WO2017113039A1

WO2017113039A1 - Iris region segmentation method and device based on active appearance model

Info

Publication number: WO2017113039A1
Application number: PCT/CN2015/000940
Authority: WO
Inventors: 王晓鹏
Original assignee: 王晓鹏
Priority date: 2015-12-30
Filing date: 2015-12-30
Publication date: 2017-07-06
Also published as: CN109074471A; CN109074471B

Abstract

An iris region segmentation method and device based on an active appearance model. The method comprises: creating an active appearance model consisting of a human eye shape model and a human eye texture model by using a plurality of pre-collected human eye sample images (S101); matching an inputted human eye image on which iris region segmentation will be performed and the pre-created active appearance model, to obtain a plurality of feature points that present a human eye contour in the inputted human eye image (S102); and selecting, from the multiple feature points, feature points used for fitting boundaries in the inputted human eye image to perform fitting, so as to obtain a segmented iris region (S103), wherein phase congruency information is used for both active appearance model creating and active appearance model matching.

Description

Iris region segmentation method and device based on active appearance model

Technical field

The present invention relates to the field of image processing, and in particular, to an iris region segmentation method and apparatus based on an Active Appearance Model (AAM).

Background technique

Today's society is a highly information-based society. On the one hand, people's demand for information is growing. On the other hand, the requirements for information security are getting higher and higher. Traditional authentication technologies include certificates, magnetic cards, passwords, etc., but these authentication techniques are not very secure. Therefore, biometric technology has emerged. Biometric identification refers to the use of certain unique features of the human body to identify these features using certain techniques to identify the identity of the person. Compared with traditional authentication technology, biometric technology has higher effectiveness, security and reliability.

Early biometrics mainly included face, fingerprint, signature, etc. These features are highly variable. Since the iris is inherently, not easy to lose, not easily damaged, and easy to identify, in recent years, iris verification has been highly recognized and paid attention by academics and industry.

The iris is an annular region between the black pupil and the white sclera in the human eye. It contains many intertwined spots, filaments, crowns, stripes, crypts and other details. When collecting iris images, due to physical limitations, pupils, eyelids, and eyelashes are usually taken together with the iris. Since iris recognition requires only an area between the pupil and the sclera that is not blocked by the eyelids and eyelashes, and other information, how to locate and segment the iris area becomes a hot spot and a difficult point in the iris recognition field.

The classical iris segmentation methods are: the integral/differential operator proposed by Danugman and the two localization algorithms proposed by Wildes combining edge detection and Hough transform.

Regarding the integral/differential operator proposed by Danugman, the advantage is that it can be calculated on the grayscale image without preprocessing the image, but it also has the following disadvantages, namely, The speed becomes very slow when the coarse positioning of the center and radius of the outer circle of the iris is not accurate. In addition, the spot formed when the iris image is acquired has an influence on the positioning accuracy, especially in uneven illumination. Positioning is easy to be mistaken when there are shadows, reflections, and occlusions. Moreover, since it is necessary to search in the three-dimensional parameter space and iterate to find the optimal solution, the calculation amount is large and the calculation speed is relatively slow.

On the other hand, the two-part positioning algorithm proposed by Wildes combining edge detection and Hough transform, because Hough transform is insensitive to noise in the image and robust, the advantage of this algorithm is that it is not sensitive to noise, but Because the computation of Hough transform is large and the extracted parameters are restricted by the quantization interval of the parameter space, the algorithm relies excessively on the accuracy of edge point detection. In addition, it needs to be in the three-dimensional parameter space when searching for the center and radius. Votes are made, so the amount of calculation and storage space is large.

It can be seen that the above two classical iris region segmentation methods are not ideal. How to quickly and accurately segment the effective iris area is still a technical issue to be solved.

Summary of the invention

The present invention has been made to solve the above technical problems, and an object thereof is to provide an accurate iris region segmentation method and apparatus for rapidly and robustly realizing segmentation of an iris region.

In order to solve the above problems, the present inventors have made intensive research to apply the active appearance model widely used in face modeling and face localization to the field of iris region segmentation, and fully consider the upper and lower eyelids to the iris. Based on the influence of occlusion, the active appearance model-based iris region segmentation method and apparatus of the present invention are proposed.

According to an aspect of the present invention, an iris region segmentation method is provided, wherein the iris region segmentation method is an iris region segmentation method based on an active appearance model, and the method includes: an active appearance model establishing step, using an advance Collecting a plurality of human eye sample images to establish an active appearance model composed of a human eye shape model and a human eye texture model; an active appearance model matching step, an input human eye image for performing iris region segmentation Matching the active appearance model established in the active appearance model establishing step to obtain a plurality of feature points presenting a contour of a human eye in the input human eye image; and a boundary fitting step from the Fitting a feature point for fitting each boundary in the input human eye image among a plurality of feature points obtained in the active appearance model matching step to obtain a segmented iris region, wherein The phase consistency information is utilized in both the active appearance model establishing step and the active appearance model matching step.

According to the iris region segmentation method of the present invention, by utilizing the phase consistency information of the human eye image in the active appearance model establishing step, the texture features of the pupil, the iris, and the upper and lower eyelids can be obtained more clearly than the texture features obtained by the existing method. The texture information can further establish a more accurate active appearance model than the prior art, and further, by utilizing the phase consistency information of the human eye image in the active appearance model matching step, the contour of the human eye can be presented very accurately, thereby enabling A more accurate iris region segmentation method is achieved than the prior art.

Preferably, in the iris region segmentation method, the active appearance model establishing step includes: a sample image phase consistency information calculating step, and calculating a phase consistent for each of the plurality of pre-acquired human eye sample images And the human eye texture model establishing step, using the calculated phase consistency information to establish a human eye texture model constituting the active appearance model.

According to the iris region segmentation method of the present invention, by using the calculated phase coincidence information to assist in marking the human eye image, it is possible to obtain a texture model that is more accurate than the texture model established by the existing method.

Preferably, in the iris region segmentation method, the human eye texture model establishing step includes: a shape dividing step of aligning a mean shape of the pre-acquired plurality of human eye sample images with the calculated phase The plurality of sample shapes obtained by marking the plurality of human eye sample images are respectively subjected to Delaunay triangulation; the texture normalization step is performed by the slice affine transformation to calculate the calculated phase consistency information Mapping to the mean shape to obtain normalized sample texture information; and principal component analysis processing steps, using the principal component analysis method to process the sample texture information to obtain a pattern Rational parameters, texture models.

Alternatively, in the iris region segmentation method, the human eye texture model establishing step includes: a texture normalization step of separately using the corresponding point-based image registration algorithm to calculate the calculated phase consistency information Mapping to the mean shape to obtain normalized sample texture information; and principal component analysis processing steps, using the principal component analysis method to process the sample texture information to obtain a texture parameter and a texture model.

Preferably, in the iris region segmentation method, the active appearance model matching step includes: an input image phase consistency information calculation step of calculating phase consistency information for the input human eye image to be subjected to iris region segmentation; An image texture calculation step of calculating a texture of the input human eye image using the calculated phase consistency information of the input human eye image; and an appearance texture matching step of calculating the texture of the input human eye image The textures of the active appearance models established in the active appearance model establishing step are matched to obtain a plurality of feature points that present a contour of a human eye in the input human eye image.

Preferably, in the iris region segmentation method, the active appearance model establishing step further includes: a sample image collecting step of pre-collecting the plurality of human eye sample images of the left and right eyes of different people; and the feature point calibration step Separating the feature points manually on the plurality of human eye sample images; the feature point alignment step aligning the corresponding feature points in the plurality of human eye sample images; and the human eye shape model establishing step, utilizing the features a feature point in the aligned plurality of human eye sample images in the point alignment step to establish a human eye shape model constituting the active appearance model; and a synthesizing step of the established human eye shape model and the The human eye texture model is combined to obtain the active appearance model.

Preferably, in the iris region segmentation method, in the feature point alignment step, Platts analysis is used to obtain an alignment image in which translation, scale, and rotation are removed.

Preferably, in the iris region segmentation method, in the human eye shape model establishing step and the human eye texture model establishing step, the human eye shape model and the human eye texture are obtained by principal component analysis. model.

According to the iris region segmentation method of the present invention, in order to obtain a shape model and a texture model, Principal component analysis is used to process the data, thereby reducing the amount of data that needs to be processed and saving computation time.

Preferably, in the iris region segmentation method, the strip boundaries include an iris boundary, a pupil boundary, and an upper and lower eyelid boundary.

Preferably, in the iris region segmentation method, when the iris boundary is fitted in the boundary fitting step, a plurality of feature points obtained in the active appearance model matching step are selected to be located on the left side of the iris At least a portion of the feature points on the boundary and the right edge of the iris are fitted.

Preferably, in the iris region segmentation method, when the pupil boundary is fitted in the boundary fitting step, selecting a plurality of feature points obtained in the active appearance model matching step is located on a pupil boundary At least a portion of the feature points are fitted to fit.

Preferably, in the iris region segmentation method, when the upper and lower eyelid boundaries are fitted in the boundary fitting step, the upper and lower eyelids are selected from a plurality of feature points obtained in the active appearance model matching step. At least a portion of the feature points at the intermediate portion of the boundary that are spaced apart from the corner of the eye are fitted.

According to another aspect of the present invention, an iris region segmentation device is provided, wherein the iris region segmentation device is an iris region segmentation device based on an active appearance model, the device comprising: an active appearance model establishing device, An active appearance model composed of a human eye shape model and a human eye texture model is configured to utilize a plurality of pre-acquired human eye sample images; an active appearance model matching device configured to input a person to be subjected to iris region segmentation The eye image is matched with the active appearance model established in the active appearance model building device to obtain a plurality of feature points that present a contour of a human eye in the input human eye image; and a boundary fitting device that is Configuring to select a feature point for fitting each boundary in the input human eye image from a plurality of feature points obtained in the active appearance model matching device to perform fitting to obtain the segmented iris region Wherein the phase 1 is utilized in both the active appearance model establishing device and the active appearance model matching device Information.

An iris region segmentation device according to the present invention, by establishing an apparatus in an active appearance model The phase consistency information of the human eye image is utilized, so that the texture features of the pupil, the iris, and the upper and lower eyelids can be obtained more clearly than the texture features obtained by the existing method, thereby being able to establish a more accurate active appearance than the prior art. In addition, by utilizing the phase consistency information of the human eye image in the active appearance model matching device, the human eye contour can be presented very accurately, and an iris region dividing device more accurate than the prior art can be realized.

Preferably, in the iris region segmentation device, the active appearance model establishing device includes: a sample image phase consistency information calculation portion configured to each of the pre-acquired plurality of human eye sample images The amplitude calculation phase consistency information; and a human eye texture model establishing section configured to use the calculated phase consistency information to establish a human eye texture model constituting the active appearance model.

According to the iris region dividing device of the present invention, by using the calculated phase coincidence information to assist in marking the human eye image, it is possible to obtain a texture model which is more accurate than the texture model established by the existing method.

Preferably, in the iris region segmentation device, the human eye texture model establishing portion includes: a shape dividing unit configured to calculate an average shape of the pre-acquired plurality of human eye sample images and utilize the calculation The plurality of sample shapes obtained by marking the plurality of human eye sample images are respectively subjected to Delaunay triangulation; the texture normalization unit is configured to pass the slice affine transformation The calculated phase consistency information is respectively mapped to the mean shape to obtain normalized sample texture information; and a principal component analysis processing unit configured to process the sample texture information by principal component analysis, Get texture parameters, texture models.

Alternatively, in the iris region segmentation device, the human eye texture model establishing portion includes: a texture normalization unit configured to utilize the corresponding point-based image registration algorithm to calculate the calculated phase Consistency information is respectively mapped to the mean shape to obtain normalized sample texture information; and a principal component analysis processing unit configured to process the sample texture information by using principal component analysis to obtain texture parameters, Texture model.

Preferably, in the iris region segmentation device, the active appearance model matching device includes: an input image phase consistency information calculation portion configured to calculate a phase coincidence for the input human eye image to be subjected to iris region segmentation Information information; an input image texture calculation section configured to calculate a texture of the input human eye image using the calculated phase consistency information of the input human eye image; and an appearance texture matching section configured to calculate Matching the texture of the input human eye image with the texture of the active appearance model established in the active appearance model building device to obtain a plurality of feature points presenting a contour of a human eye in the input human eye image .

Preferably, in the iris region segmentation device, the active appearance model establishing device further includes: a sample image collecting portion configured to pre-collect the plurality of human eye sample images of the left and right eyes of different people; a feature point calibration portion configured to manually calibrate feature points on the plurality of human eye sample images; a feature point alignment portion configured to align corresponding feature points in the plurality of human eye sample images; a human eye shape model establishing portion configured to establish a human eye shape model constituting the active appearance model by using feature points in the plurality of human eye sample images aligned in the feature point alignment portion; And a synthesizing portion configured to combine the established human eye shape model and the human eye texture model to obtain the active appearance model.

Preferably, in the iris area segmentation device, the feature point alignment portion is configured to use a Platts analysis to obtain an aligned image that removes translation, scale, and rotation.

Preferably, in the iris region segmentation device, the human eye shape model establishing portion and the human eye texture model establishing portion are configured to obtain the human eye shape model and the human eye texture by principal component analysis. model.

According to the iris region dividing device of the present invention, in order to obtain the shape model and the texture model, the data is processed using principal component analysis, whereby the amount of data to be processed can be reduced, and the calculation time can be saved.

Preferably, in the iris region segmentation device, the strip boundaries include an iris boundary, a pupil boundary, and an upper and lower eyelid boundary.

Preferably, in the iris region segmentation device, in the boundary fitting device When the iris boundary is combined, at least a part of feature points located on the left side boundary of the iris and the right side boundary of the iris are selected from a plurality of feature points obtained by the active appearance model matching device to perform fitting.

Preferably, in the iris region segmentation device, when the pupil boundary is fitted in the boundary fitting device, a plurality of feature points obtained by the active appearance model matching device are selected from a boundary of the pupil At least a portion of the feature points are fitted.

Preferably, in the iris region dividing device, when the upper and lower eyelid boundaries are fitted in the boundary fitting device, the upper and lower eyelid boundaries are selected from a plurality of feature points obtained by the active appearance model matching device At least a portion of the feature points at the intermediate portion spaced from the corner of the eye are fitted to fit.

DRAWINGS

Other features, objects, and advantages of the present invention will become more apparent from the Detailed Description of Description

1 is a flow chart of an iris region segmentation method 100 in accordance with one embodiment of the present invention;

2 is a diagram showing an implementation flow of a step S101 of establishing a human eye active appearance model according to an embodiment of the present invention;

Figure 3 is an example acquisition image of the human eye;

Figure 4 is a schematic view of selected human eye feature points;

FIG. 5 is a diagram showing phase consistency information calculated for an example captured image; FIG.

6 is a diagram showing an implementation flow of a step S1015 of establishing a human eye texture model according to an embodiment of the present invention;

Figure 7 is a schematic diagram of Delaunay triangulation of a point set;

Figure 8 is a schematic diagram of a linear affine of a slice;

Figure 9 is a diagram showing an implementation flow of a step S1015' of establishing a human eye texture model according to an alternative embodiment of the present invention;

Figure 10 is an active appearance model and a new human eye image in accordance with one embodiment of the present invention. a diagram matching the implementation flow of step S102;

11 is a view showing a series of feature points reflecting a contour of a human eye obtained by matching a new human eye image to an active appearance model to be subjected to iris region segmentation;

Figure 12 is a diagram showing the result of fitting;

Figure 13 is a block diagram of an iris region segmentation device 1300, in accordance with one embodiment of the present invention;

14 is a block diagram of an iris region segmentation device 1400 in accordance with an alternate embodiment of the present invention.

The same or similar reference numerals in the drawings denote the same or similar elements.

detailed description

Embodiments of the present invention will be described more fully hereinafter with reference to the accompanying drawings in which FIG. However, the present invention may be embodied in many different forms and the invention should not be construed as being limited to the embodiments set forth herein.

The invention is further described in detail below with reference to the accompanying drawings.

1 is a flow chart of an iris region segmentation method 100 in accordance with one embodiment of the present invention.

As shown in FIG. 1, first, in step S101, an active appearance model of the human eye is established. The active appearance model is used for boundary verification and image segmentation, which is formed by using shape information and texture information of an image to create a shape model and a texture model, and then combining the two. The purpose is to obtain the shape of the target area, the affine transformation coefficient, and the like from a pre-trained model.

The following is an example to illustrate how to build an active appearance model of the human eye.

2 is a diagram showing an implementation flow of a step S101 of establishing a human eye active appearance model according to an embodiment of the present invention.

First, the human eye sample image is acquired and the feature points are calibrated (step S1011). Specifically, a clear image I of the left and right eyes of different people is collected. The image shown in Figure 3 is a sharp image I that was acquired. After collecting N such clear images I, n feature points {(x _i , y _i ), i = 1, ..., n} are manually calibrated on each of the sharp images I. When n feature points are calibrated, feature points whose texture features change significantly (for example, upper and lower eyelid boundaries, iris boundaries, pupil boundaries, etc.) are selected. Among them, it should be noted that due to the occlusion of the eyelid, the upper and lower boundaries of the iris may not exist. Therefore, when selecting the feature points of the iris boundary, instead of selecting all the feature points on the circular boundary of the iris, only the left and right sides of the iris are selected. The feature points of the part that is blocked by the eyelids.

Figure 4 is a schematic illustration of selected human eye feature points. Due to the limitation of the physical structure, points at the pupil, eyelashes, and the like may be collected when the iris image is acquired. Therefore, in order to avoid the influence of the pupil, the eyelash, and the like, a total of 68 feature points are selected in the present embodiment. The position of the selected 68 feature points is shown in FIG. 4, wherein feature points 19 to 36 are selected at the upper eyelid boundary, a total of 18 points, and feature points 1 to 18 are selected at the lower eyelid boundary. At 18 points, feature points 57 to 68 were selected at the pupil boundary, a total of 12 points, and 10 feature points were selected at the left boundary of the iris not blocked by the eyelid and the right boundary of the iris, ie, the left side Feature points 52 to 56, 37 to 41 and feature points 42 to 51 on the right side.

After the feature points are calibrated, the phase coincidence information of the human eye image is calculated for each of the N clear images I (step S1012). Since the human eye image is mainly understood based on low-level features such as step edges and zero-crossing edges in the image, unlike the prior art, the active appearance model of the present invention is used in the process of establishing the active appearance model. Phase consistency information that contributes to the spatial resolution of edge detection. This is a method of edge detection and texture analysis using frequency domain space. Phase consistency refers to a measure of the phase similarity of each frequency component at each position of the image. It is a dimensionless quantity whose value decreases from 1 to 0, indicating a decrease from a salient feature to a no feature. Using the phase consistency information to detect the image can extract the texture features of the image, not just the edge portion. In addition, since the phase consistency information is not sensitive to the brightness and contrast of the image, the information can also be used to overcome the light and darkness. The impact of the texture structure. The phase consistency information of the human eye clear image I at the feature point x can be calculated by the following formula (1):

Where ε is a small positive constant, for example, can be set to 0.01, θ _j = jπ / J, j = {0, ..., J-1} is the direction angle of the filter, J is the direction number, n Is the number of feature points manually calibrated on each image,

with

The local amplitude and local energy along the θ _j direction are respectively calculated by the following equations (2) and (3):

among them

with

The response in the θ _j direction at each feature point x after convolution of the human eye clear image I and the two-dimensional log-Gabor filter, respectively. The transfer function of the two-dimensional log-Gabor filter in the frequency domain is defined as follows:

Where ω ₀ is the center frequency of the filter, σ _r is the bandwidth of the filter, and σ _θ is the angular bandwidth of the filter.

The phase coincidence information is calculated by the above formula for the image shown in Fig. 3 after the calibration feature point, and the result is shown in Fig. 5. As can be seen from Fig. 5, by using the phase coincidence information, it is possible to obtain an eye contour image in which the texture features of the upper and lower eyelids, the pupil, and the iris are very clear.

After the phase coincidence information of the human eye image is calculated for the above-described N clear images I, the corresponding feature points in the above-described N sharp images I are aligned (step S1013). Specifically, performing a Platts analysis on the N clear images I after the calibration feature points, wherein the N is calculated separately The center of gravity of the shape of the sharp image I is the center of gravity of the shape, the center of gravity of the N shapes is moved to the same position, and then the shape of the N clear images I is expanded or reduced to the same size, and finally passed through two clear images I The position of the corresponding point of the shape is used to calculate the difference in the rotation angle, and then the object is rotated so that the angle of the shape of the clear image I is uniform. As such, the corresponding feature points of the different images are aligned to obtain a human eye image that removes the alignment of the translation, scale, and rotation.

After the feature points in the acquired N sharp images I are aligned by the above-described step S1013, a shape model constituting the active appearance model is established (step S1014).

Specifically, first, the n feature points after each image have been aligned are connected to form a shape vector s _i , and the N clear images I are spliced into an N×2n human eye shape matrix s by using the following formula (5):

Where s _i = (x _1i , ..., x _ni , y _1i , ..., y _ni ) ^T . Next, the spliced human eye shape matrix s is averaged by using the following formula (6) to obtain an eye average matrix.

After that, the spliced human eye shape matrix s and the human eye average matrix

Performing a subtraction operation to obtain a difference matrix D={d _ij , i=1, . . . , 2n, j=1, . . . , N}, where

i=1,...,2n,j=1,...,N, then, the covariance matrix U of the difference matrix D is calculated using the following equation (7):

U=DD ^T (7)

After obtaining the covariance matrix U, the eigenvalues and eigenvectors of the covariance matrix U are calculated, and then the eigenvalues are sorted in descending order, and the eigenvectors corresponding to the first k largest eigenvalues are taken. The energy of the first k eigenvalues is made more than 95% of the total energy. These corresponding feature vectors are composed into a principal component analysis (PCA: Principal Component Analysis) projection matrix Φ _s to obtain a shape model given by the following equation (8):

among them,

It is a mean shape, Φ _s is a transformation matrix formed by principal component eigenvectors obtained by principal component analysis, and b _s is a statistical shape parameter that controls shape change. In the shape model shown in the above formula (8), in the mean shape

Based on this, a new shape model can be obtained by adjusting the statistical shape parameter b _s .

After the shape model of the human eye is established, the calculated phase consistency information is used to establish a texture model constituting the active appearance model (step S1015).

Specifically, an implementation flow of the establishing step S1015 of the human eye texture model according to an embodiment of the present invention is shown in FIG.

First, in step S1015a, the above mean shape

The N sample shapes each characterized by a series of feature points obtained by marking the N clear images I using the calculated phase coincidence information of the N clear images I are respectively subjected to Delaunay triangulation. The so-called Delaunay triangulation is a technique of joining spatial points into triangles to maximize the minimum angle of all triangles. The point of the Delaunay triangulation is that the circumscribed circle of any triangle does not include any other vertices. Figure 7 is a schematic diagram of Delaunay triangulation for a set of points. The process of a method of Delaunay triangulation shown in Figure 7 is as follows:

1) Select any point in the set of points, then select another point that is closest to the point, and then connect the two points as a directional baseline;

2) Apply the Delaunay criterion to search for the third point to the right of the directional baseline;

3) Create a Delaunay triangle, and then specify the direction in the generated triangle as the new baseline from the starting point of the baseline to the third point and the two points from the third point to the end point of the baseline;

4) Repeat steps 2) and 3) above until all baselines have been used.

Using the above-described Delaunay triangulation process to apply the above mean shape

Delaunay triangulation is performed separately from the above N sample shapes, thereby obtaining the above mean shape

And the above N sample shapes are divided into a series of triangles.

Next, in step S1015b, phase coincidence information of the acquired N sharp images I is mapped to the mean shape by slice affine transformation

Achieve normalization of textures. Due to the above mean shape

The triangles obtained by dividing the above N sample shapes by the Delaunay triangle are mutually corresponding, so that the corresponding mean shape can be calculated by the piecewise linear affine projection according to the position of each point in the triangle of the sample shape.

Position in the triangle, then map the value of the phase consistency of the point to the mean shape

The position of the corresponding point in .

Figure 8 is a schematic illustration of the linear affine of the slice. As shown in FIG. 8, the triangle on the left side and the triangle on the right side respectively represent the triangle shape obtained by dividing the sample shape and the above-described mean shape by the Delaunay triangle. The positions and correspondences of the vertices v ₁ , v ₂ , v ₃ and v' ₁ , v' ₂ , v' ₃ of the two triangles are known. For a point p in the triangle of the sample shape (the p-point coordinates are known), the linear affine transformation based on the barycentric coordinates can be used to obtain the position of the corresponding point p' in the triangle of the mean shape, and the corresponding point is completed. Mapping of phase consistency information (ie, texture information).

The phase consistency information (ie, texture information) of each of the above N sharp images I can be mapped to the above average shape by the above method.

Upper, thereby, normalizing the texture, that is, mapping the phase consistency information (ie, texture information) of the N sample shapes to the uniform reference system by the slice linear affine transformation , for the creation of the texture model for the next step.

Next, in step S1015c, all normalized sample texture information is processed by principal component analysis to obtain texture parameters, thereby obtaining a texture model. Specifically, first, averaging all normalized sample texture information to obtain an average texture

Next, principal component analysis is performed by a method similar to the above-described step S1014, and feature vectors corresponding to the first m feature values sorted by the size of the feature values are obtained. Then, these corresponding feature vectors are composed into a principal component analysis projection matrix Φ _g to obtain a texture model given by the following formula (9):

among them,

It is the average texture, Φ _g is the transformation matrix formed by the texture principal component eigenvector obtained by principal component analysis, and b _g is the statistical texture parameter that controls the texture change. In the texture model shown in the above formula (9), the average texture

Based on this, a new texture model can be obtained by adjusting the statistical texture parameter b _g .

It should be noted that the implementation flow of the step S1015 of the human eye texture model shown in FIG. 6 is merely an example, and various modifications can be made to achieve the same effect. For example, instead of performing the Delaunay triangulation and the slice affine transformation in FIG. 6, the image registration algorithm based on the corresponding points may be used to map the phase consistency information of the acquired N sharp images I to the above average shape.

Achieve normalization of textures. The implementation flow of this alternative is shown in FIG. Specifically, first, phase matching information of the acquired N sharp images I is mapped to the above average shape using a corresponding point-based image registration algorithm such as an image registration algorithm based on a thin plate spline function.

Realizing normalization of textures, wherein the basic idea of the image registration algorithm based on corresponding points is to calculate two or more images acquired under different conditions or acquired by different imaging devices by calculating an optimal spatial transformation The positions of the corresponding feature points in the image are in one-to-one correspondence (step S1015'a), and then, similarly to FIG. 6, all the normalized sample texture information is processed by principal component analysis to obtain texture parameters, thereby obtaining a texture model. (Step S1015c).

Regarding the steps described above, it should be noted that although the phase consistency information of the human eye image is calculated for the N sharp images I (step S1012), it is shown in FIG. 2 for the above-described N sharp images I. The feature points are aligned (step S1013) and the shape models constituting the active appearance model are established (step S1014), but the order of the steps between them is not limited thereto. As long as the above-mentioned step S1012 and the above-described step S1014 are performed before the texture model is established (step S1015/step S1015') and the above-described step S1013 is performed before the above-mentioned step S1014, the order of the steps can be arbitrarily changed. Or can be executed at the same time. For example, the above step S1012 and the above step S1013 may be simultaneously performed, and then the above step S1014 and the above step S1015/step S1015' may be sequentially performed; or, the above step S1012 may be performed after the above step S1013 is performed, and then the above step S1014 and the above are sequentially performed. Step S1015 / Step S1015'; Alternatively, after the above-described step S1013 and the above-described step S1014 are sequentially performed, the above-described step S1012 may be performed, and then the above-described step S1015 / step S1015' and the like are performed.

Finally, after the shape model and the texture model are created, the two models are combined into an active appearance model (step S1016). Specifically, first, b _s and b _{g are} connected according to the following formula (10) to obtain an appearance feature vector b:

Wherein, w _s is used to adjust the difference between the dimension b _s and b _g diagonal matrix. Next, principal component analysis is performed on the obtained appearance feature vector b to further eliminate the correlation between the shape and the texture, thereby obtaining an active appearance model given by the following formula (11):

among them,

Is the average appearance vector, Q is the transformation matrix formed by the principal component eigenvectors obtained by principal component analysis, and c is the appearance model parameter that controls the appearance change. As such, in the case of a given appearance model parameter c and a corresponding similar transformation matrix (such as a scaling matrix, a rotation matrix, etc.), a human eye image can be synthesized.

After completing the process of establishing the active appearance model of the human eye, returning to FIG. 1, the process proceeds to step S102 shown in FIG. In this step S102, the active appearance model obtained in step S101 is used to match a new human eye image to be subjected to iris region segmentation different from the above-described N sharp images I to obtain an accurate representation of the new human eye. A series of feature points of the human eye contour in the image.

The following is an example to illustrate how to match the active appearance model to a new human eye image.

FIG. 10 is a diagram showing an implementation flow of a matching step S102 of an active appearance model and a new human eye image according to an embodiment of the present invention.

First, phase coincidence information is calculated for a new human eye image I _n to be subjected to iris region segmentation (step S1021). The calculation method can be the same as that employed in step S1012 in FIG.

Then, using the phase coincidence information calculated in the above step S1021, the human eye image I _n is calculated to be deformed to the mean shape according to the current shape s.

The resulting texture g _s (step S1022).

Then, the appearance model parameter c in the active appearance model obtained in step S101 is continuously changed to optimize the objective function given by the following formula (12) until the appearance texture of the active appearance model and the human eye image I _n The appearance texture is consistent (step S1023):

Δ=||δ _g || ² =||g _s -g _m || ² (12)

Where g _s is the texture of the new human eye image I _n to be subjected to the iris region segmentation, and g _m is the texture of the active appearance model obtained in step S101, and

The optimization process of the objective function given by the above formula (12) is as follows:

I. Initialize the iteration number t and the appearance model parameter c, that is, set t=0 and c=0;

II. Calculating the difference between the texture of the human eye image I _{n and} the texture of the active appearance model obtained in step S101: δ _g = g _{s -} g _m ;

III. Update the appearance model parameters according to c'=c-kδ _c (where k is the adjustment coefficient, k = 1 at this time, δ _c is the variation of the appearance model parameters), and, in the new appearance model parameter c' In the case, the difference δ' _g between the texture of the human eye image I _{n and} the texture of the active appearance model is calculated;

IV. Compare δ _g and δ' _g . If δ' _g < δ _g , the current appearance model parameter value is assigned to c, that is, c = c ', and enters V; otherwise, returns to III, by sequentially changing the adjustment coefficient k (for example, let k = 1.5, 0.5 , 0.25) to continue to adjust the active appearance model;

V. Update the number of iterations t=t+1, and determine whether the difference δ′ _g between the texture of the human eye image I _{n and} the texture of the active appearance model is less than a threshold ξ, if it is less than, then exit; otherwise, return to III. If the number of iterations exceeds a predetermined number of times, the human eye is considered not to be included in the image.

A human eye image for which iris region segmentation is to be performed is shown in FIG. The texture of the human eye image shown in FIG. 11 is matched with the texture of the previously established active appearance model by the matching step S102 as described above, and a series of feature points are obtained on the human eye image. As can be seen from Fig. 11, these feature points coincide very precisely with the iris boundary, the pupil boundary, and the upper and lower eyelid boundaries in the human eye image, and these feature points very accurately represent the contour of the human eye image.

Therefore, for any new human eye image I _n to be subjected to iris region segmentation, as long as the objective function given by the above formula (12) is continuously optimized in the matching step S102 as described above until the iris region segmentation is to be performed The difference between the texture of the human eye image I _{n and} the texture of the previously established active appearance model is less than a predetermined threshold, so that a plurality of feature points that almost completely coincide with the boundaries of the human eye image I _n can be obtained, not only It can ensure the overall matching accuracy, and can also ensure the matching precision at each feature point. Furthermore, the contour of the human eye can be presented more accurately, providing accurate information for subsequent iris region segmentation.

After matching the previously established active appearance model to the new human eye image to be subjected to the iris region segmentation to obtain a series of feature points that accurately represent the human eye contour in the new human eye image, return to FIG. 1 and enter Step S103 shown in FIG. In this step S103, a plurality of feature points are selected from the series of feature points obtained by the above-described step S102 to fit the iris boundary, the pupil boundary, and the boundary of the upper and lower eyelids, respectively, using the least squares method. The result of the fitting is shown in FIG. The selection of feature points for fitting the boundaries of the iris, the pupil boundary, and the upper and lower eyelids will be described below.

●Iris boundary fitting

Since the upper and lower eyelids block the upper and lower boundaries of the iris, when the least squares method is used to fit the iris boundary, the feature points on the left side boundary of the iris and the right side boundary of the iris which are not blocked by the upper and lower eyelids are selected. For example, the 20 feature points 37 to 56 shown in FIG. 4 can be selected to fit the iris boundary. The iris boundary obtained by fitting using the 20 feature points of 37 to 56 is as shown in FIG. Of course, the selection method of the feature points for fitting the iris boundary is not limited thereto, and only a part of the feature points may be used to fit the iris boundary, for example, using the three 38, 48, 56 shown in FIG. A combination of feature points or other number of feature points to fit the iris boundary.

●Fitting the pupil boundary

Since the pupil boundaries are generally not affected by the upper and lower eyelids, all of the feature points 57 to 68 on the pupil boundary can be used for fitting. The pupil boundary obtained by fitting using the 12 feature points of 57 to 68 is as shown in FIG. Of course, the selection method of the feature points for fitting the pupil boundary is not limited thereto, and only the partial feature points may be used to fit the pupil boundary, for example, using the three numbers of 58, 58, and 66 shown in FIG. A combination of feature points or other number of feature points to fit the pupil boundary.

●Fitting of the upper eyelid boundary

As can be seen from Fig. 11, among the feature points of the upper eyelid, the features near the left and right eye corners The point is not suitable for fitting the parabola, so only the 10 feature points of 23 to 32 are used to fit the upper eyelid. The fitted upper eyelid boundary is shown in Figure 12. Of course, the selection method of the feature points for fitting the upper eyelid boundary is not limited thereto, and a smaller number of feature points than the above 10 feature points may be used to fit the upper eyelid boundary, for example, using the method shown in FIG. A combination of three

feature points

25, 28, 30 or other number of feature points to fit the upper eyelid boundary.

●Fitting of the lower eyelid boundary

As can be seen from Fig. 11, in the feature points of the lower eyelid, the feature points near the left and right eye corners are not suitable for fitting the parabola, and therefore, only the 10 feature points of 5 to 14 are used to fit the lower eyelid. The fitted lower eyelid boundary is shown in Figure 12. Of course, the selection method of the feature points for fitting the lower eyelid boundary is not limited thereto, and a lower number of feature points than the above 10 feature points may be used to fit the lower eyelid boundary, for example, as shown in FIG. The combination of the three feature points 7, 10, 12 or other number of feature points to fit the lower eyelid boundary.

After fitting the iris boundary, the pupil boundary, and the boundary of the upper and lower eyelids as described above, a common area located below the upper eyelid, above the lower eyelid, outside the pupil boundary, and within the iris boundary can be obtained. This public area is effective. The iris area (as shown in Figure 12), thereby completing the segmentation of the iris area.

According to the method of the present invention, by using the phase consistency information of the human eye image for the establishment of the texture model in the active appearance model, it is possible to obtain texture information with clearer texture features of the pupil, the iris, and the upper and lower eyelids, thereby establishing a ratio. The more accurate active appearance model of the prior art, in addition, by using the phase consistency information of the human eye image for the matching of the human eye image and the active appearance model, the contour of the human eye can be presented very accurately, and thus can be realized more than the existing one. More accurate iris segmentation method.

Hereinafter, an apparatus for realizing the iris region dividing method of the present invention will be described. Figure 13 is a block diagram of an iris region segmentation device 1300, in accordance with one embodiment of the present invention.

As shown in FIG. 13, the iris region dividing device 1300 includes an active appearance model establishing device 1301, an active appearance model matching device 1302, and a boundary fitting device 1303.

The active appearance model establishing device 1301 is a device for establishing an active appearance model of a human eye image, and includes: a clear image for collecting left and right eyes of different people. a sample image acquisition unit 1301a; a feature point calibration unit 1301 for manually calibrating feature points on each of the acquired sharp images; and a sample image phase for calculating phase consistency information of the human eye image for each clear image a sex information calculation unit 1301c; a feature point alignment unit 1301d for aligning corresponding feature points in all the acquired clear images; for constructing a shape model constituting the active appearance model by using feature points in all the clear images after alignment Human eye shape model establishing section 1301e; a human eye texture model establishing section 1301f for constructing a texture model constituting an active appearance model using the calculated phase consistency information; and a human eye shape model and a person for establishing the human eye shape The eye texture model is combined to obtain a synthesis portion 1301g of the human eye active appearance model.

Further, the human eye texture model establishing section 1301f includes: a Delaunay triangulation for all the sample shapes obtained by marking the mean shape and the phase consistency information using all the calculated clear images to mark all the clear images. a shape division unit (not shown); a texture normalization unit for mapping the phase consistency information of all the acquired clear images to the mean shape of the clear images by the slice affine transformation (in the figure) Not shown); and a principal component analysis processing unit (not shown) for processing all normalized sample texture information by principal component analysis to obtain texture parameters and texture models.

According to FIG. 13, the active appearance model establishing means 1301 inputs the human eye sample images of the left and right eyes of different people, and generates the active appearance model of the human eye through the processing of each of the parts 1301a to 1301g, and the active appearance is obtained. The model matching device 1302 outputs.

The active appearance model matching device 1302 is a device for matching a new human eye image to be subjected to iris region segmentation with an active appearance model of the human eye outputted from the active appearance model matching device 1302, and includes: An input image phase coincidence information calculation unit 1302a that calculates phase consistency information for a new human eye image to perform iris region segmentation; for calculating a new human eye to be subjected to iris region segmentation using the calculated phase coincidence information An input image texture calculation unit 1302b of the texture of the image; and a texture for the new human eye image to be subjected to the division of the iris region to be calculated The texture of the active appearance model output by the dynamic appearance model matching means 1302 matches the appearance texture matching section 1302c.

According to FIG. 13, the active appearance model matching device 1302 is configured to input an input image and an active appearance model of the iris region to be segmented, and obtain a series of feature points for presenting the contour of the input image by processing the portions 1302a to 1302c therein. The output is performed to the above-described boundary fitting device 1303.

The boundary fitting device 1303 is configured to select a plurality of suitable feature points from the series of feature points to respectively fit the iris boundary, the pupil boundary, and the boundary of the upper and lower eyelids by using a least square method, including: An iris boundary fitting portion 1303a for fitting an iris boundary of the input image; a pupil boundary fitting portion 1303b for fitting a pupil boundary of the input image; an upper eyelid boundary for fitting an upper eyelid boundary of the input image A fitting unit 1303c; and a lower eyelid boundary fitting unit 1303d for fitting a lower eyelid boundary of the input image.

As can be seen from Fig. 13, the boundary fitting device 1303 receives a series of feature points obtained by matching, and through the processing of the respective portions 1303a to 1303d, the division of the iris region of the input image is completed, and an effective iris region is obtained.

It is to be noted that the iris area dividing device 1300 shown in FIG. 13 is merely an example, and various modifications can be made thereto to achieve the same effect. As a variant, a block diagram of an iris region segmentation device 1400 in accordance with an alternate embodiment of the present invention is shown in FIG.

The iris area dividing device 1400 is different from the iris area dividing device 1300 shown in FIG. 13 only in that an active appearance model establishing means 1401 is used instead of the active appearance model establishing means 1301 shown in FIG. 13, further In place of the human eye texture model establishing unit 1301f shown in FIG. 13, the human eye texture model establishing unit 1401f is included in the active appearance model establishing means 1401.

Specifically, the human eye texture model establishing unit 1401f includes: texture normalization for mapping the phase consistency information of all the acquired clear images to the mean shape of the sharp images by using the image registration algorithm based on the corresponding points. Unit (not shown); And a principal component analysis processing unit (not shown) for processing all normalized sample texture information by principal component analysis to obtain a texture parameter and a texture model.

In addition to the above differences, the iris area dividing device 1400 shown in Fig. 14 is the same as the iris area dividing device 1300 shown in Fig. 13.

According to the apparatus of the present invention, the sample image phase coincidence information calculating section that calculates the phase coincidence information of the human eye image for each clear image is included in the active appearance model establishing means, and the calculated phase consistency information is used to establish the composition. The human eye texture model establishing portion of the texture model of the active appearance model, thereby obtaining texture information with clearer texture features of the pupil, the iris, and the upper and lower eyelids, thereby being able to establish a more accurate active appearance model than the prior art, and further, The active appearance model matching device includes an input image phase consistency information calculation portion that calculates phase consistency information for a human eye image to be subjected to iris region segmentation, and a texture and a texture of a human eye image obtained based on the calculated phase coincidence information The texture of the active appearance model is matched to the appearance texture matching portion, so that the contour of the human eye can be presented very accurately, thereby enabling an iris region segmentation device that is more accurate than the prior art.

It should be noted that the present invention can be implemented in software and/or a combination of software and hardware. For example, the various devices of the present invention can be implemented using an application specific integrated circuit (ASIC) or any other similar hardware device. In one embodiment, the software program of the present invention may be executed by a processor to implement the steps or functions described above. Likewise, the software programs (including related data structures) of the present invention can be stored in a computer readable recording medium such as a RAM memory, a magnetic or optical drive or a floppy disk and the like. Additionally, some of the steps or functions of the present invention may be implemented in hardware, for example, as a circuit that cooperates with a processor to perform various steps or functions.

It is obvious to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and the present invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the present embodiments are to be considered as illustrative and not restrictive, the scope of the invention is defined by the appended claims Meaning of the equivalent element and All variations within the scope are encompassed within the invention. Any reference signs in the claims should not be construed as limiting the claim. In addition, it is to be understood that the word "comprising" does not exclude other elements or steps. The plurality of devices or units or units recited in the device claims may also be implemented by one device or unit or unit by software or hardware. The words first, second, etc. are used to denote names and do not denote any particular order.

Claims

An iris region segmentation method is characterized in that the iris region segmentation method is an iris region segmentation method based on an active appearance model, and the method includes:

The active appearance model establishing step uses a plurality of pre-acquired human eye sample images to establish an active appearance model composed of a human eye shape model and a human eye texture model;

An active appearance model matching step of matching an input human eye image to be subjected to iris region segmentation and the active appearance model established in the active appearance model establishing step to obtain a human eye in the input human eye image Multiple feature points of the outline;

a boundary fitting step of selecting a feature point for fitting each boundary in the input human eye image from a plurality of feature points obtained in the active appearance model matching step to perform fitting to obtain a segmentation Iris area,

The phase consistency information is utilized in both the active appearance model establishing step and the active appearance model matching step.
The iris region segmentation method according to claim 1, wherein the active appearance model establishing step comprises:

a sample image phase consistency information calculating step of calculating phase consistency information for each of the plurality of pre-acquired human eye sample images;

The human eye texture model establishing step uses the calculated phase consistency information to establish a human eye texture model constituting the active appearance model.
The iris region segmentation method according to claim 2, wherein the human eye texture model establishing step comprises:

a shape dividing step of respectively determining a mean shape of the plurality of pre-acquired human eye sample images and a plurality of sample shapes obtained by marking the plurality of human eye sample images by using the calculated phase consistency information Perform Delaunay triangulation;

a texture normalization step of respectively mapping the calculated phase consistency information to the mean shape by a slice affine transformation to obtain normalized sample texture information;

The principal component analysis processing step processes the sample texture information by using a principal component analysis method to obtain a texture parameter and a texture model.
The iris region segmentation method according to claim 2, wherein the human eye texture model establishing step comprises:

a texture normalization step of mapping the calculated phase consistency information to the mean shape using an image registration algorithm based on a corresponding point to obtain normalized sample texture information;

The principal component analysis processing step processes the sample texture information by using a principal component analysis method to obtain a texture parameter and a texture model.
The iris region segmentation method according to claim 1, wherein the active appearance model matching step comprises:

Inputting an image phase consistency information calculating step of calculating phase consistency information for the input human eye image to be subjected to iris region segmentation;

Inputting an image texture calculating step of calculating a texture of the input human eye image using the calculated phase consistency information of the input human eye image;

An appearance texture matching step of matching the calculated texture of the input human eye image with the texture of the active appearance model established in the active appearance model establishing step to obtain a representation in the input human eye image Multiple feature points of the contour of the human eye.
The iris region segmentation method according to claim 1, wherein the active appearance model establishing step further comprises:

a sample image collecting step of pre-collecting the plurality of human eye sample images of the left and right eyes of different people;

a feature point calibration step of manually calibrating feature points on the plurality of human eye sample images;

a feature point alignment step of aligning corresponding feature points in the plurality of human eye sample images;

a human eye shape model establishing step of establishing a human eye shape model constituting the active appearance model by using feature points in the plurality of human eye sample images aligned in the feature point alignment step;

a step of synthesizing the established human eye shape model and the human eye texture model The rows are combined to obtain the active appearance model.
The iris region segmentation method according to claim 6, wherein in the feature point alignment step, Platts analysis is used to obtain an aligned image in which translation, scale, and rotation are removed.
The iris region segmentation method according to claim 6, wherein in the human eye shape model establishing step and the human eye texture model establishing step, the human eye shape model and the The human eye texture model.
The iris region segmentation method according to claim 1, wherein each of the boundary boundaries comprises an iris boundary, a pupil boundary, and an upper and lower eyelid boundary.
The iris region segmentation method according to claim 9, wherein when the iris boundary is fitted in the boundary fitting step, a plurality of feature points obtained in the active appearance model matching step are selected At least a portion of the feature points located on the left side border of the iris and the right side border of the iris are fitted.
The iris region segmentation method according to claim 9, wherein when the pupil boundary is fitted in the boundary fitting step, a plurality of feature points obtained in the active appearance model matching step are selected At least a portion of the feature points located on the pupil boundary are fitted.
The iris region segmentation method according to claim 9, wherein when the upper and lower eyelid boundaries are fitted in the boundary fitting step, among a plurality of feature points obtained in the active appearance model matching step Fitting is performed by selecting at least a portion of the feature points at the intermediate portion of the upper and lower eyelid boundaries that are spaced apart from the corner of the eye.
An iris region segmentation device, wherein the iris region segmentation device is an iris region segmentation device based on an active appearance model, the device comprising:

An active appearance model establishing device configured to use an image of a plurality of human eye samples collected in advance to establish an active appearance model composed of a human eye shape model and a human eye texture model;

An active appearance model matching device configured to match an input human eye image to be subjected to iris region segmentation and the active appearance model established in the active appearance model building device to obtain the input human eye image Multiple features of the human eye contour Point;

a boundary fitting device configured to select, from a plurality of feature points obtained in the active appearance model matching device, feature points for fitting respective boundaries in the input human eye image to perform fitting, To obtain the segmented iris area,

Wherein, the phase consistency information is utilized in both the active appearance model establishing device and the active appearance model matching device.
The iris region segmentation device according to claim 13, wherein the active appearance model establishing device comprises:

a sample image phase consistency information calculation section configured to calculate phase consistency information for each of the plurality of pre-acquired human eye sample images;

A human eye texture model establishing section configured to use the calculated phase consistency information to establish a human eye texture model constituting the active appearance model.
The iris region segmentation device according to claim 14, wherein the human eye texture model establishing portion comprises:

a shape dividing unit configured to obtain a mean shape of the plurality of pre-acquired human eye sample images and to mark the plurality of human eye sample images by using the calculated phase consistency information The sample shapes are respectively subjected to Delaunay triangulation;

a texture normalization unit configured to map the calculated phase consistency information to the mean shape by a slice affine transformation to obtain normalized sample texture information;

A principal component analysis processing unit configured to process the sample texture information using principal component analysis to obtain a texture parameter, a texture model.
The iris region segmentation device according to claim 14, wherein the human eye texture model establishing portion comprises:

a texture normalization unit configured to map the calculated phase consistency information to the mean shape using a corresponding point-based image registration algorithm to obtain normalized sample texture information;

a principal component analysis processing unit configured to process the sample using principal component analysis This texture information is used to obtain texture parameters and texture models.
The iris region segmentation device according to claim 13, wherein the active appearance model matching device comprises:

An input image phase coincidence information calculation section configured to calculate phase consistency information for the input human eye image to be subjected to iris region segmentation;

An input image texture calculation section configured to calculate a texture of the input human eye image using the calculated phase consistency information of the input human eye image;

An appearance texture matching portion configured to match the calculated texture of the input human eye image with the texture of the active appearance model established in the active appearance model building device to obtain the input person Multiple feature points of the human eye contour in the eye image.
The iris region segmentation device according to claim 13, wherein the active appearance model establishing device further comprises:

a sample image acquisition unit configured to pre-collect the plurality of human eye sample images of the left and right eyes of different people;

a feature point calibration portion configured to manually calibrate feature points on the plurality of human eye sample images;

a feature point alignment portion configured to align corresponding feature points in the plurality of human eye sample images;

a human eye shape model establishing portion configured to establish a human eye shape model constituting the active appearance model by using feature points in the plurality of human eye sample images aligned in the feature point alignment portion; as well as

A synthesis section configured to combine the established human eye shape model and the human eye texture model to obtain the active appearance model.
The iris region segmentation device according to claim 18, wherein the feature point alignment portion is configured to perform a Platts analysis to obtain an aligned image in which translation, scale, and rotation are removed.
The iris region dividing device according to claim 18, wherein the human eye shape model establishing portion and the human eye texture model establishing portion are configured to utilize a main assembly The human eye shape model and the human eye texture model are obtained by sub-analysis.
The iris region segmentation device according to claim 13, wherein each of the boundary boundaries includes an iris boundary, a pupil boundary, and an upper and lower eyelid boundary.
The iris region dividing device according to claim 21, wherein when the iris boundary is fitted in the boundary fitting device, a plurality of feature points obtained by the active appearance model matching device are selected to be located At least a portion of the feature points on the left side boundary of the iris and the right side boundary of the iris are fitted.
The iris region dividing device according to claim 21, wherein when the pupil boundary is fitted in the boundary fitting device, a plurality of feature points obtained by the active appearance model matching device are selected to be located At least a portion of the feature points on the pupil boundary are fitted.
The iris region dividing device according to claim 21, wherein when the upper and lower eyelid boundaries are fitted in the boundary fitting device, a plurality of feature points obtained by the active appearance model matching device are selected At least a portion of the feature points located at an intermediate portion of the upper and lower eyelid boundaries at a distance from the corner of the eye are fitted.