 is characterized in that, in described step (3), the effective coverage of described disconnected wrinkle ridge line is made of two rough disconnected wrinkle ridge line segments, and wherein said rough disconnected wrinkle ridge line segment obtains by the following method:

(31) described disconnected line two breakpoints become 90 degree and 270 degree, two vertical direction with its place piece direction on, extend, obtain four initial points by the certain proportion in described disconnected wrinkle ridge line cycle;

(32) calculate the mid point between two initial points on the same vertical direction by described disconnected line curvature, as initial point;

(33) on same vertical direction, calculate two mid points between the adjacent initial point by described disconnected line curvature, as initial point;

(34) repeating step (33) is up to described two the rough disconnected wrinkle ridge line segments that obtain being connected into by initial point.

10. broken grain splicing method according to claim 1 is characterized in that, in described step (4), according to described fingerprint image and directional diagram thereof, simulates described disconnected wrinkle ridge line to the centre respectively from described disconnected line two breakpoints.

Images