CN101847202B - Correction algorithm for image distortion of optical fingerprint collector - Google Patents

Abstract

The invention relates to a correction algorithm for the image distortion of an optical fingerprint collector. Fingerprint images acquired by a traditional optical fingerprint collector more or less have different degrees of image distortion which is non-linear and easy to influence the matching effect. The correction algorithm for the image distortion of an optical fingerprint collector, which is disclosed in the invention, finishes image correction by deducing an image distortion function through a corresponding relation between distorted images acquired on the optical collector of square grids and grid control points of square grid images. Linear and non-linear distorted fingerprint images acquired by the optical fingerprint collector are corrected by the invention so that the same fingerprint can be accurately matched under homologous and nonhomologous conditions, and fingerprint images are propelled to develop towards a standardization direction.

Description

The correcting algorithm of image distortion of optical fingerprint collector

Technical field

The present invention relates to IMAQ and coupling in the automatic system of fingerprint recognition, belong to Digital Image Processing and area of pattern recognition, what relate generally to is the correcting algorithm of image distortion of optical fingerprint collector.Be suitable for optically detecting device image and the matching process of plane stamp fingerprint image in the computing machine Automated Fingerprint Identification System.

Background technology

Automated Fingerprint Identification System (AFIS) conduct-kind of important biological characteristic authentication technique is widely used on market, and particularly at public safety field, constantly the market demand has further promoted the development of AFIS technology.The precondition of fingerprint recognition is obtained the digital finger-print image exactly, is mainly accomplished by various fingerprint capturers.Present fingerprint capturer mainly contains optical fingerprint collector, semiconductor fingerprint collector, ultrasound wave fingerprint capturer three major types.Advantages such as optical fingerprint collector is good with stability, bearing temperature changes, wear-resistant, with low cost become the main force's product on the market; The semiconductor fingerprint collector comprises three major types such as silicon electric capacity fingerprint image sensor, semiconductor pressure-sensitive sensor and conductor temperature induction pick-up; Its advantage is that the fingerprint image that obtains is more accurate; But the semiconductor fingerprint collector generally receives the influence of static easily; Fingerprint is responsive, not wear-resistant to doing, wetting; The ultrasound wave fingerprint capturer is because its cost is too high, also rare in AFIS at present.

Optical fingerprint collector receives the influence of optical imagery mechanism, all exists image deformation to a certain degree, mainly contains trapezoidal deformation and rectangle deformation.Present fingerprint capturer alignment technique also carries out to both of these case.In fact, the deformation of the fingerprint image that optical fingerprint collector obtains is far above these two kinds of line styles and changes, and has non-linear significantly distortion near the edge of image.Receive the influence of scalloping, same optical fingerprint collector when gathering unidirectional fingerprint and mating, can be accomplished coupling; Same optical fingerprint collector just can't mate the fingerprint of different directions exactly; The fingerprint image of image that optical fingerprint collector obtains and plane stamp then can't mate at all.

For these reasons; Do not gathered the aspect effect of fingerprint in order to make optical fingerprint collector; For the fingerprint image that various fingerprint collecting equipment are obtained has versatility; Just need proofread and correct, thereby realize that same fingerprint under homology and nonhomologous situation, can both accomplish coupling accurately the fingerprint image of various line styles and nonlinear deformation.

Summary of the invention

The object of the invention promptly produces thus, proposes a kind of correcting algorithm of image distortion of optical fingerprint collector, has solved the matching problem that optical fingerprint collector anamorphose causes.Through the line style that optical fingerprint collector is obtained and the correcting algorithm of nonlinear distortion fingerprint image, make same fingerprint under homology and nonhomologous situation, all can realize accurate match.

The present invention realizes that the technical scheme that above-mentioned purpose is taked is: utilize square net taking pictures or scanning on the image that obtains; On the image of gathering with optical fingerprint collector, write down the apex coordinate and the corresponding relation of each grid respectively, the anisotropic of fault image distortion problem; Decomposition is in lattice; Handle according to linear distortion, and then the coordinate of arbitrfary point in fault image in the derivation square net, concrete steps are:

A) standard picture that forms of the square net of certain density δ be I (i, j), its size is m * n, writes down and takes pictures or the dpi of scanning collection, four of each grid summits are the reference mark of distortion correction in the square net;

B) align fingerprint optically detecting device in the horizontal and vertical directions with square net; Collection fault image J (i, j), its size is u * v; Picture format is converted into grayscale mode; (i j) has consistent direction and identical dpi, and the summit of each grid is the reference mark of distortion correction in the fault image with image I in maintenance;

C) registering images I (i, j) (i, central point j) write down the coordinate at each corresponding grid reference mark in two images respectively with image J;

D) (i, the lattice reference mark coordinate in j) is respectively P to establish image I ₁(x ₁, y ₁), P ₂(x ₂, y ₂), P ₃(x ₃, y ₃), P ₄(x ₄, y ₄), so, for image I (i, j) in any 1 P in the square lattice ₀(x ₀, y ₀), the ratio that projects on the line segment in level and vertical direction is respectively α and β, can obtain:

(i, j) the control corresponding point is respectively P to image J ₁' (x ' ₁, y ' ₁), P ₂' (x ' ₂, y ' ₂), P ₃' (x ' ₃, y ' ₃), P ₄' (x ' ₄, y ' ₄), some P ₀(x ₀, y ₀) at image J (i, the corresponding point P in j) ₀(x ₀, y ₀), have too: $\mathrm{\α}=\frac{x_0^{\′}-x_1^{\′}}{x_2^{\′}-x_1^{\′}}=\frac{x_0^{\′}-x_4^{\′}}{x_3^{\′}-x_4^{\′}};$ $\mathrm{\β}=\frac{y_0^{\′}-y_1^{\′}}{y_4^{\′}-y_1^{\′}}=\frac{y_0^{\′}-y_2^{\′}}{y_3^{\′}-y_2^{\′}};$

E) P ₀(x ₀, y ₀) in the horizontal direction projection ratio α and in the projection ratio beta of vertical direction, with P ₀(x ₀, y ₀) in the horizontal direction the projection ratio and equate respectively in the projection ratio of vertical direction, can solve the distortion function equation and be:

$\left\{\begin{array}{c}{x}_0^{\′}=\frac{(x_d-x_c)((y_b-y_a)x_a-(x_d-x_c)y_a)-(x_b-x_a)((y_c-y_c)x_c-(x_d-x_c)y_c)}{(y_b-y_a)(x_d-x_c)-(y_b-y_c)(x_b-x_a)}\\ y_0^{\′}=\frac{(y_b-y_a)((x_d-x_c)y_c-(y_d-y_c)x_c)-(y_d-y_c)((x_b-x_a)y_a-(y_b-y_a)x_a)}{(y_b-y_a)(x_d-x_c)-(y_d-y_c)(x_b-x_a)}\end{array}\right.;$

Wherein,

$\left\{\begin{array}{c}{x}_a=x_1^{\′}+\frac{(x_0-x_1)(x_2^{\′}-x_1^{\′})}{x_2-x_1}\\ x_b=x_3^{\′}+\frac{(x_0-x_3)(x_4^{\′}-x_3^{\′})}{x_4-x_3}\\ x_c=x_4^{\′}+\frac{(x_0-x_4)(x_1^{\′}-x_4^{\′})}{x_1-x_4}\\ x_d=x_2^{\′}+\frac{(x_0-x_2)(x_3^{\′}-x_2^{\′})}{x_3-x_2}\\ y_a=y_1^{\′}+\frac{(y_0-y_1)(y_2^{\′}-y_1^{\′})}{y_2-y_1}\\ y_b=y_3^{\′}+\frac{(y_0-y_3)(y_4^{\′}-y_3^{\′})}{y_4-y_3}\\ y_c=y_4^{\′}+\frac{(y_0-y_4)(y_1^{\′}-y_4^{\′})}{y_1-y_4}\\ y_d=y_2^{\′}+\frac{(y_0-y_2)(y_3^{\′}-y_2^{\′})}{y_3-y_2}\end{array}\right.$

The present invention makes same fingerprint under homology and nonhomologous situation, all can realize accurate match through the line style that optical fingerprint collector is obtained and the correction of nonlinear distortion fingerprint image.Arbitrary dimension, any optical fingerprint collector image of deformation be can proofread and correct, producer, the specification of collector, the restriction of type do not received.Pass through experiment confirm; Algorithm of the present invention has solved the matching problem that optical fingerprint collector anamorphose etc. causes; Can be with the fingerprint image and the plane stamp scan fingerprint image of the optical fingerprint collector collection that originally can't mate; Mate exactly, promotion fingerprint collecting image is developed to the standardization direction has positive meaning.

Description of drawings

Fig. 1 is image distortion of optical fingerprint collector correcting principle figure of the present invention;

Fig. 2 is the reference mark synoptic diagram that square net node of the present invention forms;

Fig. 3 is an image distortion of optical fingerprint collector synoptic diagram of the present invention;

Fig. 4-the 5th, the rectification effect of the fingerprint image that optical fingerprint collector of the present invention is gathered.

Embodiment

Combine accompanying drawing that the present invention is further described at present:

The hardware environment that is used for embodiment of the present invention is: the optical fingerprint collector that CAMA-e Shenzhen company produces, ACPIMultiprocessor PC-4400+ computing machine, 2.00GB internal memory, 64M video card; The software environment of operation is: Window XP, can realize the method that the present invention proposes with C and MATLAB 6.5 programming languages.

1. design the image dsitortion reference mark

The distortion of fingerprint image need be accomplished through the distortion parameter of measuring optical fingerprint collector, and the distribution of fingerprint feature point has randomness, need substitute measurement with square net.

Align fingerprint optically detecting device in the horizontal and vertical directions with square net, (i, j), its size is m * n, picture format is converted into grayscale mode, the dpi of record acquisition device to gather fault image I.

The gray level image that square net forms be J (i, j), its size be u * v, keep and image I (i j) has the direction and identical dpi of unanimity.

(i, j) (i, central point j) write down the coordinate at each corresponding grid reference mark in two images respectively to registering images I with image J.

2. find the solution distortion function

Square net is meant the square net with certain density δ.Its effect is: (i, nonlinear distortion problem j) are decomposed in lattice, handle according to linear distortion image I.Four summits of each lattice are exactly the reference mark, image J (i, j) in, corresponding reference mark is arranged.And then the derivation distortion function, the correcting process of completion nonlinear distortion image.

2.1 the linear transformation of square net image

If (i, the lattice reference mark coordinate in j) is respectively P to image J ₁(x ₁, y ₁), P ₂(x ₂, y ₂), P ₃(x ₃, y ₃), P ₄(x ₃, y ₃), so, for image J (i, j) in any 1 P in the square lattice ₀(x ₀, y ₀), be respectively α and β at horizontal line section with ratio on the vertical line segment, then:

$\mathrm{\α}=\frac{x_0-x_1}{x_2-x_1}=\frac{x_0-x_4}{x_3-x_4};$

$\mathrm{\β}=\frac{y_0-y_1}{y_4-y_1}=\frac{y_0-y_2}{y_3-y_2};$

If (j) there is not nonlinear deformation in i to image I, and (i, (i, j) the control corresponding point is respectively P to image I with image J in j) ₁' (x ' ₁, y ' ₁), P ₂' (x ' ₂, y ' ₂), P ₃' (x ' ₃, y ' ₃), P ₄' (x ' ₄, y ' ₄).Point P ₀(x ₀, y ₀) (i exists in the square lattice in j) and hints obliquely at coordinate points P in image I ₀' (x ' ₀, y ' ₀), then:

$\mathrm{\α}=\frac{x_0^{\′}-x_1^{\′}}{x_2^{\′}-x_1^{\′}}=\frac{x_0^{\′}-x_4^{\′}}{x_3^{\′}-x_4^{\′}};$

$\mathrm{\β}=\frac{y_0^{\′}-y_1^{\′}}{y_4^{\′}-y_1^{\′}}=\frac{y_0^{\′}-y_2^{\′}}{y_3^{\′}-y_2^{\′}};$

Can try to achieve the linear transformation function is:

$\left\{\begin{array}{c}{x}_0^{\′}=\frac{x_1^{\′}(x_2-x_0)+x_2^{\′}(x_0-x_1)}{x_2-x_1}\\ y_0^{\′}=\frac{y_1^{\′}(y_4-y_0)+y_4^{\′}(y_0-y_1)}{y_4-y_1}\end{array}\right.;$

In real the application, P ₁(x ₁, y ₁), P ₂(x ₂, y ₂), P ₃(x ₃, y ₃), P ₄(x ₃, y ₃) and P ₁' (x ' ₁, y ' ₁), P ₂' (x ' ₂, y ' ₂), P ₃' (x ' ₃, y ' ₃), P ₄' (x ' ₄, y ' ₄) can round four summits of width of cloth image, the linear transformation function can be reduced to: $\left\{\begin{array}{c}{x}_0^{\′}=\frac{x_0\×u}{m}\\ y_0^{\′}=\frac{y_0\×v}{n}\end{array}\right.;$

2.2 the nonlinear transformation of square net image

If (j) there is nonlinear deformation in i to image I, and (i hints obliquely at a P in j) to image I ₀' (x ' ₀, y ' ₀) be a some P ₁' (x ' ₁, y ' ₁) and P ₂' (x ' ₂, y ' ₂) ratio is point and the P of α in the straight-line equation ₃' (x ' ₃, y ' ₃) and P ₄' (x ' ₄, y ' ₄) ratio is the straight line that the point of α forms, with P ₁' (x ' ₁, y ' ₁) and P ₃' (x ' ₂, y ' ₂) ratio is point and the P of β in the straight-line equation ₂' (x ' ₃, y ' ₃) and P ₄' (x ' ₄, y ' ₄) ratio is the intersection point of the straight line that forms of the point of β.Its parametric equation can be expressed as:

P ₀′(x′ ₀，y′ ₀)＝f(P _{1，2，3，4}(x _{1，2，3，4}，y _{1，2，3，4})，P _{1，2，3，4}′(x′ _{1，2，3，4}，y′ _{1，2，3，4})，P ₀(x ₀，y ₀)，P ₀′(x′ ₀，y′ ₀))

Can try to achieve distortion function according to the straight-line equation formula:

$\left\{\begin{array}{c}{x}_0^{\′}=\frac{((y_c-y_a)x_a-(x_c-x_a)y_a)(x_b-x_d)-(x_a-x_c)((y_d-y_b)x_b-(x_d-x_b)y_b)}{(y_d-y_a)(x_b-x_d)-(y_d-y_b)(x_a-x_c)}\\ y_0^{\′}=\frac{(y_d-y_a)((y_d-y_b)x_b-(x_d-x_b)y_b)-((y_c-y_a)x_a-(x_c-x_a)y_a)(y_d-y_b)}{(y_d-y_a)(x_b-x_d)-(y_d-y_b)(x_a-x_a)}\end{array}\right.$

Wherein:

$\left\{\begin{array}{c}{x}_a=x_1^{\′}+\frac{(x_0-x_1)(x_2^{\′}-x_1^{\′})}{x_2-x_1}\\ x_b=x_2^{\′}+\frac{(y_0-y_2)(x_3^{\′}-x_2^{\′})}{y_3-y_2}\\ x_c=x_4^{\′}+\frac{(x_0-x_4)(x_3^{\′}-x_4^{\′})}{x_3-x_4}\\ x_d=x_1^{\′}+\frac{(y_0-y_1)(x_4^{\′}-x_1^{\′})}{y_4-y_1}\\ y_a=y_1^{\′}+\frac{(x_0-x_1)(y_2^{\′}-y_1^{\′})}{x_2-x_1}\\ y_b=y_2^{\′}+\frac{(y_0-y_2)(y_3^{\′}-y_2^{\′})}{y_3-y_2}\\ y_c=y_4^{\′}+\frac{(x_0-x_4)(y_3^{\′}-y_4^{\′})}{x_3-x_4}\\ y_d=y_1^{\′}+\frac{(y_0-y_1)(y_4^{\′}-y_1^{\′})}{y_4-y_1}\end{array}\right.$

3. the application of distortion function

For the optical fingerprint collector of any type, no matter there is the distortion of any kind of, utilize the distortion that distortion function of the present invention can correcting image.After the process square net was once corrected, distortion function can calculate each picture element corresponding relationship in hinting obliquely at non-deformation pattern in the fault image.For fingerprint image, can directly accomplish the rectification overall process of distortion fingerprint image according to the correction model of optical finger print collection.

The present invention aligns fingerprint optically detecting device with square net in the horizontal and vertical directions when using, and collection distortion gray level image I (i, j).The gray level image (500dpi) that square net forms be J (i, j), its size is 320 * 320, (i j) has the direction of unanimity with image I.(i is j) with image J (i, central point j), two for registering images I

The coordinate at reference mark is respectively in the image:

Table 1. optical fingerprint collector image square net reference mark coordinate

Utilize the distortion function of the correcting algorithm of image distortion of optical fingerprint collector, behind the coordinate of input reference mark, can obtain the relation table of hinting obliquely at of entire image pixel:

Table 2. part optical fingerprint collector image is corrected the table of hinting obliquely at of back pixel

For the fingerprint image of input, can be through the method for tabling look-up, the fingerprint image after obtaining to proofread and correct, thus can realize the coupling of non-distortion fingerprint images such as distortion correction image and stamp fingerprint image.

Claims (2)

1. the correcting algorithm of an image distortion of optical fingerprint collector; It is characterized in that: the standard digital image of square net as reference; Image to square net forms on optical fingerprint collector is corrected, through each grid apex coordinate of record standard square net image and collector images acquired, the anisotropic of fault image distortion problem; Decomposition is in lattice; Handle according to linear distortion, and then the coordinate of arbitrfary point in fault image in the derivation square net, concrete steps are:

A) standard picture that forms of the square net of certain density δ be I (i, j), its size is m * n, writes down and takes pictures or the dpi of scanning collection, four of each grid summits are the reference mark of distortion correction in the square net;

B) align fingerprint optically detecting device in the horizontal and vertical directions with square net; Collection fault image J (i, j), its size is u * v; Picture format is converted into grayscale mode; (i j) has consistent direction and identical dpi, and the summit of each grid is the reference mark of distortion correction in the fault image with image I in maintenance;

C) registering images I (i, j) (i, central point j) write down the coordinate at each corresponding grid reference mark in two images respectively with image J;

D) (i, the lattice reference mark coordinate in j) is respectively P to establish image I ₁(x ₁, y ₁), P ₂(x ₂, y ₂), P ₃(x ₃, y ₃), P ₄(x ₄, y ₄), so, for image I (i, j) in any 1 P in the square lattice ₀(x ₀, y ₀), the ratio that projects on the line segment in level and vertical direction is respectively α and β, can obtain:

(i, j) the control corresponding point is respectively p ' to image J ₁(x ' ₁, y ' ₁), P ' ₂(x ' ₂, y ' ₂), P ' ₃(x ' ₃, y ' ₃), P ' ₄(x ' ₄, y ' ₄), some P ₀(x ₀, y ₀) at image J (i, the corresponding point P ' in j) ₀(x ' ₀, y ' ₀), have too:

E) P ₀(x ₀, y ₀) in the horizontal direction projection ratio α and in the projection ratio beta of vertical direction, with P ' ₀(x ' ₀, y ' ₀) in the horizontal direction the projection ratio and equate respectively in the projection ratio of vertical direction, can solve the distortion function equation and be:

wherein

。

2. the correcting algorithm of image distortion of optical fingerprint collector according to claim 1 is characterized in that: said e) distortion function derived of step can be through hinting obliquely at the rectification data of the fixing a kind of collector of table.

Images