CN1851751A - Digital waterprint method for iris image - Google Patents

Abstract

The present invention is digit watermark method used in irides picture, relating to biologic recognition technology and digit watermark field. Said method contains First determining irides effective area in irides picture, through analysing irides texture extracting texture feature information code, after encoding spread spectrum used as watermark information embedding to original picture transform domain having invariability to geometry distortion attack, to form watermark embedded irides picture. In watermark detection stage, said method is a blindness watermark detection algorithm without original picture. Said digit watermark method can be used in irides picture integrality and safeness detection.

Description

Be applied to the digital watermark method of iris image

Technical field

The present invention relates to biological identification technology and digital image processing field, particularly digital watermark technology.

Background technology

In the modern society of advanced IT application, the importance that identity is differentiated and authenticated is more and more outstanding.Based on biological characteristic such as fingerprint, iris, the identity authentication technique of sound etc. has unique, stable because of it, unchangeable, advantage such as carry easily, become the focus that identity is differentiated in recent years, wherein iris is used to identity more and more widely and differentiates because texture-rich is difficult to advantages such as forgery.

With the digital watermark technology security that strengthens biological attribute data that combines with biometric technology is an emerging research topic.Digital watermarking is that a kind of to embed certain information in protected multi-medium data be watermark, to reach identification, to protect a kind of technology of this multi-medium data owner's equity.Little, the concentration of energy of texture information region area of iris image, conventional digital watermark can make iris texture suffer by a relatively large margin change, influence the application of iris image.In addition, in the digital watermark of routine, watermark information normally characterizes a fixed sequence program or some pseudo-random sequences of user's copyright information, watermark is distorted insensitive for the part of texture, suffering malice to distort the back watermark can't detect or the testing result mistake, therefore can't guarantee the security of iris image, guarantee that the texture information zone is not distorted.Therefore, conventional digital Image Watermarking Technique and be not suitable for iris image.In general, effective iris image digital watermarking should comprise the feature of two aspects:

1) sentience not: refer to that for common digital watermarking conversion that embed watermark causes carrier data should be ND for observer's vision system.And the invisibility of iris watermark is not only with respect to people's sensorium, and requires the embedding of watermark not change the textural characteristics of iris image, is applied to different iris authentication systems as the iris image of embed watermark, and its feature does not change.

2) half frangibility: the watermark that is applied to iris image requires algorithm to distort for the malice of iris texture to have susceptibility, and iris suffers malice to distort the back watermark should to detect or detect to indicate and distort.The malice of iris is distorted the modification that is meant the iris image texture, causes the change of iris feature phasor.Watermark simultaneously also should be able to be born Flame Image Process and geometric distortion and be attacked, as compression of images, filtering, noise pollution, size change over etc.After through these operations, the watermarking algorithm of strong robustness should still can extract the watermark of embedding or the existence of proof watermark from watermark carrier.

Summary of the invention

What the present invention will solve is the defective that exists in the prior art; make the embedding of watermark not change the textural characteristics of iris image; make watermark stronger robustness can be arranged simultaneously to Flame Image Process and geometric distortion attack; thereby the characteristic information that makes hiding iris texture is not answered above-mentioned attack and is destroyed, and protects the integrality and the security of iris image effectively.

The invention discloses a kind of digital watermark method that is applied to iris image, this method belongs to half frangibility digital watermark technology: when watermark has robustness the part of iris texture distorted and have frangibility, promptly suffer malice to distort the back watermark and can't detect or the testing result mistake.In the conventional digital watermark technology, watermark information normally characterizes a fixed sequence program or some pseudo-random sequences of user's copyright information.In being applied to the digital watermark method of iris image, extract the ID Code of iris image, the transform domain that is embedded into original image behind the coding spread spectrum as watermark information so just can make iris feature information set up mapping relations between spatial domain and transform domain.The mapping relations of the characteristic information by analytic transformation territory and spatial domain, just can judge iris image spatial domain whether be subjected to distort.

The digital watermark method that is applied to iris image comprises following algorithm: the watermark information generating algorithm generates the watermark information sequence that comprises iris image textural characteristics information, in order to be embedded in host's iris image; Watermarking algorithm is fused to described watermark information in the iris image, guarantees described host image is only made the modification that is difficult for discovering.The embedding of watermark is chosen in the geometric attack field of invariants to be carried out; Watermark detection algorithms judges whether a sub-picture comprises described watermark information.

Description of drawings

The digital watermark method that Fig. 1 is applied to iris image generates block diagram;

The digital watermark method that Fig. 2 is applied to iris image detects block diagram;

Fig. 3 illustration the process flow diagram of the embedding algorithm of watermark among the present invention

Fig. 4 illustration the process flow diagram of watermark detection algorithms

Embodiment

1. be applied to the generation of the digital watermarking of iris image.Fig. 1 illustration be applied to the embedding block diagram of the digital watermark method of iris image.Method comprises following steps:

(1) from the iris picture, extracts iris texture characteristic information.

A. be that uncertainties such as illumination when avoiding owing to iris image acquiring, position are extracted the influence that brings to iris feature, need carry out pre-service iris image.Pre-service comprises location and two steps of normalization of iris.At first adopt Hough transfer pair iris to position, obtain the round parameter of iris boundary.Because people's pupil changes with the ambient lighting environmental change, at this moment also need by to iris dimensions normalization eliminate the influence of pupil convergent-divergent.Through forming the size unification after the normalized is the iris texture information band of 512*32.Iris texture information band with two dimension projects on the horizontal direction at last, forms the iris texture signal of one dimension.

B. the wavelet transformation zero passage detection is obtained iris texture characteristic.Dyadic wavelet transform can be carried out multi-level decomposition to signal, the energy of iris image signal focuses mostly in medium and low frequency, and the feature of selecting the result of wavelet transformation on medium level to extract iris can make iris feature that general attacked by noise is had resistivity preferably.Choose the result of the wavelet transformation on three intermediate frequency yardsticks of iris image, carry out zero passage detection, obtain the zero crossing positional information, and wavelet transformation result's average forms zero passage characteristic sequence (e between adjacent zero crossing _n, z _n), through behind the quantization encoding, obtained comprising the one-dimensional sequence { m of iris texture characteristic ₁, m ₂... m _k, m _i∈ 1,1}.It will be as iris feature watermark sequence to be embedded.

(2) embedding of watermark in field of invariants

In order to make iris watermark under geometric distortion attack condition, still have a robustness, watermark embed process should be chosen in the field of invariants of a geometric distortion and carry out, this transform domain should be that translation, shearing, rotation, convergent-divergent are had unchangeability, makes watermark attack the tool robustness to above-mentioned geometric distortion.(Fig. 3 illustration among the present invention the process flow diagram of the embedding algorithm of watermark)

The a.Radon conversion

Position with Hough transfer pair iris image, obtain circular iris image.As the center, iris region is carried out the Radon conversion with the center of circle of circular iris region.According to the character of Radon conversion as can be known, image has the convergent-divergent unchangeability in the Radon territory, and simultaneously, the rotation of image shows as the circulation translation of Radon transform domain.

B.2 tie up the DFT conversion

The Radon territory transformation results of iris image is carried out 2 dimension DFT conversion, know easily that according to the character of DFT conversion the circulation translation in Radon territory will only influence the phase field of DFT conversion, and the amplitude territory is unaffected.(R Θ) will be one the translation in the space will be sheared that rotate, geometric distortions such as convergent-divergent are attacked the transform domain that remains unchanged, and watermark information will embed in this transform domain through resulting DFT amplitude territory M after Radon conversion and the DFT conversion.

C. the embedding of watermark

Consider the energy distribution feature of iris image and the subjective feeling of human eye, we select the Mid Frequency embed watermark information in DFT amplitude territory.By the processing of the first step, we have obtained comprising the one dimension watermark information sequence { m of iris texture characteristic ₁, m ₂..., m _K, m _i∈ 1,1}.In order to strengthen the robustness of watermark, need carry out direct sequence spread spectrum to watermark information.Get the pseudo-random sequence P that length is L _nAs user code, the selection of sequence length n is relevant with the selection width of the length of watermark and DFT territory frequency band.W={w behind the spread spectrum ₁, w ₂..., w _K*L, w _i{ 1,1} is that an average is 0 pseudo-random sequence to ∈.(R Θ) satisfies in the territory to select M $f_1<\sqrt{{R_i}^2+{{\mathrm{\&Theta;}}_i}^2}{<f}_2$ Frequency band be that on the Mid Frequency K*L point embeds according to following formula:

M′(R _i，Θ _i)＝M(R _i，Θ _i)(1+αw _i)

Here α is the embedment strength of watermark, 1≤i≤K*L.Watermark need be embedded into M (R equally _i,-θ _i), this is to tie up the amplitude territory of DFT conversion about former point symmetry because of 2.Slipped into the iris image of watermark according to the flow process of Fig. 3.

2. blind watermark detection algorithms.The testing process of the watermark among the present invention does not need original image to participate in (as shown in Figure 2), is a kind of blind watermark detection algorithms.The testing process of watermark is divided two parts: the Detection of Existence of watermark and the security of iris detect.

(1) Detection of Existence of watermark.

At first locate the iris center, determine the iris circular domain that watermark embeds.Circular domain is done the Radon conversion, obtain the amplitude territory by Radon territory figure is done 2-D DFT conversion again.Select K*L coefficient M ' (R on the set frequency band in the amplitude territory _i, Θ _i).By setting up the spread-spectrum watermark w that obtains from the feature characteristic information coding of spatial domain extraction _iAnd M ' (R _i, Θ _i) the similarity evaluation function

$S=\frac{{\mathrm{\&Sigma;}}_{i=1}^N{w}_i{M}^{\&prime;}(R_i,{\mathrm{\&Theta;}}_i)}{\sqrt{{\mathrm{\&Sigma;}}_{i=1}^N{\left(M^{\&prime;}(R_i,{\mathrm{\&Theta;}}_i)\right)}^2}}$

$T=\frac{\mathrm{\&alpha;}{\mathrm{\&Sigma;}}_{i=1}^N{M}^{\&prime;}(R_i,{\mathrm{\&theta;}}_i)}{3\sqrt{{\mathrm{\&Sigma;}}_{i=1}^N{M}^{\&prime;}{\left((R_i,{\mathrm{\&theta;}}_i)\right)}^2}}$

Following formula be based on the threshold calculations function that detects of similarity evaluation function.By comparing the size of S and threshold value, determine whether valid watermark exists.

(2) the iris security detects.Under the prerequisite that valid watermark exists, whether the textural characteristics etc. that needs to determine iris image suffers that malice distorts, i.e. the legitimacy of detected image.

Select K*L coefficient M ' (R on the set frequency band in the amplitude territory _i, Θ _i), use user code P _nObtain watermark sequence M after the despreading _k(Fig. 4 illustration the process flow diagram of watermark detection algorithms) is because watermark information M _kBe the zero passage wavelet conversion coefficient of the iris image one dimension sample value behind the coding, therefore can obtain zero passage wavelet conversion coefficient (e after the decoding _n', z _n'), the iris one dimension sample value that ream weight is built is f _Ext(x), x ∈ 1,2 ... N}.Simultaneously embed watermark image is obtained one dimension sample point f ' (x) in the spatial domain projection, f ' (x) is rebuild on the low resolution in being used in after the wavelet decomposition and obtains f _Recon' (x).Set up following evaluation function

$D=\underset{m}{\min }\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}\frac{f_{{\mathrm{ext}}_n}\left(n\right)\&CenterDot;f_{{\mathrm{recon}}_j}^{\&prime;}(n+m)}{\left|\right|f_{{\mathrm{ext}}_n}\left(n\right)\left|\right|\&CenterDot;\left|\right|f_{{\mathrm{recon}}_j}^{\&prime;}(n+m)\left|\right|}$

Here m ∈ Z.M value difference can realize that still watermark embeds the optimum matching of feature and space and information when making the iris image rotation.Can estimate watermark according to evaluation function and rational thresholding and get legitimacy.

Claims (6)

1. digital watermark method that is applied to iris image.Feature is to comprise:

The watermark information generating algorithm generates the watermark information sequence that is embedded in the host image;

Watermarking algorithm is fused to described watermark information in the described host image, guarantees described host image only is difficult for the modification discovered;

Watermark detection algorithms judges whether a sub-picture comprises described watermark information.

2. digital watermark method described in the claim 1, wherein said host image is an iris image.

3. digital watermark method described in the claim 1, wherein said watermark information comprises the structural texture information of described host image itself.

4. digital watermark method described in the claim 1, wherein said watermark information generation method may further comprise the steps:

(1) iris image that obtains is carried out pre-service, comprise the location of iris, polar coordinates form normalized iris information band at last to the conversion of rectangular coordinate system;

(2) normalized iris information band is projected to the iris texture signal that forms one dimension on the horizontal direction.This one dimension iris texture is carried out multi-level wavelet transformation.The result who chooses wavelet transformation on several medium levels then carries out zero passage detection, and wavelet transformation result's mean value is as the characteristic sequence of this iris between the positional information of record zero crossing and adjacent zero crossing;

(3) use the characteristic sequence quantization encoding of user code, form the one-dimensional sequence that comprises iris texture characteristic, as the watermark information in original iris image to be embedded to obtaining.

5. digital watermark method described in the claim 1, wherein said watermarking algorithm may further comprise the steps:

(1) utilize the Hough conversion to locate the center and the cylindrical parameter of pupil in the iris picture.Select the circular domain embed watermark information in the iris cylindrical scope.

(2) the circular domain image is carried out the Radon conversion.

(3) carry out 2 dimension DFT conversion in the Radon territory.

(4) watermark information described in the claim 3 is carried out direct sequence spread spectrum, produce an average and be 0 pseudo-random sequence, be embedded into then in the amplitude territory of selected DFT.

(5), obtained embedding the iris picture of iris texture watermark with process DFT inverse transformation of the frequency domain data behind the embed watermark and Radon inverse transformation.

6. digital watermark method described in the claim 1, wherein said watermark detection algorithms may further comprise the steps:

(1) use user code to generate the watermark information of image to be detected according to the generation of watermark information described in the claim 4 method.

(1) existence of detection watermark.Determine the iris cylindrical scope that watermark embeds by Iris Location, this circular domain is done the Radon conversion, obtain the amplitude territory by Radon territory figure being done 2 dimension DFT conversion again.Select coefficient point on the set frequency band to calculate the similarity value of this image watermark information and described coefficient point in the amplitude territory, to determine whether valid watermark exists.

(2) the iris security detects, make the user code embed watermark sequence that decoding obtains from the iris image that embeds the iris feature watermark, carry out calculation of similarity degree with the iris feature sequence that directly from the iris image of embed watermark, adopts the zero passage wavelet transformation to extract, thereby determine the integrality of iris texture characteristic.

Images