CN101751564A - Intravenous grain extraction method based on maximal intra-neighbor difference vector diagram - Google Patents

Abstract

The invention discloses an intravenous grain extraction method based on maximal intra-neighbor difference vector diagram, which firstly calculates the maximal intra-neighbor difference and directions of all pixel points in an original intravenous image to construct a maximal intra-neighbor difference vector diagram, and further calculates the maximal intra-neighbor vector difference of the maximal intra-neighbor difference vector diagram, finally obtains proper thresholds from the histogram of the maximal intra-neighbor vector difference diagram and extracts the intravenous grain of the original image from the maximal intra-neighbor vector difference diagram. The invention avoids the image pretreatment process, also can eliminate the influence of intravenous image coherent features on grain extraction such as unevenness and fuzzy boundary, can completely extract the intravenous grain and ensures the grain to be even and have less noise points, and also greatly facilitates the following feature extraction.

Description

Veinprint extracting method based on maximal intra-neighbor difference vector

Technical field

The present invention relates to a kind of veinprint extracting method, particularly a kind of veinprint extracting method based on maximal intra-neighbor difference vector.

Background technology

Characteristics of human body's identification is that a kind of human body physiological characteristics and behavior pattern utilized carried out identification and authentication technology, wherein, vein identification technology is as a kind of emerging biometrics identification technology, because of it has untouchable, characteristics and live body detectabilities such as high resistance interference and antifalsification, obtain increasing researchist's concern in the last few years, become a research focus in characteristics of human body's identification.The technical foundation that the vein pattern is obtained is near infrared spectroscopy (NIRS) and near infrared imaging, a certain position when 720nm-950nm near infrared light irradiation human body, its meeting transdermal surface, wherein intravasation then can be absorbed by the deoxygenation protoheme in the blood, and the near infrared light that enters in its hetero-organization can be reflected or be scattered, thereby the imaging meeting of vasculature part is than other position dimnesses, and then forms the near infrared vein image.But since subcutaneous fat, the influence of bone and illumination, the vein image that obtains by this acquisition mode exists light and shade to differ, and the lines thickness is inhomogeneous, the relatively fuzzyyer situation that influences the subsequent treatment effect that waits of part lines.

These characteristics at vein image, the various countries researchist has proposed some veinprint extraction algorithms, usually be divided into two classes: a class is to utilize or improve classic method that vein image is cut apart and obtain veinprint, as utilize the adaptive threshold method, combination Hessian wave filter, matched filter, Mathematical Morphology Method, methods such as multi-scale geometric analysis are cut apart.Wherein, the adaptive threshold method can extract veinprint from the image of uneven illumination, but can comprise some irregular shade and noises in the veinprint that extracts.Matched filter and Mathematical Morphology Method just have effect preferably to the image of veinprint even thickness, and still have more noise in the veinprint that extracts.Another kind of is that complete feature according to veinprint has proposed some new veinprint extracting method targetedly, as utilize the T1 Repeated Line Tl tracing to extract veinprint, this method at random begin to survey from diverse location, pursue coordinate tracking along black line after detecting local black line, thereby repeatedly repeat to survey and cutting apart with all the black line track of following the tracks of stack realization veinprints, therefore this method can effectively extract veinprint from distinct image not too, but for wherein thin or short veinprint, then because the repetition detection times deficiency on it, and make this part veinprint effectively to extract, lose easily; For example a kind of again dividing method based on local maximum curvature, this method with the central point weighted stacking of transversal section local maximum curvature value correspondence, and connect all the other central points and obtain final veinprint respectively on four different directions.More than these algorithms have separately characteristics respectively, but the veinprint that most of algorithm extracts is thicker, and is difficult to be partitioned into thinner veinprint, with real veinprint certain distance is arranged, be unfavorable for the extraction of vein pattern, therefore, can produce certain influence follow-up processing.The scholar who has has proposed framework extraction method based on watershed algorithm at vein image, this algorithm is at first by directly applying to watershed algorithm the vein gray level image, acquisition comprises the veinprint of many unnecessary crestal lines and noise, handle by morphology then and suppress noise and delete unnecessary lines, but this method also exists some problems: can not extract the veinprint that contains floating end points, and can be a veinprint with two veinprint mistakes knowledge of close together.Also have the scholar to propose a kind of based on maximal intra-neighbor difference (Maximal intra-neighbor difference, MIND) vein image partitioning algorithm, its core is to make full use of the neighborhood information of vein image and the MIND image that newly-designed distance function calculates former figure, and with obtained the enhancing image through the former figure weighting summation after the histogram modification, afterwards, by calculate the average image that strengthens image and with strengthen image and be weighted and relatively obtain final segmentation result, in addition, this method can improve segmentation effect according to the partitioning parameters in the histogram Adaptive adjusting algorithm of MIND image, but still has some noises in the veinprint that this method obtains.At these problems, there is the scholar to propose a kind of vein image partitioning algorithm based on field of direction distributive law, this algorithm makes full use of the space attribute of former vein image feature and field of direction image, and with the cut apart criterion of field of direction distributive law as differentiation vein and background, lines for the thickness inequality can be obtained effect preferably, but still have more noise between close lines, and since noise concentrate relatively and may cause a small amount of pseudo-lines.

Summary of the invention

The objective of the invention is to propose a kind of veinprint extracting method based on maximal intra-neighbor difference vector, this method need not be carried out pre-service to image, by former figure is converted to VGMIND, make full use of the Vector Message of each neighborhood of pixel points among the VGMIND, eliminate lines thickness inequality and lines obscurity boundary to cutting apart the interference that brings, the veinprint that extracts clear evenly and noise spot few, be convenient to follow-up feature extraction.

For reaching above-mentioned purpose, the present invention adopts following technical scheme: a kind of veinprint extracting method based on maximal intra-neighbor difference vector, it is characterized in that, and comprise the steps:

The first step calculates the maximal intra-neighbor difference and the direction thereof of all pixels in the former vein image, obtains the maximal intra-neighbor difference vector of former vein image; The mutual relationship of this polar plot reflection pixel and its neighborhood, from then on image can be rough extracts veinprint, but has more noise.

Second goes on foot, and calculates the interior phasor difference of maximum neighborhood of all pixels in the described maximal intra-neighbor difference vector, obtains the interior phasor difference figure of maximum neighborhood of former vein image; To give prominence to veinprint more clearly;

The 3rd step, obtain the histogram of phasor difference figure in the described maximum neighborhood, from described histogram, obtain threshold value, and extract the veinprint of former vein image with described threshold value phasor difference figure in the maximum neighborhood.

In order to realize the present invention better, maximal intra-neighbor difference that calculates all pixels in the former vein image and direction thereof in the described first step obtain the maximal intra-neighbor difference vector of former vein image, comprise the steps:

In the A step, determine any pixel neighborhood of a point in the former vein image according to the radius of neighbourhood and distance function, and select partition neighborhood template T _Div-neiIt is the adjacent fan-shaped stick mass at center with this pixel that described neighborhood is divided into 16 _iI=1 wherein ... 16, radius of neighbourhood R:1≤R≤11, the unit of R is a pixel; Distance function adopts the Euclidean distance that rounds up: $d_{R\_\mathrm{Circle}}=\mathrm{Round}\left(\sqrt{{(x_1-x_2)}^2+{(y_1-y_2)}^2}\right),$ In the formula, (x ₁, y ₁), (x ₂, y ₂) be coordinate any 2 in the former vein image;

In the B step, calculate each fan-shaped stick mass in this pixel neighborhood of a point _iInterior all gray values of pixel points p (x, y) sum S _i=∑ p (x, y); Pixel (x, y) ∈ mass wherein _i, (x y) is pixel (x, gray-scale value y), i=1 to p ... 16; Calculate semicircle zone block in this pixel neighborhood of a point _iIn all gray values of pixel points p (x, y) and semicircle complementary zone block with it _{I '}In all gray values of pixel points p (x, y) and between difference T _i:

$T_i=\frac{1}{8}(\underset{j=i}{\overset{i+7}{\mathrm{\Σ}}}{S}_{\mathrm{mod}(j-\mathrm{1,16})+1}-\underset{j=i+8}{\overset{i+15}{\mathrm{\Σ}}}{S}_{\mathrm{mod}(j-\mathrm{1,16})+1}),$ Wherein, semicircle zone block _iBe meant from fan-shaped stick mass _iBeginning is clockwise to fan-shaped stick mass _{Mod (i+6,16)+1}8 continuums that continuous fan-shaped stick is formed; Its complementary semicircle zone block _{I '}Be meant from fan-shaped stick mass _{I '}8 continuums that continuous fan-shaped stick is formed that beginning is calculated clockwise; I=1 ... 16, i '=mod (i+7,16)+1;

The C step is at two tuples vector l=[(T ₁, direc ₁), (T ₂, direc ₂) ... (T ₁₆, direc ₁₆)], T _i+ T _{Mod (i+7, 16)+1}Select the maximum of T in all two tuple elements in=0 _Max=max (T _i), and obtain maximum of T thus _MaxCorresponding direction ${\mathrm{direc}}_{T_{\max }}={\mathrm{direc}}_i\left(i\right|T_i=T_{\max }),$ T _MaxAnd two corresponding direction in the tuple ${\mathrm{direc}}_{T_{\max }}={\mathrm{direc}}_i\left(i\right|T_i=T_{\max })$ Be the maximal intra-neighbor difference vector value and the direction vector thereof of the pixel of asking; Wherein, vectorial l is by 16 two tuple (T _i, direc _i) constitute; T _iBe semicircle zone block _iIn all gray values of pixel points p (x, y) and semicircle complementary zone block with it _{I '}In all gray values of pixel points p (x, y) and between difference, direc _iBe semicircle zone block _iThe corresponding area direction, i=1 ... 16, i '=mod (i+7,16)+1; Region direction direc _iBe meant and semicircle zone block _iDiameter vertical, and point to semicircle zone block _iThe direction of semi arch direction;

The D step repeats above-mentioned A and goes on foot the step to C, obtains the maximal intra-neighbor difference vector value and the direction vector thereof of all pixels in the former vein image, promptly obtains the maximal intra-neighbor difference vector of former vein image.

The partition neighborhood template T of described A in the step _Div-nei1., make fan-shaped stick mass behind the piecemeal select as follows: _iAnd mass _{Mod (i+7,16)+1}With this pixel is former point symmetry; 2., make each fan-shaped stick mass behind the piecemeal _iIn comprise pixel number equate to greatest extent.

Calculate the interior phasor difference of maximum neighborhood of all pixels of described maximal intra-neighbor difference vector in described second step, obtain the interior phasor difference figure of maximum neighborhood of former vein image, comprise the steps:

The I step, determine any pixel neighborhood of a point in the described maximal intra-neighbor difference vector according to the radius of neighbourhood, selecting the partition neighborhood template that described neighborhood is divided into 16 is the adjacent fan-shaped stick mass at center with this pixel _iCalculate each fan-shaped stick mass respectively _iInterior all vector v (x, y) sum S _i ^v:

$S_i^v=\mathrm{\Σv}(x,y),(x,y)\∈{\mathrm{mass}}_i,i=1\·\·\·16;$

And calculate each semicircle zone block _iInterior all vector v (x, y) sum S _i ^b,

$S_i^b=\frac{1}{8}\underset{j=i}{\overset{i+7}{\mathrm{\Σ}}}{S}_{\mathrm{mod}(j-\mathrm{1,16})+1}^v,i=1\·\·\·16;$

The II step is with 16 S _i ^bBe divided into 8 pairs, each is to S _i ^bAnd S _{I '} ^bPairing semicircle zone block _iAnd block _{I '}8 couples of T are calculated in complementation then _i ^v, T _i ^vSize be S _i ^bAnd S _{I '} ^bThe mould value T of the two phasor difference _i ^m, its symbol is according to sign function sgn (S _i ^b, S _{I '} ^b) decision,

$T_i^v=\mathrm{sign}(S_i^b,S_{i^{\′}}^b)\×T_i^m,i=1\·\·\·8$

$T_i^m=|S_i^b-S_{i^{\′}}^b|,i=1\·\·\·8$

$\mathrm{sgn}(S_i^b,S_{i^{\&prime;}}^b)=\left\{\begin{array}{c}1,|\mathrm{angle}\left(S_i^b\right)-{\mathrm{direc}}_i|<\mathrm{\&pi;}/2\&cap;|\mathrm{angle}\left(S_{i^{\&prime;}}^b\right)-{\mathrm{direc}}_{i^{\&prime;}}|<\mathrm{\&pi;}/2\\ -1,\mathrm{else}\end{array}\right.,i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

In the III step, calculate maximum of T _Max ^v, be phasor difference in the maximum neighborhood of this pixel,

$T_{\max }^v=\max \left(T_i^v\right),i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

IV goes on foot, and repeats the above-mentioned I step to the III step, calculates the interior phasor difference of maximum neighborhood of all pixels in the maximal intra-neighbor difference vector, promptly obtains the interior phasor difference figure of maximum neighborhood of former vein image;

Wherein, radius of neighbourhood R:1≤R≤11, the unit of R is a pixel; Partition neighborhood template T _Div-nei1., make fan-shaped stick mass behind the piecemeal select as follows: _iAnd mass _{Mod (i+7,16)+1}With this pixel is former point symmetry; 2., make each fan-shaped stick mass behind the piecemeal _iIn comprise pixel number equate to greatest extent; Semicircle zone block _iBe meant from fan-shaped stick mass _iBeginning is clockwise to fan-shaped stick mass _{Mod (i+6,16)+1}8 continuums that continuous fan-shaped stick is formed; Its complementary semicircle zone block _{I '}Be meant from fan-shaped stick mass _{I '}8 continuums that continuous fan-shaped stick is formed that beginning is calculated clockwise; I=1 ... 16, i '=mod (i+7,16)+1.

Obtain the histogram of phasor difference figure in the maximum neighborhood in described the 3rd step; Obtain lines from described histogram and extract threshold value; And refer to the veinprint that described threshold value phasor difference figure in maximum neighborhood extracts former vein image: with value corresponding to flex point between first maximum point and first minimum point the histogram of phasor difference figure in the described maximum neighborhood; Extract threshold value a as lines; Then with the pixel of gray value 〉=a among the phasor difference figure in the described maximum neighborhood all assignment be 255; The pixel of gray value＜a all assignment is 0 $P_{\mathrm{pattern}}(i,j)=\left\{\begin{array}{c}0,P_{\mathrm{GMINVD}}(i,j)<a\\ 255,P_{\mathrm{GMINVD}}(i,j)\&GreaterEqual;a\end{array}\right.,$ Promptly obtain not carrying out as yet the veinprint that smoothing denoising is handled.

Described radius of neighbourhood R=7 is a preferred version.

The present invention with respect to the major advantage and the beneficial effect of prior art is:

Veinprint extracting method based on maximal intra-neighbor difference vector disclosed by the invention, need not carry out pre-service to image, by former figure is converted to VGMIND, make full use of the Vector Message of each neighborhood of pixel points among the VGMIND, eliminated lines thickness inequality and lines obscurity boundary to cutting apart the interference that brings.Experimental result shows, this method has been eliminated the influence that vein image inherent characteristics such as thickness inequality and obscurity boundary are extracted lines, thereby can complete extraction go out veinprint, lines clear evenly and noise spot few, be convenient to follow-up feature extraction.

Description of drawings

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

Fig. 2 is the former vein image of embodiment 1;

Fig. 3 is the partition neighborhood template synoptic diagram that embodiment 1 selects;

Fig. 4 is a semicircle area relative region direction synoptic diagram in the neighborhood chosen of embodiment 1;

Fig. 5 a is the maximal intra-neighbor difference direction vector figure that embodiment 1 obtains;

Fig. 5 b is the maximal intra-neighbor difference vector value figure that embodiment 1 obtains;

Fig. 6 is the interior phasor difference figure of maximum neighborhood that embodiment 1 obtains;

Fig. 7 is the histogram of phasor difference figure in the maximum neighborhood that obtains of embodiment 1, and wherein A point is first maximum value, and the C point is first minimal value, flex point B of existence between A point and C point;

Fig. 8 is the veinprint synoptic diagram that embodiment 1 extracts the former vein image that obtains;

Fig. 9 a is a vein image; Fig. 9 b is the picture structure synoptic diagram that adopts Fig. 9 a of Niblack method extraction; Fig. 9 c is the picture structure synoptic diagram that adopts Fig. 9 a that extracts based on the method for MIND; Fig. 9 d is the picture structure synoptic diagram that adopts Fig. 9 a that extracts based on the method for DRDF; Fig. 9 e is the picture structure synoptic diagram that adopts Fig. 9 a of method extraction of the present invention; Fig. 9 f is the picture structure synoptic diagram through Fig. 9 a that adopts method extraction of the present invention after the pre-service again;

Figure 10 a is a vein image; Figure 10 b is the picture structure synoptic diagram that adopts Figure 10 a of Niblack method extraction; Figure 10 c is the picture structure synoptic diagram that adopts Figure 10 a that extracts based on the method for MIND; Figure 10 d is the picture structure synoptic diagram that adopts Figure 10 a that extracts based on the method for DRDF; Figure 10 e is the picture structure synoptic diagram that adopts Figure 10 a of method extraction of the present invention; Figure 10 f is the picture structure synoptic diagram through Figure 10 a that adopts method extraction of the present invention after the pre-service again;

Figure 11 a is a vein image; Figure 11 b is the picture structure synoptic diagram that adopts Figure 11 a of Niblack method extraction; Figure 11 c is the picture structure synoptic diagram that adopts Figure 11 a that extracts based on the method for MIND; Figure 11 d is the picture structure synoptic diagram that adopts Figure 11 a that extracts based on the method for DRDF; Figure 11 e is the picture structure synoptic diagram that adopts Figure 11 a of method extraction of the present invention; Figure 11 f is the picture structure synoptic diagram through Figure 11 a that adopts method extraction of the present invention after the pre-service again;

Figure 12 a is a vein image; Figure 12 b is the picture structure synoptic diagram that adopts Figure 12 a of Niblack method extraction; Figure 12 c is the picture structure synoptic diagram that adopts Figure 12 a that extracts based on the method for MIND; Figure 12 d is the picture structure synoptic diagram that adopts Figure 12 a that extracts based on the method for DRDF; Figure 12 e is the picture structure synoptic diagram that adopts Figure 12 a of method extraction of the present invention; Figure 12 f is the picture structure synoptic diagram through Figure 12 a that adopts method extraction of the present invention after the pre-service again.

Embodiment

The present invention is described in further detail below in conjunction with embodiment and accompanying drawing, but embodiments of the present invention are not limited thereto.

Embodiment 1

Use veinprint extracting method of the present invention and extract as shown in Figure 2 vein image structure, the process flow diagram of realization as shown in Figure 1:

The first step is got radius of neighbourhood R=7 (unit of R is a pixel), determines any pixel neighborhood of a point in the former vein image according to the radius of neighbourhood and distance function, and selects partition neighborhood template T as shown in Figure 3 _Div-neiIt is the adjacent fan-shaped stick mass at center with this pixel that described neighborhood is divided into 16 _i, i=1 ... 16, fan-shaped stick mass behind the piecemeal _iAnd mass _{Mod (i+7,16)+1}With this pixel is former point symmetry; Among Fig. 3, any selected pixel of 0 expression, the unit of each blockage is a pixel, pixel in 1～7 this neighborhood of pixel points of expression, pixel in the neighborhood is divided into 16 adjacent fan-shaped sticks, to show differentiation, the pixel number in each fan-shaped stick is basic identical with different shadings;

Wherein, distance function $d_{R\_\mathrm{Circle}}=\mathrm{Round}\left(\sqrt{{(x_1-x_2)}^2+{(y_1-y_2)}^2}\right)$

In the following formula, (x ₁, y ₁), (x ₂, y ₂) be coordinate any 2 in the former vein image;

In second step, calculate each fan-shaped stick mass in this pixel neighborhood of a point _iInterior all gray values of pixel points p (x, y) sum S _i=∑ p (x, y), pixel (x, y) ∈ mass wherein _i, i=1 ... 16, (x y) is pixel (x, gray-scale value y) to p;

Calculate semicircle zone block in this pixel neighborhood of a point _iIn all gray values of pixel points p (x, y) and semicircle complementary zone block with it _{I '}In all gray values of pixel points p (x, y) and between difference T _i:

$T_i=\frac{1}{8}(\underset{j=i}{\overset{i+7}{\mathrm{\&Sigma;}}}{S}_{\mathrm{mod}(j-\mathrm{1,16})+1}-\underset{j=i+8}{\overset{i+15}{\mathrm{\&Sigma;}}}{S}_{\mathrm{mod}(j-\mathrm{1,16})+1}),i=1\&CenterDot;\&CenterDot;\&CenterDot;16$

Wherein, semicircle zone block _iBe meant from fan-shaped stick mass _i, i=1 ... 16 beginnings are clockwise to fan-shaped stick mass _{Mod (i+6,16)+1}8 continuums that continuous fan-shaped stick is formed; Its complementary semicircle zone block _{I '}(i '=mod (i+7,16)+1) be meant from fan-shaped stick mass _{I '}8 continuums that continuous fan-shaped stick is formed that beginning is calculated clockwise;

The 3rd step is at two tuples vector l=[(T ₁, direc ₁), (T ₂, direc ₂) ... (T ₁₆, direc ₁₆)], T _i+ T _{Mod (i+7,16)+1}Select the maximum of T in all two tuple elements in=0 _Max=max (T _i), and obtain maximum of T thus _MaxCorresponding direction ${\mathrm{direc}}_{T_{\max }}={\mathrm{direc}}_i\left(i\right|T_i=T_{\max }),$ T _MaxAnd two corresponding direction in the tuple ${\mathrm{direc}}_{T_{\max }}={\mathrm{direc}}_i\left(i\right|T_i=T_{\max })$ Be the maximal intra-neighbor difference vector value and the direction vector thereof of the pixel of asking;

Wherein, vectorial l is by 16 two tuple (T _i, direc _i) constitute T _iBe semicircle zone block _iIn all gray values of pixel points p (x, y) and semicircle complementary zone block with it _{I '}In all gray values of pixel points p (x, y) and between difference, direc _iBe semicircle zone block _iThe corresponding area direction, i=1 ... 16; Region direction direc _iBe meant and semicircle zone block _iDiameter vertical, and point to semicircle zone block _iThe reciprocal direction of semi arch, as shown in Figure 4,1 expression region direction direc ₁, 2 expression region direction direc ₂, 3 expression region direction direc ₃16 expression region direction direc ₁₆

The 4th step repeated above-mentioned three steps of the first step to the, obtained the maximal intra-neighbor difference vector value and the direction vector thereof of all pixels in the former vein image, and the maximal intra-neighbor difference vector that promptly obtains former vein image is shown in Fig. 5 a and Fig. 5 b.

The 5th step, determine any pixel neighborhood of a point in the described maximal intra-neighbor difference vector according to radius of neighbourhood R=7 (unit of R is a pixel), select partition neighborhood template T _Div-neiIt is the adjacent fan-shaped stick mass at center with this pixel that described neighborhood is divided into 16 _iCalculate each fan-shaped stick mass respectively _iInterior all vector v (x, y) sum S _i ^v:

$S_i^v=\mathrm{\&Sigma;v}(x,y),(x,y)\&Element;{\mathrm{mass}}_i,i=1\&CenterDot;\&CenterDot;\&CenterDot;16;$

And calculate each semicircle zone block _iInterior all vector v (x, y) sum S _i ^b,

$S_i^b=\frac{1}{8}\underset{j=i}{\overset{i+7}{\mathrm{\&Sigma;}}}{S}_{\mathrm{mod}(j-\mathrm{1,16})+1}^v,i=1\&CenterDot;\&CenterDot;\&CenterDot;16;$

Then with 16 S _i ^bBe divided into 8 pairs, each is to S _i ^bAnd S _{I '} ^bPairing semicircle zone block _iAnd block _{I '}8 couples of T are calculated in complementation then _i ^v, T _i ^vSize be S _i ^bAnd S _{I '} ^bThe mould value T of the two phasor difference _i ^m, its symbol is according to sign function sgn (S _i ^b, S _{I '} ^b) decision,

$T_i^v=\mathrm{sign}(S_i^b,S_{i^{\&prime;}}^b)\&times;T_i^m,i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

$T_i^m=|S_i^b-S_{i^{\&prime;}}^b|,i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

$\mathrm{sgn}(S_i^b,S_{i^{\&prime;}}^b)=\left\{\begin{array}{c}1,|\mathrm{angle}\left(S_i^b\right)-{\mathrm{direc}}_i|<\mathrm{\&pi;}/2\&cap;|\mathrm{angle}\left(S_{i^{\&prime;}}^b\right)-{\mathrm{direc}}_{i^{\&prime;}}|<\mathrm{\&pi;}/2\\ -1,\mathrm{else}\end{array}\right.,i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

Calculate maximum of T _Max ^v, be phasor difference in the maximum neighborhood of this pixel,

$T_{\max }^v=\max \left(T_i^v\right),i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

Calculate the interior phasor difference of maximum neighborhood of all pixels in the maximal intra-neighbor difference vector, promptly obtain the interior phasor difference figure (as shown in Figure 6) of maximum neighborhood of former vein image; Wherein, partition neighborhood template T _Div-nei1., make fan-shaped stick mass behind the piecemeal select as follows: _iAnd mass _{Mod (i+7,16)+1}With this pixel is former point symmetry; 2., make each fan-shaped stick mass behind the piecemeal _iIn comprise pixel number equate to greatest extent; Semicircle zone block _iBe meant from fan-shaped stick mass _iBeginning is clockwise to fan-shaped stick mass _{Mod (i+6,16)+1}8 continuums that continuous fan-shaped stick is formed; Its complementary semicircle zone block _{I '}Be meant from fan-shaped stick mass _{I '}8 continuums that continuous fan-shaped stick is formed that beginning is calculated clockwise; I=1 ... 16, i '=mod (i+7,16)+1.

The 6th step, obtain the histogram of phasor difference figure in the maximum neighborhood, value with the flex point C correspondence between first maximum point A in the histogram (as Fig. 7) and first minimum point B, extract threshold value a as pattern, then with the pixel of gray-scale value 〉=a among the phasor difference figure in the described maximum neighborhood all assignment be 255, the pixel of gray-scale value＜a all assignment is 0

$P_{\mathrm{pattern}}(i,j)=\left\{\begin{array}{c}0,P_{\mathrm{GMINVD}}(i,j)<a\\ 255,P_{\mathrm{GMINVD}}(i,j)\&GreaterEqual;a\end{array}\right.$

Promptly obtain not carrying out as yet the veinprint that smoothing denoising is handled, as shown in Figure 8.

Embodiment 2

More existing Niblack method, the effect of the vein image shown in Fig. 9 a being extracted based on the method for MIND, based on method and the veinprint extracting method disclosed by the invention of DRDF, in order more clearly to compare to extracting the result, we not to the image after extracting carry out smoothly, subsequent treatment such as denoising.Fig. 9 b is the picture structure synoptic diagram that adopts Fig. 9 a of Niblack method extraction; Fig. 9 c is the picture structure synoptic diagram that adopts Fig. 9 a that extracts based on the method for MIND; Fig. 9 d is the picture structure synoptic diagram that adopts Fig. 9 a that extracts based on the method for DRDF; Fig. 9 e is the picture structure synoptic diagram (step of extraction is with embodiment 1) that adopts Fig. 9 a of method extraction of the present invention; For relatively preprocessor is to the influence of the effect of method of the present invention, Fig. 9 f is the picture structure synoptic diagram through Fig. 9 a that adopts method extraction of the present invention after the pre-service again; By result we as can be seen: with respect to the Niblack method with based on for the method for MIND, more clear based on the veinprint that method and the method disclosed by the invention of DRDF are extracted, also more meet veinprint actual in the original image, noise is less, and with respect to for the method for DRDF, the veinprint that method disclosed by the invention is extracted has been eliminated the pseudo-lines that brings based on the method for DRDF and (has been seen Fig. 9 d, the rectangle tab area is that many noises are concentrated the pseudo-lines that causes between close lines among the figure), and noise is still less.In addition, in order to compare the influence of pre-service to method disclosed by the invention, adopt a kind of image enchancing method commonly used earlier: contrast-limited self-adapting histogram equilibrium (Contrast Limited Adaptive Histogram Equalization, CLAHE) method is carried out pre-service to vein image shown in Fig. 9 a, and then use method of the present invention that it is carried out veinprint and extract, experimental result is seen Fig. 9 f, comparison diagram 9e and Fig. 9 f, carry out earlier as can be seen re-using method of the present invention and directly using the not obviously difference of this paper method extraction veinprint after the pre-service, therefore method of the present invention is applicable to the extraction of vein image directly being carried out veinprint.

Embodiment 3

More existing Niblack method, the effect of the vein image shown in Figure 10 a being extracted based on the method for MIND, based on method and the veinprint extracting method disclosed by the invention of DRDF, in order more clearly to compare to extracting the result, we not to the image after extracting carry out smoothly, subsequent treatment such as denoising.Figure 10 b is the picture structure synoptic diagram that adopts Figure 10 a of Niblack method extraction; Figure 10 c is the picture structure synoptic diagram that adopts Figure 10 a that extracts based on the method for MIND; Figure 10 d is the picture structure synoptic diagram that adopts Figure 10 a that extracts based on the method for DRDF; Figure 10 e is the picture structure synoptic diagram (step of extraction is with embodiment 1) that adopts Figure 10 a of method extraction of the present invention; For relatively preprocessor is to the influence of the effect of method of the present invention, Figure 10 f is the picture structure synoptic diagram through Figure 10 a that adopts method extraction of the present invention after the pre-service again; By result we as can be seen: with respect to the Niblack method with based on for the method for MIND, more clear based on the veinprint that method and the method disclosed by the invention of DRDF are extracted, also more meet veinprint actual in the original image, noise is less, and with respect to for the method for DRDF, the veinprint that method disclosed by the invention is extracted has been eliminated the pseudo-lines that brings based on the method for DRDF and (has been seen Figure 10 d, circle markings zone is concentrated for noise and is located the pseudo-lines that causes among the figure), and noise is still less.In addition, in order to compare the influence of pre-service to method disclosed by the invention, adopt a kind of image enchancing method commonly used earlier: contrast-limited self-adapting histogram equilibrium (Contrast Limited Adaptive Histogram Equalization, CLAHE) method is carried out pre-service to vein image shown in Figure 10 a, and then use method of the present invention that it is carried out veinprint and extract, experimental result is seen Figure 10 f, comparison diagram 10e and Figure 10 f, carry out earlier as can be seen re-using method of the present invention and directly using not obviously difference of method extraction veinprint of the present invention after the pre-service, therefore method of the present invention is applicable to the extraction of vein image directly being carried out veinprint.

Embodiment 4

More existing Niblack method, the effect of the vein image shown in Figure 11 a being extracted based on the method for MIND, based on method and the veinprint extracting method disclosed by the invention of DRDF, in order more clearly to compare to extracting the result, we not to the image after extracting carry out smoothly, subsequent treatment such as denoising.Figure 11 b is the picture structure synoptic diagram that adopts Figure 11 a of Niblack method extraction; Figure 11 c is the picture structure synoptic diagram that adopts Figure 11 a that extracts based on the method for MIND; Figure 11 d is the picture structure synoptic diagram that adopts Figure 11 a that extracts based on the method for DRDF; Figure 11 e is the picture structure synoptic diagram (step of extraction is with embodiment 1) that adopts Figure 11 a of method extraction of the present invention; For relatively preprocessor is to the influence of the effect of method of the present invention, Figure 11 f is the picture structure synoptic diagram through Figure 11 a that adopts method extraction of the present invention after the pre-service again; By result we as can be seen: with respect to the Niblack method with based on for the method for MIND, more clear based on the veinprint that method and the method disclosed by the invention of DRDF are extracted, also more meet veinprint actual in the original image, noise is less, and with respect to for the method for DRDF, the veinprint that method of the present invention is extracted has been eliminated the pseudo-lines that brings based on the method for DRDF and (has been seen Figure 11 d, circle markings zone is concentrated for noise and is located the pseudo-lines that causes among the figure), and noise is still less.In addition, in order to compare the influence of pre-service to method disclosed by the invention, adopt a kind of image enchancing method commonly used earlier: contrast-limited self-adapting histogram equilibrium (Contrast Limited Adaptive Histogram Equalization, CLAHE) method is carried out pre-service to vein image shown in Figure 11 a, and then use method of the present invention that it is carried out veinprint and extract, experimental result is seen Figure 11 f, comparison diagram 11e and Figure 11 f, carry out earlier as can be seen re-using method of the present invention and directly using not obviously difference of method extraction veinprint of the present invention after the pre-service, therefore method of the present invention is applicable to the extraction of vein image directly being carried out veinprint.

Embodiment 5

More existing Niblack method, the effect of the vein image shown in Figure 12 a being extracted based on the method for MIND, based on method and the veinprint extracting method disclosed by the invention of DRDF, in order more clearly to compare to extracting the result, we not to the image after extracting carry out smoothly, subsequent treatment such as denoising.Figure 12 b is the picture structure synoptic diagram that adopts Figure 12 a of Niblack method extraction; Figure 12 c is the picture structure synoptic diagram that adopts Figure 12 a that extracts based on the method for MIND; Figure 12 d is the picture structure synoptic diagram that adopts Figure 12 a that extracts based on the method for DRDF; Figure 12 e is the picture structure synoptic diagram (step of extraction is with embodiment 1) that adopts Figure 12 a of method extraction of the present invention; For relatively preprocessor is to the influence of the effect of method of the present invention, Figure 12 f is the picture structure synoptic diagram through Figure 12 a that adopts method extraction of the present invention after the pre-service again; By result we as can be seen: with respect to the Niblack method with based on for the method for MIND, more clear based on the veinprint that method and the method disclosed by the invention of DRDF are extracted, also more meet veinprint actual in the original image, noise is less, and with respect to for the method for DRDF, the veinprint that method of the present invention is extracted has been eliminated the pseudo-lines that brings based on the method for DRDF and (has been seen Figure 12 d, the rectangle tab area is the pseudo-lines that noise concentrates the place to cause among the figure), and noise is still less.In addition, in order to compare the influence of pre-service to method disclosed by the invention, adopt a kind of image enchancing method commonly used earlier: contrast-limited self-adapting histogram equilibrium (Contrast Limited Adaptive Histogram Equalization, CLAHE) method is carried out pre-service to vein image shown in Figure 12 a, and then use method of the present invention that it is carried out veinprint and extract, experimental result is seen Figure 12 f, comparison diagram 12e and Figure 12 f, carry out earlier as can be seen re-using method of the present invention and directly using not obviously difference of method extraction veinprint of the present invention after the pre-service, therefore method of the present invention is applicable to the extraction of vein image directly being carried out veinprint.

The foregoing description is a preferred implementation of the present invention, but embodiments of the present invention are not restricted to the described embodiments, and for example radius of neighbourhood R gets other value in 1≤R≤11 scopes or the like, all can realize the present invention.Other any do not deviate from change, the modification done under spirit of the present invention and the principle, substitutes, combination, simplify, and all should be the substitute mode of equivalence of the present invention, is included within protection scope of the present invention.

Claims (6)

1. the veinprint extracting method based on maximal intra-neighbor difference vector is characterized in that, comprises the steps:

The first step calculates the maximal intra-neighbor difference and the direction thereof of all pixels in the former vein image, obtains the maximal intra-neighbor difference vector of former vein image;

Second goes on foot, and calculates the interior phasor difference of maximum neighborhood of all pixels in the described maximal intra-neighbor difference vector, obtains the interior phasor difference figure of maximum neighborhood of former vein image;

The 3rd step, obtain the histogram of phasor difference figure in the described maximum neighborhood, from described histogram, obtain threshold value, and extract the veinprint of former vein image with described threshold value phasor difference figure in the maximum neighborhood.

2. the veinprint extracting method based on maximal intra-neighbor difference vector according to claim 1, it is characterized in that, maximal intra-neighbor difference that calculates all pixels in the former vein image and direction thereof in the described first step, obtain the maximal intra-neighbor difference vector of former vein image, comprise the steps:

In the A step, determine any pixel neighborhood of a point in the former vein image according to the radius of neighbourhood and distance function, and select partition neighborhood template T _Div-neiIt is the adjacent fan-shaped stick mass at center with this pixel that described neighborhood is divided into 16 _iI=1 wherein ... 16, radius of neighbourhood R:1≤R≤11, the unit of R is a pixel; Distance function adopts the Euclidean distance that rounds up: $d_{R\_\mathrm{Circle}}=\mathrm{Round}\left(\sqrt{{(x_1-x_2)}^2+{(y_1-y_2)}^2}\right),$ In the formula, (x ₁, y ₁), (x ₂, y ₂) be coordinate any 2 in the former vein image;

In the B step, calculate each fan-shaped stick mass in this pixel neighborhood of a point _iInterior all gray values of pixel points p (x, y) sum S _i=∑ p (x, y); Pixel (x, y) ∈ mass wherein _i, (x y) is pixel (x, gray-scale value y), i=1 to p ... 16;

Calculate semicircle zone block in this pixel neighborhood of a point _iIn all gray values of pixel points p (x, y) and semicircle complementary zone block with it _{I '}In all gray values of pixel points p (x, y) and between difference T _i:

$T_i=\frac{1}{8}(\underset{j=i}{\overset{i+7}{\mathrm{\&Sigma;}}}{S}_{\mathrm{mod}(j-\mathrm{1,16})+1}-\underset{j=i+8}{\overset{i+15}{\mathrm{\&Sigma;}}}{S}_{\mathrm{mod}(j-\mathrm{1,16})+1})$

Wherein, semicircle zone block _iBe meant from fan-shaped stick mass _iBeginning is clockwise to fan-shaped stick mass _{Mod (i+6,16)+1}8 continuums that continuous fan-shaped stick is formed; Its complementary semicircle zone block _{I '}Be meant from fan-shaped stick mass _{I '}8 continuums that continuous fan-shaped stick is formed that beginning is calculated clockwise; I=1 ... 16, i '=mod (i+7,16)+1;

The C step is at two tuples vector l=[(T ₁, direc ₁), (T ₂, direc ₂) ... (T ₁₆, direc ₁₆)], T _i+ T _{Mod (i+7,16)+1}Select the maximum of T in all two tuple elements in=0 _Max=max (T _i), and obtain maximum of T thus _MaxCorresponding direction ${\mathrm{direc}}_{T_{\max }}={\mathrm{direc}}_i\left(i\right|T_i=T_{\max }),$ T _MaxAnd two corresponding direction in the tuple ${\mathrm{direc}}_{T_{\max }}={\mathrm{direc}}_i\left(i\right|T_i=T_{\max })$ Be the maximal intra-neighbor difference vector value and the direction vector thereof of the pixel of asking;

Wherein, vectorial l is by 16 two tuple (T _i, direc _i) constitute; T _iBe semicircle zone block _iIn all gray values of pixel points p (x, y) and semicircle complementary zone block with it _{I '}In all gray values of pixel points p (x, y) and between difference, direc _iBe semicircle zone block _iThe corresponding area direction, i=1 ... 16, i '=mod (i+7,16)+1; Region direction direc _iBe meant and semicircle zone block _iDiameter vertical, and point to semicircle zone block _iThe direction of semi arch direction;

The D step repeats above-mentioned A and goes on foot the step to C, obtains the maximal intra-neighbor difference vector value and the direction vector thereof of all pixels in the former vein image, promptly obtains the maximal intra-neighbor difference vector of former vein image.

3. the veinprint extracting method based on maximal intra-neighbor difference vector according to claim 2 is characterized in that, the partition neighborhood template T of described A in the step _Div-nei1., make fan-shaped stick mass behind the piecemeal select as follows: _iAnd mass _{Mod (i+7,16)+1}With this pixel is former point symmetry; 2., make each fan-shaped stick mass behind the piecemeal _iIn comprise pixel number equate to greatest extent.

4. the veinprint extracting method based on maximal intra-neighbor difference vector according to claim 1, it is characterized in that, calculate the interior phasor difference of maximum neighborhood of all pixels of described maximal intra-neighbor difference vector in described second step, obtain the interior phasor difference figure of maximum neighborhood of former vein image, comprise the steps:

The I step, determine any pixel neighborhood of a point in the described maximal intra-neighbor difference vector according to the radius of neighbourhood, select partition neighborhood template T _Div-neiIt is the adjacent fan-shaped stick mass at center with this pixel that described neighborhood is divided into 16 _iCalculate each fan-shaped stick mass respectively _iInterior all vector v (x, y) sum S _i ^v:

$S_i^v=\mathrm{\&Sigma;v}(x,y),(x,y)\&Element;{\mathrm{mass}}_i,i=1\&CenterDot;\&CenterDot;\&CenterDot;16;$

And calculate each semicircle zone block _iInterior all vector v (x, y) sum S _i ^b,

$S_i^b=\frac{1}{8}\underset{j=i}{\overset{i+7}{\mathrm{\&Sigma;}}}{S}_{\mathrm{mod}(J-\mathrm{1,16})+1}^v,i=1\&CenterDot;\&CenterDot;\&CenterDot;16;$

The II step is with 16 S _i ^bBe divided into 8 pairs, each is to S _i ^bAnd S _{I '} ^bPairing semicircle zone block _iAnd block _{I '}8 couples of T are calculated in complementation then _i ^v, T _i ^vSize be S _i ^bAnd S _{I '} ^bThe mould value T of the two phasor difference _i ^m, its symbol is according to sign function sgn (S _i ^b, S _{I '} ^b) decision,

$T_i^v=\mathrm{sign}(S_i^b,S_{i^{\&prime;}}^b)\&times;T_i^m,i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

$T_i^m=|S_i^b-S_{i^{\&prime;}}^b|,i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

$\mathrm{sgn}(S_i^b,S_{i^{\&prime;}}^b)=\left\{\begin{array}{cc}1,& |\mathrm{angle}\left(S_i^b\right)-{\mathrm{direc}}_i|<\mathrm{\&pi;}/2\&cap;|\mathrm{angle}\left(S_{i^{\&prime;}}^b\right)-{\mathrm{direc}}_{i^{\&prime;}}|<\mathrm{\&pi;}/2\\ -1,& \mathrm{else}\end{array}\right.,i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

In the III step, calculate maximum of T _Max ^v, be phasor difference in the maximum neighborhood of this pixel,

$T_{\max }^v=\max \left(T_i^v\right),i=1\&CenterDot;\&CenterDot;\&CenterDot;8$

IV goes on foot, and repeats the above-mentioned I step to the III step, calculates the interior phasor difference of maximum neighborhood of all pixels in the maximal intra-neighbor difference vector, promptly obtains the interior phasor difference figure of maximum neighborhood of former vein image;

Wherein, radius of neighbourhood R:1≤R≤11, the unit of R is a pixel; Partition neighborhood template T _Div-nei1., make fan-shaped stick mass behind the piecemeal select as follows: _iAnd mass _{Mod (i+7,16)+1}With this pixel is former point symmetry; 2., make each fan-shaped stick mass behind the piecemeal _iIn comprise pixel number equate to greatest extent; Semicircle zone block _iBe meant from fan-shaped stick mass _iBeginning is clockwise to fan-shaped stick mass _{Mod (i+6,16)+1}8 continuums that continuous fan-shaped stick is formed; Its complementary semicircle zone block _{I '}Be meant from fan-shaped stick mass _{I '}8 continuums that continuous fan-shaped stick is formed that beginning is calculated clockwise; I=1 ... 16, i '=mod (i+7,16)+1.

5. the veinprint extracting method based on maximal intra-neighbor difference vector according to claim 1, it is characterized in that, in described the 3rd step in the maximum neighborhood histogram of phasor difference figure obtain lines and extract threshold value, and be meant: with the value of the flex point correspondence between first maximum point and first minimum point the histogram of phasor difference figure in the described maximum neighborhood with the veinprint that described threshold value phasor difference figure in maximum neighborhood extracts former vein image, extract threshold value a as lines, then with the pixel of gray-scale value 〉=a among the phasor difference figure in the described maximum neighborhood all assignment be 255, the pixel of gray-scale value＜a all assignment is 0

$P_{\mathrm{pattern}}(i,j)=\left\{\begin{array}{c}0,P_{\mathrm{GMINVD}}(i,j)<a\\ 255,P_{\mathrm{GMINVD}}(i,j)\&GreaterEqual;a\end{array}\right.$

Promptly obtain not carrying out as yet the veinprint that smoothing denoising is handled.

6. according to claim 2 or 4 described veinprint extracting method, it is characterized in that described radius of neighbourhood R=7 based on maximal intra-neighbor difference vector.

Images