KR20160000531A - Pupil acquisition method using binary of adjacent sum and control device for extracting pupil using the same - Google Patents

Abstract

A pupil extraction method using neighbor aggregation binarization is disclosed. To this end, the present invention comprises: a step of removing glints appearing in image information acquired by photographing an eye part; a neighbor aggregation binarization step of comparing the glint-removed image information with a set reference grey level value so as to correct the image information; a step of determining a pupil region by using the center of gravity of the image information obtained by the neighbor aggregation binarization step; and a step of aggregating boundary points between a pupil and an iris by using the center of gravity in the determined pupil region.

Description

TECHNICAL FIELD [0001] The present invention relates to a pupil extraction method using adjacent sum binarization and a pupil extraction control method using the same.

The present invention relates to a pupil extraction method using adjacent sum binarization and a pupil extraction control apparatus using the same. More particularly, the present invention relates to a pupil extraction method and a pupil extraction method using a curved surface and a circular shape detection method for extracting a pupil from an iris It requires a lot of calculation amount, it takes much time to confirm the identity, it is difficult to recognize the iris due to a malfunction due to a complicated calculation amount, and the cost of installation is increased due to an increase in CPU usage due to a large amount of computation A problem has occurred.

In order to solve the problems of the related art, the present invention relates to a pupil extracting method which uses adjoining summation binarization instead of simple binarization and which can determine the center of gravity in the pupil by a simple operation compared with the conventional art, and a pupil extraction controlling apparatus using the same .

The present invention provides a pupil extraction method using adjacent sum binarization that measures the boundaries of pupil and iris with a smaller amount of computation than the prior art, extracts pupils based on the measured values, and a pupil extraction control apparatus using the same. have.

To facilitate understanding of the present invention, a general description of a general iris recognition technique is as follows.

There are many ways to identify individuals. Among them, biometrics is getting the spotlight as the most stable and accurate method of identifying individuals. Biometrics is a method of identifying individuals based on individual physical (biological) and behavioral characteristics, such as fingerprints, faces, irises, and long passages. Physical characteristics include signatures, voices, . ≪ / RTI > Personal identification and security based on the characteristics of such individuals can not be delivered by theft or leakage, and there is no danger of being changed or lost. Therefore, the audit function can be completely constructed .

In particular, among various biometric methods, iris is known to be most excellent in terms of uniqueness, invariance, and stability in discriminating individuals, and has a tendency to be applied to fields requiring high security because of low recognition rate.

The iris is formed all before birth at 3 years of age. It is known that iris does not change during a lifetime without special trauma. It has more various patterns than fingerprints and is known as the most perfect person identification method so far. Therefore, it is expected that the market potential will be very high because the user convenience is very high and the forgery is impossible and the stability is very high.

On the other hand, in a method of recognizing a specific individual using an iris, it is indispensable to detect pupil and iris for real-time iris recognition from image information of a human eye.

John G. Daugman's patent for iris recognition also finds a circle that closely approximates the boundary between the iris and the pupil, approximates the boundary between the iris and the sclera with the same principle, The iris feature information extracted through the Gabor transform from each selected region is stored in an iris code as a storage means Then, the iris feature information extracted at the time of identity verification is compared with the registered iris feature information of each selection region. If it is determined that the difference is within the threshold value, it is determined that the same person is registered.

That is, it is important to accurately extract the boundary line between the pupil and the iris. In this case, the extraction of the pupil plays a very important role. For this reason, it is necessary to accurately extract the pupil to accurately extract the iris The iris region data can be collected and compared with the previously stored iris region data to confirm the identity.

On the other hand, Dogman's patented technique for detecting the boundary between pupil and iris defined a pupil boundary as a circle for a monochrome image as a circle.

In the above equation, r is the radius of the pupil boundary, and x ₀ , y ₀ Is the center coordinate, and I (x, y) is the brightness of the pixel at that coordinate.

That is, a point at which the change in the brightness of the circumference indicates the maximum is determined as a boundary. It is difficult to process the image in real time because it requires a large amount of calculation because each radius of each pixel in the central search area must be searched for by searching for the boundary while changing the center and the radius.

In addition, although the shape of the pupil is defined as a circle, since the majority of the pupil is actually a distorted circular shape, there is a problem that the circular boundary is different from the actual pupil boundary.

As a result, when extracting the boundary between the pupil and the iris by using the Dogman patent, it is time consuming to search the boundary when the center and the radius are changed, and all the pixels existing in the center search area The time and CPU usage are increased, and there is a problem that the iris recognition probability is high due to a large amount of computation in the iris recognition due to many variables and integrations included in the expression itself.

The general method of the iris recognition system using Dogman patent, the problems of the respective methods, and the background of the present invention will be described as follows.

(2) Detection using Hough Transform; (3) Detection using Canny Contour Detector; (4) Histogram; Detection.

As described above, the pupil boundary detection method using the Dogman patent is a method of finding a point having the maximum change rate of the circumference by using various centers and radii. However, since the pupil has a characteristic of brightness and a circular shape, Since the circular detector regards the pupil as a circle, it can not accurately capture the geometric change of the pupil due to the contraction and relaxation of the iris muscle. To compensate for this, it is necessary to shoot under constant illumination at all times, and expensive equipment is required .

In order to find the maximum value, the radius and the center must be changed and many calculations must be performed. Therefore, the memory cost and the calculation cost are inevitably increased.

The present invention provides an algorithm capable of detecting a simpler pupil and iris boundary so as to reduce many calculations, memory costs, and computation costs in the iris recognition using the Dog patent.

On the other hand, the Hough Transform method, which is one of the methods for detecting using the edge component of a pupil, is a method of detecting a complex pattern in a binary image.

This Hough Transform technique is performed by determining the parameter values that determine the characteristics of a specific pattern. The components of the spatially scattered patterns are transformed into characteristics that occupy a narrow space in the parameter space of the pattern through the Hough Transform . Therefore, Hough Transform can resolve problems easily by solving problems that are difficult to solve in image space into parameter space. However, it has a merit in memory cost and computation cost compared to the circular detection method described above, but it has a drawback in that it can not solve the error that occurs when detecting a pupil of a geometric shape that is not a circular shape.

On the other hand, even if the pupil is a geometric shape, the present invention detects a center of gravity in the pupil exactly at an arbitrary point in the pupil through a simpler algorithm than the conventional one, and uses the minimum circular method starting from this arbitrary point to determine the pupil and iris Boundary detection is faster and more accurate than the Hough Transform method, which can detect the boundary between pupil and iris.

On the other hand, the Canny filter used in the detection using the Canny outline detector is a type of edge detection filter, which smoothes the image using a Gaussian filter and then detects a boundary using a Sobel operator This method is disadvantageous in that it is not easy to detect an outline when the binarization process is difficult and the eyebrow or iris pattern is strong.

On the other hand, the present invention introduces the adjacent sum binarization that corrects each pixel by comparing the sum of the gray level values of the surrounding pixels of each pixel constituting the image information, not the simple binarization, with the set reference value, and furthermore, By removing the glints present in the pupil or pupil and iris boundaries before the process, we provide a method to more efficiently handle the binarization process, which is difficult to handle in detection using the Canny outline detector.

Finally, the histogram analysis method is a method of extracting pupil by binarizing the image by finding a value corresponding to the pupil part in the histogram of the input image. The method has a simple and fast implementation, and this method has a low calculation cost and simple implementation However, it is not easy to find the maximum and minimum points, and a method of determining the pupil area using illumination may cause a lot of errors. In addition, when the user wearing glasses or the image acquisition equipment are different, the histogram exposes various problems such as sensitive change.

On the other hand, the present invention does not find the maximum and minimum points as in the histogram analysis as described above, and does not search the pupil area using the illumination. The adjacent summation binarization process is performed instead of the simple binarization, and an arbitrary point in the pupil area is selected The algorithm itself is simple to implement.

The method and apparatus for extracting the boundary between the pupil and iris are disclosed in various prior art documents.

For example, Korean Patent Registration No. 10-0373850 (Mar. 23, 2003) entitled " Identification system using iris and method and a recording medium storing a computer program source for the method ", Korean Patent Registration No. 10- And a method of extracting a pupil using an eye image. However, in the present invention, the gray level value of the surrounding pixels is used centering on one pixel rather than a simple binarization Or the glint present in the pupil or iris boundary can be efficiently removed by using the gray level value of the pixels in the pupil region existing around the glint, There is a technical limitation in that an algorithm for quickly and accurately detecting any one point in the image is not implemented.

Meanwhile, in the conventional iris recognition technology, a control unit for performing an algorithm for recognizing an iris is installed, and the control unit searches for a gray level value among iris images captured in real time through a camera.

In this case, a lighting device for photographing the iris is installed around the device for performing iris recognition irrespective of the surrounding brightness, and illuminates the focus area of the camera for photographing the iris. The light of the lighting device is reflected by the iris and the pupil The light is glint.

In extracting the pupil, it is difficult to extract the pupil when a part of the pupil is covered with eyebrows or eyelids, and thus iris detection becomes difficult. Of course, there is a problem in extracting the pupil as well as eyebrows or eyelids The difficult thing is Glint.

In other words, the brightness and size of the glint vary, and iris recognition methods for accurately determining the pupil, iris, and sclera with grayscale values or gray level values can not accurately recognize the iris or pupil due to the glint, The intensity of contrast is rapidly changing. The most unexpected part is that the iris is considered to be present around the glint, which makes it impossible to confirm the identity of the glint, Or glint at the pupil and iris boundaries.

Of course, the gray level value of the pixels constituting the glint can be corrected by selecting an arbitrary value belonging to the range of the gray level value of the pupil existing around the glint, but this is not accurate, The gray level value of the pixels constituting the glint is constant at the arbitrarily selected gray level value and the gray level value of the pixels constituting the pupil existing in the vicinity of the glint is within a certain range There arises a problem that iris recognizing apparatuses do not recognize the pupil due to unbalance or abrupt change in the gray level value.

The technique of iris recognition in relation to removal of glint is disclosed in various prior arts. Korean Patent Registration No. 10-0826876 (Apr. 25, 2008) entitled " Iris Detection Method and Apparatus therefor " The present invention provides a method and apparatus for detecting an iris region even in image information obscured by an iris, but still recognizes the problem of the glint existing in the pupil or the pupil and the iris boundary, According to the "iris recognition system, method thereof and security system for a wireless communication device using the same, " Korean Registered Patent No. 10-1178855 (2012.08.27), which has a limitation in extraction, And the number of pixels constituting the illumination portion in the pupil is counted, whereby a clearer image can be selected. However, if we look at the logic, i) it is judged using a sudden change in gray level value against the pupil and illumination, ii) one threshold value is required for filtering the input image, and iii) Values are set so that the number of pixels smaller than the second threshold value is counted, and iv) a third threshold value is set in the final stage, Not only this complexity, but also various problems such as an increase in the CPU capacity due to the delay of the detection time occur.

The present invention provides a method for eliminating glint which is hardest to process at the time of detecting the boundary between pupil and iris as described above. After eliminating the glint, it is possible to perform an approximate summation binarization instead of a simple binarization, We propose a method to find the center of gravity of the pupil region by a simple algorithm compared with the conventional one.

It should be understood that the foregoing description of the background art is merely for the purpose of promoting an understanding of the background of the present invention and is not to be construed as adhering to the prior art already known to those skilled in the art.

Korean Patent Registration No. 10-0373850 (Mar. 23, 2003) Korean Patent Registration No. 10-0376415 (Mar. 2003.05) Korean Patent Registration No. 10-0826876 (Apr. 25, 2008) Korean Registered Patent No. 10-1178855 (Aug. 27, 2012)

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above-mentioned problems of the prior art as follows: (1) Efficient removal of a glint present inside a pupil or between a pupil and an iris boundary; (2) (3) Pupil extraction method using neighboring summation binarization that solves the problem of extracting the boundary between pupil and iris by curved and circular detection, which was used in conventional pupil extraction, and its use And an object thereof is to provide a pupil extraction control device.

A pupil extraction method using adjacent sum binarization is introduced.

To this end, according to the present invention, there is provided an image processing method comprising the steps of: removing, in image information obtained by photographing an eye part, a glint appearing in the image information; A neighboring summation binarization step of comparing the image information from which the glint has been removed with a preset reference gray level value to correct the image information; Determining a pupil region using the gravity center of the image information by the adjacent summation binarization step; And summing the pupil and iris boundary points using the center of gravity in the determined pupil region.

The step of removing the glint may include the steps of: identifying a gray level value of a pupil region adjacent to the glint in the image information; correcting the glint using the gray level value; And removing the glint present in the pore.

Wherein the step of verifying the gray level value comprises: dividing the image information into a plurality of pixels; setting a reference pixel within the glint so that a gray level value of a surrounding pixel surrounding the reference pixel is greater than a gray level Value range, and then determines an average value of the gray level values of the surrounding pixels included in the gray level value range of the pupil region to check the gray level value of the pixel.

And the reference pixel is a center of gravity of the glint.

And sets the left neighboring neighboring pixel of the reference pixel as a new reference pixel if the gray level value of the neighboring pixel is out of the gray level value range of the pupil region, Sets the upper neighboring peripheral pixel as a new reference pixel when the gray level value of the upper neighboring neighboring pixel is out of the gray level value range of the pupil region, And sets the lower neighboring neighboring pixel as a new reference pixel if the gray level value of the right adjacent neighboring pixel is out of the gray level value range of the pupil region.

The step of correcting the glint is characterized in that the gray level value of the identified adjacent pupil region is applied to the inside of the glint.

The correction process may include tracking a path through which the correction process is performed to extract pixels corresponding to the remaining glint, setting the remaining glint to a new reference pixel, And corrects the gray level value of the new reference pixel with an average value of the gray level values of the surrounding pixels included in the gray level value range of the reference pixel.

Wherein the correction process tracks a path where the correction process is performed so that when the gray level value of the surrounding pixels surrounding all the pixels existing in the glint is included in the gray level value range of the pupil area, Is completed.

Wherein the step of correcting the glint comprises: if the glint is at the boundary of the pupil and the iris, prior to applying the gray level value of the identified adjacent pupil region to the inside of the glint, A first point and a second point among a plurality of points of the plurality of straight lines intersecting with the pupil region are determined and then the first point and the second point are connected to each other so as to be close to the pupil region And the glint region is corrected to the pupil region.

Wherein the first point is a point where the plurality of straight lines are rotated in one direction and two points are detected on the way of detecting points included in the gray level value range in the pupil region, Wherein the second point is a point where the plurality of straight lines are rotated in the other direction and one point included in the gray level value range within the pupil area is continuously detected and is no longer detected Is a pixel positioned on the pupil side at a time point when the pupil is not in the pupil plane.

Wherein removing the glint present in the pupil is performed by merging the corrected glint into the glint indicated in the image information.

Wherein the neighboring summation binarization step divides the image information into a plurality of pixels, sets each of the divided pixels as a main pixel, and sets a sum of a gray level value of the main pixel and surrounding pixels surrounding the main pixel, A first secondary binarization step of comparing a first reference value and correcting a gray level value of the main pixel to expand a bright area of the image information; and a second binarization step of dividing the image information corrected by the first binarization step into a plurality of pixels , Sets each of the divided pixels as a main pixel, compares the sum of the gray level values of the main pixel and surrounding pixels surrounding the main pixel with a set second reference value, and corrects the gray level value of the main pixel And a second binarization step of expanding the dark region of the image information.

The step of determining the pupil region may include calculating an N-1 imaginary center of gravity of an image filtered according to a reference set along rows and columns in the profile information of the row and column of the adjacent summed binarized image; Reversing the neighboring summation binarized image and adopting only the pixels located within the circle estimated with the N-1 imaginary center of gravity as the center of the circle in the reversed image, Calculating a virtual center of gravity; And when the Nth imaginary center of gravity satisfies the following equation, the pupil region is determined.

(X _n And Y _n The coordinates of the Nth virtual gravity center, and a and b are constants)

The calculating of the (N-1) imaginary gravity center may include: vertically reversing the adjacent sum image; A horizontal image extracting step of consecutively contiguous dark area pixels along a row of the reversed image; A vertical image extracting step of consecutively contiguous dark area pixels along the column of the reversed image; And calculating the N-1 imaginary center of gravity using coordinate values of the pixels constituting the superimposed image after extracting the superimposed image of the horizontal image and the vertical image.

The calculating of the Nth imaginary center of gravity may include calculating an N-1 left pixel and an N-1 right pixel located at both ends of a group of pixels at the top when extending vertically along the N-1 imaginary center of gravity, Calculating; Calculating an N-1 virtual circle center (X _{n -1,} Y _{n -1} ) determined by an X coordinate average value of the pixels corresponding to the both ends and a Y coordinate of the (N-1) imaginary center of gravity; Calculating a circle having a radius (r _{n -1} ) between a distance between the (N-1) _th imaginary circle center and the (N-1) _th left pixel or the (N-1) _th right pixel; The N th imaginary center of gravity is calculated using the coordinates of the pixels constituting the adopted dark region after adopting only the pixels in the dark region located inside the circle.

The formula satisfying that the pupil region is determined satisfies the following expression.

(r _{n -1} - r _n ) / r _{n -1} 5/100

(r _n The center of gravity of the Nth virtual gravity is the center of the circle, and the radius of this circle)

The r _n Calculating an Nth left pixel and an Nth right pixel located at both ends of a group of pixels that meet at the top when extending vertically along the Nth imaginary center of gravity; Calculating an X-coordinate average value of pixels corresponding to the both ends and an N- _th virtual circle center (X _n, Y _n ) determined as a Y-coordinate of the Nth virtual gravity center; The distance between the Nth imaginary circle center and the Nth left pixel or the Nth right pixel is calculated as a radius r _n .

The step of summing the boundary points of the pupil and the iris includes the steps of checking the points where the light and shade are changed while rotating in one direction about the center of gravity of the determined pupil region and extracting the pupils through the minimum circular method using these points .

On the other hand, a pupil extraction control device is introduced.

To this end, according to the present invention, there is provided a stereoscopic image processing apparatus, comprising: a glint extracting unit for extracting an image of an eye part of an individual by using a gray level value of a pupil area adjacent to a glint appearing in the photographed image information; An adjacent summation binarizer for comparing the image information composed of a plurality of pixels from which the glint is removed with a set reference value to expand a dark region and a bright region of the image information; A pupil processor for determining a pupil region using the center of gravity of the binarized image information; And a pupil detection unit for extracting the pupil by summing the boundary points of the pupil and the iris by using the center of gravity in the determined pupil region.

Wherein the glint extracting unit comprises: a database for storing the photographed image information; a glint extracting unit for extracting a glint from the photographed image information, binarizing the image information, and expressing the image information in black and white, wherein the glint is expressed as a gray level value Determines whether or not the pupil corresponds to an inner region of the pupil of the image information region, and if the glint is within the pupil region, corrects the glint to the gray level value range of the pupil to determine the pupil region and extracts the pupil region An iris processing unit for determining an iris region; And a comparison unit for determining whether or not the image information previously stored in the database unit matches the iris region determined by the glint processing unit.

The iris processing unit may further include a glint extraction module that divides the image information into a plurality of pixels and determines the glint value and extracts the image if the gray level value of the pixel is equal to or greater than a reference gray level value.

The iris processing unit may mask the extracted glint to the image information represented by the binarized and monochrome image to determine whether the glint corresponds to the inner region of the pupil of the image information region, And a determination module.

Wherein the iris processing unit divides the image information into a plurality of pixels and sets a reference pixel within the glint so that a gray level value of a surrounding pixel surrounding the reference pixel is included in a gray level value range of the pupil area And a glint correction module for correcting the average value of the gray level values of the surrounding pixels included in the gray level value range of the pupil region to the gray level value of the reference pixel.

In the glint correction module, the reference pixel is a center of gravity of the glint.

And sets the left neighboring neighboring pixel of the reference pixel as a new reference pixel if the gray level value of the neighboring pixel is out of the gray level value range of the pupil region, Sets the upper neighboring peripheral pixel as a new reference pixel when the gray level value of the upper neighboring neighboring pixel is out of the gray level value range of the pupil region, And sets the lower neighboring neighboring pixel as a new reference pixel if the gray level value of the right adjacent neighboring pixel is out of the gray level value range of the pupil region.

The correction process may include tracking a path through which the correction process is performed to extract a pixel corresponding to the remaining glint, setting the remaining glit to a new reference pixel, And corrects the gray level value of the new reference pixel with an average value of the gray level values of the surrounding pixels included in the gray level value range of the reference pixel.

Wherein the correction process tracks a path where the correction process is performed so that when the gray level value of the surrounding pixels surrounding all the pixels existing in the glint is included in the gray level value range of the pupil area, Is completed.

If the glint is present at the boundary between the pupil and the iris, a plurality of straight lines passing through the center of gravity of the glint are assumed, and a first point and a second point among a plurality of points, And then connecting the first point and the second point to correct the glint region close to the pupil region to the pupil region.

Wherein the first point is a point where the plurality of straight lines are rotated in one direction and two points are detected on the way of detecting points included in the gray level value range in the pupil region, Wherein the second point is a point where the plurality of straight lines are rotated in the other direction and one point included in the gray level value range within the pupil area is continuously detected and is no longer detected Is a pixel positioned on the pupil side at a time point when the pupil is not in the pupil plane.

The adjacent summation binarization unit divides the image information into a plurality of pixels, sets each of the divided pixels as a main pixel, and outputs a sum of a gray level value of the main pixel and neighboring pixels surrounding the main pixel, A first order binarization module for comparing a reference value of the main pixel with a gray level value of the main pixel to expand a bright region of the image information, and a second binarization module for dividing the image information corrected by the first binarization step into a plurality of pixels And the gray level value of the main pixel is corrected by comparing the total sum of the gray level values of the main pixel and the surrounding pixels surrounding the main pixel with the set second reference value, And a second binarization module for expanding a dark region of the image information.

The pupil processing unit may include an (N-1) th calculation module for calculating an (N-1) imaginary center of gravity of an image filtered according to a reference set along a row and a column in the profile information of the row and column of the adjacent summed binarized image, And then the neighboring sum binarized image is reversed and only the pixels located within the circle estimated with the N-1 imaginary center of gravity as the center of the circle in this reversed image are used, and then the Nth virtual And a first determination module that determines the pupil area by determining whether the Nth imaginary center of gravity satisfies the following equation.

(X _n And Y _n The coordinates of the Nth virtual gravity center, and a and b are constants)

Wherein the N-1 th calculation module comprises: an N-1 calculation module for reversing the adjacent summed image up and down, and for successively contiguous neighboring images along a column with a horizontal image consisting of successive contiguous dark pixels along a row of the reversed image Extracts a vertical image composed of pixels in a dark region, extracts a superimposed image of the horizontal image and the vertical image, and calculates the N-1 imaginary center of gravity using coordinate values of the pixels constituting the superimposed image .

The N th calculation module calculates an (N-1) th left pixel and an (N-1) th right pixel positioned at both ends of a pixel group meeting at the uppermost end when extending vertically along the (N-1) imaginary center of gravity, 1 imaginary circle center (X _{n -1,} Y _{n -1} ) determined by the X coordinate average value of the pixels corresponding to the (N-1) _th virtual gravity center and the Y coordinate of the (N-1) imaginary gravity center, (R _{n -1} ) between the first virtual circle center and the (N-1) _th left pixel or the (N-1) _th right pixel, and then selects only the pixels in the dark region located in the circle And the Nth imaginary center of gravity is calculated using the coordinates of pixels constituting the adopted dark region.

The equation satisfying that the pupil area is determined further includes a second determination module that determines whether or not the following expression is satisfied.

(r _{n -1} - r _n ) / r _{n -1} 5/100

(r _n The center of gravity of the Nth virtual gravity is the center of the circle, and the radius of this circle)

The second-order judging module judges whether the r _n An Nth left pixel and an Nth right pixel located at both ends of a group of pixels that meet at the uppermost end when vertically extending along the Nth imaginary center of gravity are calculated, (X _{n ,} Y _n ) determined based on the average value and the Y coordinate of the center of the N _th gravity _, and then calculates the center of the N virtual circle center (X _{n ,} Y _n ) between the Nth virtual circle center and the Nth left pixel or the Nth right center pixel And the distance is calculated as a radius (r _n ).

The pupil detection unit sums the boundary points of the pupil and the iris, checks the points where the light and shade are changed while rotating in one direction about the center of gravity of the determined pupil region, .

According to the pupil extraction method using the adjacent sum binarization and the pupil extraction control apparatus using the same according to the present invention, the following various effects can be realized.

First, there is an advantage in that the glint can be processed by correcting the gray level value of the glint, which is the most difficult to handle in the iris recognition process, by using the gray level value of the pixels constituting the pupil existing around the glint.

Second, there is an advantage in that iris recognition becomes impossible due to the glint existing inside the pupil or between the pupil and the iris, and the recognition time is delayed.

Third, since the logic for removing the glints existing in the pupil or at the pupil and iris boundaries is simpler than the conventional one, there is an advantage that the CPU capacity for performing this logic is reduced and the installation cost is reduced.

Fourth, there is an advantage that the bright and dark regions of the eye image can be more clearly distinguished through the adjacent sum binarization rather than the simple binarization.

Fifth, it is possible to extract various points of the pupil region by using a simple algorithm compared to the conventional method, thereby shortening the time required for iris recognition.

1 is an image of image information obtained by photographing an eye part including an iris and a pupil.
FIG. 2 is an image obtained by extracting a glint of a predetermined brightness or more from an image of image information obtained by photographing an eye part.
3 is an image representing the center of gravity of each of the plurality of extracted glint regions;
Fig. 4 is an image obtained by binarizing an area of an eye. Fig.
5 is an image showing the correction process of the glint present in the pupil.
6 is an image for explaining a correction process;
Figs. 7 to 9 show an image in which the correction process progresses sequentially.
10 is an image showing the center of gravity of the glint present at the pupil-iris boundary;
11 is an image for explaining the process of determining the first point and the second point;
12 is an image for explaining a closed figure by connecting a first point and a second point;
13 is an image represented by the pupil region where correction is made;
14 is an enlarged image obtained by applying the corrected image to the original image.
15 illustrates image information divided into a plurality of pixels to illustrate a first binarization step;
16 is an image corrected by the first binarization step.
17 illustrates image information divided into a plurality of pixels to illustrate a second binarization step.
18 is an image corrected by the first binarization step.
FIGS. 19 to 22 are diagrams showing a process of calculating the (N-1) imaginary center of gravity.
23 to 26 are diagrams showing a process of calculating the Nth virtual gravity center.
27 to 28 are images for explaining the process of summing the boundaries of pupil and iris.
29 is an overall configuration diagram of the pupil extraction control device;

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a preferred embodiment of a pupil extraction method using neighboring cumulative binarization and a pupil extraction control apparatus using the same according to the present invention will be described with reference to the accompanying drawings.

In the pupil extraction method using neighboring cumulative binarization according to the present invention, in the image information obtained by photographing the eye part, the step of removing the glint appearing in the image information is first performed.

For this purpose, it is necessary to check the gray level value of the pupil region adjacent to the glint in the image information, to correct the glint using the gray level value, and to remove the glint present in the pupil with the corrected glint .

On the other hand, FIG. 1 is an image showing image information obtained by photographing an eye part, and it can be confirmed that a plurality of glints (g) are scattered not only inside the pupil but also on the skin.

A pupil region is located in the periphery of the glint, and the pupil region has a gray level value of "30" or less.

The present invention first removes the glint, which is the most difficult and difficult to process in extracting the pupil to determine the boundary between the pupil and iris.

In order to remove these glints, a correction is made to convert the gray level value of the glint using the gray level value of the pupil region existing around the glint, so that the gray level value of the glint is similar to the gray level value of the pupil region To remove the glint.

Before describing the glint removing step, a pixel constituting the whole image of the present invention and a gray level value representing the brightness of the pixel of the present invention will be described in order to facilitate understanding of the present invention.

In order to conveniently acquire a certain eye image size, a rectangular capture area having a resolution constant in software is displayed on the screen of the display device. By simultaneously inputting images of both eyes of the user in accordance with the rectangular area, And acquires an image.

At this time, in order to acquire a uniform iris image of the international standard size, when the both eyes are included, the rectangular capture area is at least 640 pixels * 480 pixels, and the diameter of the iris of one eye is generally known to be 200 to 220 pixels.

In the present invention, the image size of an image including one eye region is calculated to be 640 × 480 (pixels) in the horizontal and vertical directions so as to conform to the international standard. The drawings attached to the present invention satisfy this size specification I suppose.

1 also shows an area of an eye including an iris and a pupil, and this image is also set to have a pixel size of 320 pixels by 240 pixels.

On the other hand, the gray level value is obtained by calculating the degree of darkness of each pixel based on the value of "0 to 255", and the closer the pixel value is to "0", the closer to "255" .

Generally, it is known that the gray level value range of the pixels constituting the glint by the light source is in the range of "200 to 255." In the present invention, the threshold value "225" is set to extract such glint, It can be confirmed that the gray level value of the pixels (8 bits) constituting the lint is "225" or more.

Further, it is assumed that the pupil region is composed of pixels having a black color system with a threshold value of 30 or less, and using the average value of the pixels belonging to the gray level value range of the pupil region, The process of correcting the level value again is performed.

The process of extracting glit from the image information is performed before confirming the gray level value of the pupil region adjacent to the glint in the image information.

In the image information of the attached eye of FIG. 1, pixels having a gray level value of 225 or more are extracted. In the glint extraction process, the threshold value is set so that the glints existing in the pupil or the pupil and iris boundaries are extracted Quot; 225 ".

It should be noted that this threshold value can be set differently according to the place where the iris recognition apparatus is installed and the illuminance of the light source illuminating the iris, and the image from which the glint is extracted through such threshold value is shown in FIG.

As shown, a plurality of regions composed of pixels having a gray level value of "225" or more are detected, and each of these regions may be composed of a plurality of pixels, and one region may be a glint have.

For example, in FIG. 2, it can be seen that there are a plurality of glint regions composed of a plurality of pixels at the center of the pupil, and a glint located on the skin side is present below the glint regions.

Meanwhile, as shown in FIG. 3, after the glint extracting process as described above, the glint extracts a plurality of glints, each consisting of a plurality of pixels, And a separate image representing the center of gravity is displayed.

The method of calculating the center of gravity of each glint region is calculated by the following equation.

X _c = (? X _i x M _xi ) / ΣM _{xi - - - - - - - - - - - - - - - -} (1)

Y _c = (? Y _i M _Yi ) / ΣM _{Yi - - - - - - - - - - - - - - - -} (2)

(X _c : X coordinate of the center of gravity, Y _c : Y coordinate of the center of gravity, X _i : X coordinate of each pixel, Y _i : Y coordinate of each pixel, M _xi : weight of each pixel in the X axis direction, M _Yi : weight of each pixel in the Y axis direction )

In the above equation, the weight of each pixel in the X-axis direction and the weight in the Y-axis direction are assumed to be 1.

Of course, when calculating the center of gravity of a glint region composed of a plurality of pixels in addition to the above formula, (sum of X coordinates of each pixel) / (number of pixels), (sum of Y coordinates of each pixel) Number).

The reason for calculating the center of gravity of each of the glint regions as described above is to judge whether glint exists inside the pupil or simply exists near the skin.

That is, as shown in FIG. 4, after the center of gravity is calculated as described above, a simple binarization operation is performed in which the image of the eye is simply distinguished as black and white (bright portion and dark portion) The area is discriminated in black and white, and after masking the image of FIG. 3 in which the center of gravity is expressed in FIG. 4, the center of gravity (the center of gravity is white) The coordinates of the point to be found are substituted in FIG. 4, and it is judged whether the area containing the point is a dark area, that is, a part surrounded by a pupil.

Itneunjineun surrounded by a pupil when the pupil in the direction relative to the other words, the center of gravity contrast is changed and detected by the multi-directional (360 ^o) on the basis of the center-of-gravity position, it will be a dark area is detected in all directions, in the end The center of gravity can be judged as a point formed inside the pupil.

Of course, the points that are not present inside the pupil with respect to the center of gravity in this process will not be detected even if they are detected in multiple directions. Therefore, this point is not a point formed in the pupil but a glint And will be excluded from the glint correction target.

That is, the object of the present invention is to correct the gray level value of the pixels constituting the glint in the pupil or the pupil and the iris boundary to the range of the gray level value of the pupil region, It is possible to reduce the calculation time required for correction by excluding it from the correction target.

A step of extracting the glint and confirming the gray level value of the pupil region adjacent to the glint is performed and a correction process is performed to apply the gray level value of the identified adjacent pupil region to the inside of the glint.

First, the image information obtained by photographing the eye portion is divided into a plurality of pixels, a reference pixel is set in the glint, and whether the gray level value of the surrounding pixels surrounding the reference pixel is included in the gray level value range of the pupil region The gray level value of the surrounding pixels included in the gray level value range of the pupil region is obtained and the gray level value of the pixel is checked.

The process will be described with reference to FIG. 6, which is an image represented by the gray level value of the glint present in the pupil.

As shown, the gray level value of the pixels constituting the glint is indicated as "225" or higher, and the gray level value of the pixels constituting the pupil region is expressed as "30"

The present invention sets the reference value to be at least "30" in the dark region of the pupil. However, as described above, it can be changed according to the illuminance of the iris and the illuminance of the iris.

6, pixels constituting the glint indicated by "241, 240, 227, 251" and the like, and pixels representing the pupil region composed of "17, 14, 12, 10" or the like around these pixels.

On the other hand, a pixel represented by an "X" mark appears at the center of the pixels constituting this glint, which is the center of gravity within the glint region.

Thus, the present invention sets a selected reference pixel among the pixels constituting the glint, which means the center-of-gravity pixel of the region constituted by the glint.

A correction process is performed based on the center-of-gravity pixel, and whether or not the gray level value of the surrounding pixel surrounding the reference pixel is included in the gray level value of the pupil region is determined first.

In other words, if the gray level values of a total of eight surrounding pixels surrounding the reference pixel indicated by "X" in FIG. 6 are listed, the gray level value of these peripheral pixels becomes "241,240,227,251,245,229,233,243" 30 "or less.

A simple example will be described with reference to FIG.

As shown in the figure, a glint indicative of a gray level value "251" is displayed at the center of the pixels constituting the glint.

Pixels included in the gray level value range of the pupil region such as "16,14,12" are surrounded by the center pixel.

Of course, the center of gravity of the glint will of course be the pixel itself located at the center of the figure, and the average value of the surrounding pixels is (16 + 14 + 12 + 8 + 10 + 9 + 13) /8=11.625, , And the gray level value is represented by an integer, so if the decimal point or less is omitted, the average value is calculated as "11 ".

That is, when the gray level value of the pixel represented by the glint is "251 ", it is corrected to" 11 "

The average value of "11" is the gray level value of the identified neighboring pupil region, and the process of correcting this level value to the pixel of "251 " and switching to" 11 "

On the other hand, if we look at the corrected gray level value "11", the gray level value of the actual pixel of the pupil area obscured by the glint may not match the correction value "11".

However, the gray level value ("11") of the corrected pixel and the gray level value ("16, 14, 12, 8, 10, 9, 13, 11" It can be confirmed that it is corrected to the gray level value.

With this correction, the presence of the glint represented by the gray level value of "251 " conventionally makes it impossible to determine whether the region is a pupil region, a pupil-iris boundary, or an iris region The problem is that the correction process of the present invention allows the pixels constituting the inside of the glint to be corrected approximately in the range of the gray level value corresponding to the pupil region and extracts the determined pupil from the iris, It is possible to confirm the identity of the previously stored iris accurately and quickly.

On the other hand, as shown in Fig. 6, when pixels existing around the center-of-gravity pixel do not have pixels whose gray level value of the pupil region is "30" or less, And a new reference pixel is set. The concrete procedure will be described below.

As shown, the gray level value of neighboring surrounding pixels surrounding the reference pixel represented by "X " is expressed to be larger than the gray level value" 30 "of the pupil region.

In this case, it is impossible to correct the average value of the gray level values of the surrounding pixels surrounding the reference pixel, and the " X "

At this time, the left neighboring pixel adjacent to the left side of the pixel indicated by "X" is moved to "251", and the pixel indicated by "251" is set as the new reference pixel.

The gray level value of the neighboring pixels around the new reference pixel again detects whether a gray level value smaller than the gray level value range of the pupil area is "30 ".

For example, as shown in Fig. 6, there are pixels represented by "16,22,17" around a new reference pixel indicated by "251 & And the average value of these three pixels is calculated as "251 ", and the calculated average value will be the correction value of the new reference pixel.

That is, (16 + 22 + 17) /3=18.333 is obtained, and if less than the decimal point is omitted, the correction value becomes "18 ".

On the other hand, when moving to the left in this reference pixel with the correction value of "18 ", there is a pixel of" 12 ", as shown in Fig. 7, and the gray level value of this pixel is included in the gray level value in the pupil area , This pixel is excluded from the correction target, and the upper neighboring peripheral pixel is shifted and this upper neighboring peripheral pixel is set as a new reference pixel.

That is, a pixel representing the gray level value of "241 " located above the pixel indicated by" 18 "in Fig. 7 is moved to, and the pixel indicated by" 241 "

(4 + 17 + 14 + 16 + 16) + (16 + 16 + 16 + 16) 22 + 18) /6=15.1667 and the correction value will be "15 ".

That is, as shown in Fig. 8, the pixel indicated by "241" is corrected and corrected to "15 & A pixel indicating the gray level value of "17 " exists on the upper side, and it can not move to the left or the upper side.

In this case, the pixel is moved to the right adjacent pixel, and the shifted right adjacent pixel is set as a new reference pixel.

In other words. As shown in Fig. 8, a pixel indicated by "240" located on the right side of the pixel corrected to "15 " is set as a new reference pixel, and is again included in the gray level value of the pupil region (17 + 14 + 12 + 15 + 18) /5=15.2, the correction value becomes "15 ".

In this way, the pixel shifted from the pixel corrected to "15 " again to the right adjacent pixel is shifted to the right neighboring pixel" 227 ", and pixels included in the gray level value of the pupil region (15 + 14 + 12 + 10 + 16) /5=13.4, the correction value will be "13".

If the gray level values of adjacent pixels on the left, top, and right sides of the reference pixel corrected to "13 " are again all gray level values of" 30 "or less, In this case, the lower adjacent pixel is moved to the lower adjacent pixel, and the lower adjacent pixel is set as a new reference pixel.

That is, the pixel indicated by "245" in Fig. 8 will be the new reference pixel, and the average value of the pixels included in the gray level value of the pupil region among the peripheral pixels of this reference pixel is found, (15 + 13 + 16 + 17) / 4 = 15.25 and the correction value will be "15 ".

In this way, as shown in Fig. 9, when a pixel of one pixel is corrected, pixels moving in the order of left, up, right, and down move to the pixel to be corrected, If present, the pixels inside the glint are corrected with the average value of those pixels.

Referring back to FIG. 9, when the moving path of the pixels to be corrected is examined, movement starts at the pixel position corrected to 15, which is the center of gravity, and 15 → 18 → 15 → 15 → 13 → 15 → 15 → 19 " to " 17 ", and if we look at the surrounding pixels of the pixel indicated by "17 ", the gray level value of all pixels is displayed as" 30 "

In this case, a path corresponding to the remaining glint is traced by tracing the path in which the correction process has proceeded, and the remaining glint is set as a new reference pixel, and the gray level value of the pupil region among surrounding pixels surrounding the pixel In the presence of the pixels included in the range, the gray level value of the new reference pixel is corrected to an average value of the gray level values of the pixels.

That is, the trace of the path where the correction process is performed means that the path is reversed on the pixel (indicated by "17") (the pixel located in the third row and the third column from the bottom of the image shown in FIG. 9) After moving from " 17 " to "19 ", i.e., from a pixel indicated by" 17 "to " 19" (pixel located in row 3 and row 4) Pixels adjacent to the right side can not move to the left or the upper side because the pixels adjacent to the left side and the upper side are represented by gray level values of "30" or less, while the pixels adjacent to the right side are "243" , And after shifting to the right side, it is corrected to an average value of pixels indicated by "30" or less around the pixel indicated by "243 ".

The average value of (17 + 14 + 19 + 17 + 24 + 27 + 5) /7=17.2857 will be "17".

If the movement path is traced backward with respect to the pixel corrected to "17 ", the pixel shifts to the pixel indicated by" 230 ", and the average value of the surrounding pixels of the pixel indicated by "230 & 16 + 0 + 15 + 4 + 17 + 17 + 8) /8=12.8571 and the correction value will be "12".

On the other hand, in the correction process described above, all the pixels in the glint are corrected. In the case of correcting the pixels constituting the glint continuously, it takes an infinite time to verify the identity, Needs to be.

To this end, according to the present invention, the entire progress path of the correction process is reversed as described above, and the gray level value of the surrounding pixels surrounding all pixels existing in the glint is included in the gray level value range of the pupil region In this case, the correction process is completed.

That is, as shown in FIG. 9, when all the pixels constituting the glint are corrected and the corrected path is traced back, the gray of the pixels surrounding each of the pixels constituting the glint When all the level values are displayed as "30" or less, all correction processes of the pixels existing in the glint are completed.

In the meantime, a so-called merging process is performed in which a correction process of checking the gray level value and obtaining an average value, applying an average value to the inside of the glint as described above, and then assigning the corrected glints to the original image information To remove the glint.

That is, as shown in FIG. 14, after correcting the gray level value of the pixels constituting the glint by the gray level value of the pupil area adjacent to the glint, the corrected glint is merged into the glint shown in the image information the pixels constituting the glint become similar to the gray level values of the pixels constituting the pupil region.

Through the process of merging the corrected glint into the image information and removing the glint, all the pixels in the glint are corrected in a similar or approximate manner to the gray level values of the surrounding pixels. Conventional problems that the pupil is not recognized due to the presence of abnormal glints are prevented in advance.

In addition, after the glint confirms the corrected pupil, the pupil is extracted from the iris, and the identity of the pupil is confirmed by comparing the extracted iris with the previously stored iris.

On the other hand, when the glint is present at the pupil-iris boundary as shown in FIG. 10, the glint present in the pupil is corrected in the above-described process. On the other hand, On the right side, there is a dark part indicating a pupil, and there is a limitation in the correction by the above-mentioned method.

To this end, the present invention is characterized in that a plurality of straight lines passing through the center of gravity of the glint are assumed, and a first point and a second point among a plurality of points meeting with the pupil region among a plurality of straight lines are determined, 2 points are connected to correct a glint region close to the pupil region to a pupil region. A detailed process will be described below.

First, whether or not the glint exists at the pupil-iris boundary is gradually shifted from the upper left corner to the next line (row) in the image of FIG. 3 in which the center of gravity is expressed, and the center of gravity The coordinate of the found point is substituted in FIG. 4, and the area containing the point is detected in multiple directions (360 ^° ) to determine whether it is surrounded by a dark area and a bright area.

At this time, as shown in FIG. 11, a dark region will be detected on the right side with respect to the center of gravity located inside the pupil, and a bright region will be detected on the left side. It will be judged as the center of gravity of the formed glint.

If it is determined that the center of gravity exists at the pupil and iris boundaries as described above, a plurality of straight lines passing through the center of gravity of the glint may be assumed, and each of the plurality of straight lines may include a gray level "30"

That is, as shown in FIG. 11, a plurality of straight lines passing through the center of gravity are detected in the dark regions existing in the pupil region. Hereinafter, the specific process will be described. However, .

Then, as shown in FIG. 12, the first point and the second point are connected to form a closed figure, and then a glint region close to the pupil region is corrected to a pupil region.

The reason for carrying out the above process is that when the glint is present in the pupil, it is surrounded by dark areas outside the glint, that is, points having a gray level value of 30 or less, It is possible to correct the level value. However, when the glint exists at the pupil-iris boundary, there is a pupil region with a gray level value of "30" or less in the rightward direction as shown in Fig. 11, This is because the iris region having a bright region, i.e., a gray level value larger than "225 " exists, and it is impossible to correct the gray level value of the pixels existing in the glint as in the case where the glint exists inside the pupil.

Hereinafter, with reference to FIG. 11, a first point and a second point are determined, and a method of connecting the first and second points will be described in detail.

The first point (A) is a point (A, B) in which a plurality of straight lines are rotated in one direction and two points are detected during the detection of points (S1 area) included in the gray level value range in the pupil area Wherein the second point (D) is a pixel (A) positioned on the pupil side when it is sensed as one point (C), and the second point (D) Is a pixel positioned on the pupil side at a time point when one point included in the pixel is continuously detected (S3 region).

That is, as shown in the figure, it is assumed that a plurality of straight lines (solid lines) passing through the center of gravity are rotated (a solid line shown in the figure) in a counterclockwise direction (approximately 0 ^o to 180 ^o ).

On the other hand, the plurality of straight lines (solid lines) start from the center of gravity, and rotate in the counterclockwise direction to detect points that are less than "30"

While continuously detecting these points (S1 area), it will detect the "A" point shown in the figure, and when detecting this " A " spot, will be.

For example, it is assumed that the straight line that detects the "A" point and the "B" point has a slope of about 150 ^° based on the center of gravity, and is slightly larger than this angle (for example, 151 ^° ) If a straight line is formed, it will detect "C", which is one point of the eyebrow area located at the upper left of the center of the pupil, without passing through the pupil area as shown.

At this time, after detecting the two points A and B, the pixel A positioned at the pupil side among the two points at the time of detection of one point C is determined as the first point A.

In other words, in the process of detecting, a plurality of straight lines starting from the center of gravity rotates counterclockwise to continuously detect points having a gray level value of 30 or less in the pupil region. At this time, Will be formed continuously (S1) until they reach the "A" point, and the "B" point on the straight line passing "A" point will also be detected.

Then, if a straight line larger than the slope of the straight line passing through the "A" point and the "B" point is assumed again, this straight line will detect the point "C" The point is determined as the first point.

In this way, although the first point A can be assumed, the first point A can also be determined in the following manner.

After passing through the center of gravity and detecting the points included in the gray level value range in the pupil area, the distance between these points is continuously calculated.

For example, if the slope of a straight line passing through the "A" and "B" points is assumed to be 150 ^o , points detected by a plurality of straight lines up to this angle will be assumed to be "S1" points in the figure.

At this time, each distance of the detected points is adjacent points, and the distance will be converged to almost zero.

Thereafter, the distance between the "A" point and the "B" point will be a predetermined distance when passing between the "A" point and the "B" point, It will be a point candidate point.

Thereafter, a plurality of straight lines will detect points present in the "S2" region, and the distances of these detected points will converge to "0" as well as the points present in the "S1"

That is, the distance of each of successive points to be detected is calculated, and the distance of each of them is converged to "0 ", and at that point, the distance is calculated as a predetermined distance, Quot; A "point, which detects a point at which a straight line passing through the center of gravity at the time of calculation at a predetermined distance is included in the gray level value of the pupil region, as the first point.

Likewise, in the method of assuming the second point (D), when a plurality of straight lines are rotated in the other direction, one point included in the gray level value range in the pupil region is consecutively detected and is no longer detected And a pixel located on the pupil side of the pixel.

That is, as shown, a plurality of straight lines starting from the center of gravity are rotated clockwise (approximately -0 ^o to -180 ^o ) to detect points included in the gray level value in the pupil region.

That is, it will detect the points existing in the area "S3 ", and at some point it will pass the point" D ", and after passing the point " D " I will not.

At this time, when the points continuously included in the gray level value in the pupil region are detected and can no longer be detected, the "D" point located on the pupil side will be determined as the second point.

Although the second point can be assumed in the same manner as described above, the second point can also be assumed by the following method.

Similarly to the previously described first point determination method, when the distances of the respective points in the area "S3" as shown in the figure are calculated, they will converge to almost zero, and when the point passes the point "D " The distance will be calculated as "∞".

That is, the second point is that when the distance of each of the points included in the gray level value range in the pupil region is converged to "0 ", and the point located at the pupil side is calculated as " .

As described above, when the first point and the second point are assumed, the first point and the second point are connected by a straight line as shown in FIG.

At this time, it is preferable that the gray level value of this straight line belongs to the area of "gray ", and the pixels adjacent to the pupil side are also preferably gray level values belonging to the" gray " .

In this way, a gray closed closed figure is created, and the pixels existing therein are corrected in the same manner as the correction method of the glint existing in the pore described above. As a result, the glint It will also be calibrated.

The pupil region corrected in this manner is shown in Fig. 13, and Fig. 14 is an enlarged image obtained by applying the corrected image to the original image.

As can be seen, the glint inside the pupil is almost certainly corrected, but the glint at the pupil-iris boundary has some errors compared to the glint present inside the pupil, Since the correction is performed, the conventional problem of not extracting the pupil due to the boundary does not occur.

As described above, the step of removing the glint from the image information of the eye part has been described.

Hereinafter, a neighboring summation binarization step in which image information from which glint is removed is compared with a set reference gray level value is described.

The adjoining summation binarization step first divides the image information into a plurality of pixels, sets each of the divided pixels as a main pixel, and calculates a sum of a gray level value of a main pixel and surrounding pixels surrounding the main pixel, A first binarization step of comparing the reference value and correcting the gray level value of the main pixel to expand the bright region of the image information is performed.

The first binarization step realizes the effect that the bright region is expanded around the pixel representing the bright region.

The first binarization step will be described in detail with reference to FIG.

15, the image information is divided into a plurality of pixels, and each of the pixels is represented by a gray level value.

The gray level value of the main pixel is compared with the gray level value sum of the main pixel and the surrounding pixels surrounding the main pixel to compare the gray level value sum of the main pixel with the gray level value sum already set, The level value is corrected.

The set first reference value is set to the gray level value "256 ", but the first reference value may be changed depending on the place where the iris recognition device is installed or the illuminance of the light source illuminating the iris.

The correction process is as follows for a pixel located in the fourth row and third column (from the top) and representing the gray level value "70 ".

16 + 48 + 75 + 22 + 70 + 63 + 17 + 35 + 55 = 401 "when the sum of the main pixel representing" 70 "and the eight pixels surrounding the main pixel is calculated, Quot; 255 " represented by the brightest gray level value among the gray level values of "0 to 255 "

That is, a pixel initially represented by "70" is corrected to "255 ". Similarly, when the pixel representing "22" located in the fourth row and the second column is corrected, this value is "256 + , It is corrected to a gray level value of "0 " representing the darkest brightness.

The gray level value corrected by this first binarization step is shown in the right drawing of Fig.

As shown, pixels having a gray level value of "0" representing the darkest region are also increased, whereas pixels representing a gray level value "255" It can be seen that pixels of "255" representing the brightest gray level value around the pixel are further increased.

The newly corrected image by this first binarization step is shown in FIG.

On the other hand, a second binarization step is performed based on the image corrected by the first binarization step, wherein the second binarization step includes a step of binarizing the image information corrected by the first binarization step to a plurality of pixels And the gray level value of the main pixel is corrected by comparing the sum of the gray level values of the main pixel and the surrounding pixels surrounding the main pixel with the set second reference value, The dark region of the image information is expanded.

The specific correction process will be described with reference to FIG.

The main pixel is subjected to all the pixels corrected by the first binarization step, and the second reference value is set to the gray level value "2040 ".

On the other hand, the second reference value is eight times the gray level value "255" representing the brightest region, and its technical meaning is that the gray level values of the main pixel and surrounding pixels surrounding the main pixels are both & , Eight pixels out of a total of nine pixels including the main pixel and surrounding pixels surrounding the main pixel are corrected to a gray level value of "255 " , It means that the main pixel is corrected to a gray level value of "0 " representing the darkest region.

That is, if the sum of the gray level values of the main pixel and surrounding pixels surrounding the main pixel is less than or equal to "2040 ", the main pixel is corrected to a gray level value of" 0 "Quot; 255 "in the case where all nine pixels are represented by" 255 ").

In this way, the image information corrected by the secondary binarization step is shown on the right side of Fig. 17, and as shown in Fig. 17, when most of the pixels are corrected to the gray level value of "0 " .

The image information corrected in this way is shown in Fig. 18, and the adjacent binarization, which is an essential step of the present invention, realizes the following effects.

First, the pixel representing the brightest region in the image information is extended to some extent, but the pixel representing the dark region is further expanded in the center of the pupil. Thus, when extracting an arbitrary point within the pupil, .

Next, an effect that unnecessary in the image information and dark parts of local parts such as eyebrows or points, which are unnecessary elements when extracting a certain point inside the pupil, can be eliminated.

Meanwhile, the step of determining the pupil region using the center of gravity of the image information corrected by the adjacent summation binarization step as described above is performed.

The technical meaning of determining the pupil region is that the dark region is concentrated at the center of the pupil and only the pupil appears at the image information as much as possible. This pupil region is determined, and the center of gravity of the corrected image information is used And this center of gravity is set to an arbitrary point of the pupil region at a later time.

The step of determining the pupil area may include calculating the N-1 imaginary center of gravity of the filtered image according to the criteria established along the rows and columns in the profile information of the row and column of the adjacent summed binarized image do.

The step of calculating the (N-1) imaginary gravity center may include: vertically reversing the adjacent sum image, a horizontal image extracting step of consecutively adjacent dark region pixels along a row of the reversed image, A vertical image extracting step of extracting a superimposed image of a horizontal image and a vertical image by using a coordinate value of pixels constituting the superimposed image, The (N-1) imaginary center of gravity is calculated.

The specific procedure will be described with reference to FIGS. 19 to 22. FIG.

First, as shown in Fig. 19, the process of reversing the adjacent summed image up and down is performed.

The reason for rotating the image 180 to ⁰ degrees vertical is, there is the detection of consecutive pixels in the dark area adjacent to, look at the adjacent combined image, bar eyebrows are located on the pupil, the detected initial to non-eyebrow pupil region of the process To be detected.

In the next step, a step of extracting a horizontal image consisting of successive contiguous dark pixels along a row of the reversed image is performed.

That is, pixels in the dark region are extracted along the line from the top to the bottom in the image shown in FIG.

A continuously adjacent dark region means a group of pixels in which two or more consecutively adjacent pixels are represented by a gray level value "0 " representing the darkest brightness in particular, and such a group of pixels includes two or more It is obvious that a group of pixels can be detected.

At least two or more consecutive adjacent pixel groups can be determined to be not unnecessary images.

That is, through the above steps, not only the pupil region but also the eyebrow portion other than the pupil region also have a gray level value of "0 ", and the group of pixels not adjacent to each other by two or more pixels are excluded, The unnecessary images are filtered to find the center of gravity of the image.

The image by the horizontal image extraction step is shown in FIG. 20, and it can be seen that the filtered image still shows unnecessary images in addition to the eyebrow area as well as the pupil area.

An unnecessary image including the eyebrow area causes an error in finding the center of gravity of the pupil area. Such an unnecessary image is a result of detecting only pixels in a dark region in the row direction. To reinforce this, A vertical image extraction step is performed so as to correspond to the image extraction step.

The vertical image extraction step refers to a process of detecting pixels in a dark region continuously adjacent to the image shown in FIG. 19, and the method is the same as the horizontal image extraction step, and is omitted here.

It can be seen that the image filtered by this vertical image extraction step is shown in FIG. 21 and the group of pixels representing dark areas along the column are shown compared to the filtered image in FIG. 20, as shown.

Meanwhile, the horizontal image and the vertical image are extracted as described above, and then the superimposed image is extracted from each of the images.

As shown in FIG. 22, the superimposed image is shown in FIG. 22, and unnecessary images other than the eyebrow area as well as the pupil are filtered in comparison with the images shown in FIGS. 20 and 21. As a result, An image composed of pixels having a gray level value "0 " centered on the pupil region is calculated, and the center of gravity in the pupil can be more accurately calculated through this process.

When the superimposed image is calculated as described above, the center of gravity is obtained by using the coordinate value of the pixels constituting the superimposed image, and the center of gravity thereof is the N-1 imaginary center of gravity.

The method of calculating the center of gravity has already been described and is omitted here.

Meanwhile, a process of calculating the N-1 imaginary center of gravity and substituting the center of gravity into the reversed neighboring summation image is performed, and the N-1 imaginary center of gravity is calculated as a center A process of calculating the center of gravity of the adopted image is carried out after only pixels located inside the center of gravity of the image are taken.

A process of calculating the Nth virtual center of gravity will be described with reference to FIGS. 23 to 25 as follows.

The step of calculating the Nth imaginary center of gravity comprises: (N-1) left pixel (b) located at both ends of the group of pixels which meet at the top when vertically extending along the (N-1) imaginary center of gravity The step of calculating the N-1 right pixel c is performed.

That is, as shown in FIG. 23, the adjacent summation binarization process is performed, and the gray level value "0 ", which is vertically extended along the N-1 imaginary center of gravity a in the reversed image, Pixels are detected.

The pixel will be the (N-1) -th left pixel b, and the pixel extending horizontally along the (N-1) th left pixel b and the rightmost positioned gray level value & And set to the right pixel c.

23 shows the (N-1) th left pixel b and the (N-1) th right pixel c, and thereafter the X-coordinate average value of the pixels located at both ends and the Y- (X _{n -1,} Y _{n -1} ) (d) of the (N-1) imaginary circle determined by the coordinates.

1, the (N-1) th imaginary circle center d is shown in Fig. 24, and the (N-1) th imaginary circle center d is the center of the circle, And the distance between the right pixel c and the right pixel c is a radius (r _{n -1} ).

24 shows a circle composed of the radius (r _{n -1} ) as described above. After adopting only the pixels in the dark region located inside the circle, the coordinates of the pixels constituting the adopted dark region are used to calculate the Nth Followed by a step of calculating the virtual center of gravity.

That is, only the pixels representing the gray level value "0 " existing in the circle constituted by the radius (r _{n -1} ) are continued. By this, in finding the center of gravity of the pupil region, unnecessary images are gradually filtered Effect is realized.

As shown in FIG. 25, it can be seen that much of the eyebrow area is filtered as compared to FIG. 24, and a large number of unnecessary images are also filtered out in addition to those shown in FIG.

The Nth imaginary center of gravity is calculated using the coordinate values of the filtered pixels in FIG. 25. The method of calculating the center of gravity using the coordinate values of the pixels has already been described herein, and the filtered images 26, and it gradually becomes similar to the pupil region.

On the other hand, after calculating the Nth imaginary center of gravity as described above, the pupil region is determined when the following expression, which is the center of the Nth imaginary center of gravity, is satisfied. The following formula should be used.

(X _n And Y _n The coordinates of the Nth virtual gravity center, and a and b are constants)

That is, the N-1 imaginary center of gravity and the Nth imaginary center of gravity are obtained and the N + 1 imaginary center of gravity is obtained again. If the above process is repeatedly performed, as shown in FIG. 26, The image converged to the pupil region will be detected and at a certain moment the point at which the center of gravity does not change further will come. At this point, even if the repetitive filtering process as described above is performed, the coordinates of the center of gravity It will not change.

That is, when the coordinates of the virtual center of gravity no longer change, the filtered image is determined as a pupil region. However, when the coordinates of the center of gravity converge to constants (a, b) Can be determined as a pupil region, but the above equation ensures that the pupil region is determined even if the following expression is satisfied.

(r _{n -1} - r _n ) / r _{n -1} 5/100

(r _n The center of gravity of the Nth virtual gravity is the center of the circle, and the radius of this circle)

That is, the N-1 virtual circle when the radius of the circle to the radius of the circle, and the "r _{n -1"} as the center thereof, the N-th virtual circle center-to-center to the center of one "r _n", of the radius Even if the error rate is within 5%, the extracted image is determined as the pupil region.

Each of the circle radii is continuously calculated, and when the error rate of the radius is within 5%, the corresponding image is finally determined as the pupil region.

Of course, the error rate can be set arbitrarily, but if the error rate is 5%, the number of pixels is not so large, and if the pupil satisfying the error rate is determined, There will be.

On the other hand, in the above expression, "r _n Quot; is calculated in the same manner as in the process of calculating "r _{n -1 &} quot ;. More specifically, the process of calculating the " And an N- _th imaginary circle center (X _{n ,} Y _n ) determined by an X-coordinate average value of pixels corresponding to both ends and a Y-coordinate of the Nth imaginary center of gravity; and The distance between the center of the N virtual circle and the Nth left pixel or the Nth right pixel is calculated as a radius r _n and the method is the same as the method of calculating r _{n -1} and is omitted here.

When the pupil region is determined as described above, the step of summing the boundary points of the pupil and iris using the determined center of gravity of the pupil region is performed.

The step of summing the boundary points of the pupil and the iris is characterized in that pupil is extracted by means of a minimum circular method using points determined by checking the points where the light and shade are changed while rotating in one direction around the center of gravity of the determined pupil region .

The specific steps will be described with reference to FIGS. 27 and 28 as follows.

As shown in the figure, the points varying in brightness vary in one direction (clockwise or counterclockwise) with respect to the center of gravity.

The point where the light and shade are changed means the boundary between the pupil and the iris. However, since the eyebrows are occasionally blocked to the pupil side, it is assumed that the eyebrows exist in the range of 45 ⁰ to 135 ⁰ , The data of the points where the contrast of the part changes is discarded.

On the other hand, the exact center point of the pupil and the radius of the circle corresponding to the pupil are measured using the generally known minimum circular method based on the boundary points composed of points where the light and darkness change with respect to the center of gravity in the pupil region, The pupil is extracted.

Meanwhile, a pupil extraction control apparatus using the above method is introduced.

To this end, the present invention includes a glint extractor 100, an adjacent summation binarizer 200, a pupil processor 300, and a pupil detector 400, as shown in FIG.

Meanwhile, the glint extracting unit 100 removes the glint using the gray level value of the pupil region adjacent to the glint shown in the photographed image information, and the neighboring summation binarization unit 200 performs a process of removing the glint The pupil processing unit 300 determines the pupil region using the center of gravity of the binarized image information, and determines the pupil region using the center of gravity of the binarized image information. (400) extracts the pupil by summing the boundary points of the pupil and iris using the center of gravity in the determined pupil region.

The glint extracting unit 100 extracts the glint from the captured image information, and displays the image information in a monochrome manner. The glint extracting unit 100 binarizes the image information, And if the glint is within the pupil region, the pupil region is determined by correcting the glint to the gray level value range of the pupil, and the pupil region is extracted to extract the iris And a comparing unit 130 for determining whether the iris processing unit 120 and the iris processing unit 120 match the image information stored in the database 110 and the iris region determined by the glint processing unit.

Meanwhile, the iris processing unit 120 includes a glint extraction module 121 for dividing the image information into a plurality of pixels, determining that the gray level of the pixel is equal to or greater than the reference gray level value, Further comprising a glint position determination module (122) for masking the lint to binarized image information expressed in monochrome to determine whether the glint corresponds to the inner region of the pupil of the image information region, A reference pixel is set in the glint to determine whether the gray level value of the surrounding pixel surrounding the reference pixel is included in the gray level value range of the pupil area and then the gray level value range of the pupil area And a glint correction module 123 that corrects the average value of the gray level values of the surrounding pixels included in the reference pixel to the gray level value of the reference pixel.

The detailed process of the glint extracting unit 100 having the above-described structure has already been described and is omitted here.

Referring again to FIG. 29, the adjacent summation binarization unit 200 divides the image information into a plurality of pixels, sets each of the divided pixels as a main pixel, and sets the main pixel and surrounding A first binarization module 210 for comparing the sum of the gray level values of the pixels and the set first reference value to expand the bright region of the image information by correcting the gray level value of the main pixel, The image information is divided into a plurality of pixels, each of the divided pixels is set as a main pixel, and a sum of a gray level value sum of a main pixel and surrounding pixels surrounding the main pixel is compared with a set second reference value, And a second binarization module 220 for expanding the dark region of the image information by correcting the gray level value.

Further, the pupil processing unit 300 calculates an N-1 imaginary center of gravity of the filtered image according to the reference set along the rows and columns in the profile information of the row and column of the adjacent summed binarized image, (310), reversing the adjacent sum binarized image, and adopting only the pixels located within the circle estimated with the N-1 imaginary center of gravity as the center of the circle in the reversed image, And a first determination module (330) for determining whether the Nth imaginary center of gravity satisfies the following equation to determine a pupil area: < EMI ID = 1.0 > do.

(X _n And Y _n The coordinates of the Nth virtual gravity center, and a and b are constants)

However, the equation satisfying that the pupil region is determined further includes a second determination module 340 that determines whether or not the following expression is satisfied.

(r _{n -1} - r _n ) / r _{n -1} 5/100

(r _n The center of gravity of the Nth virtual gravity is the center of the circle, and the radius of this circle)

The specific operation of the pupil extraction control apparatus having the above-described configuration has already been described, and will not be described here.

While the present invention has been particularly shown and described with reference to specific embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the following claims It will be apparent to those of ordinary skill in the art.

100: Glint extractor 110: Database part
120: iris processing unit 130:
121: Glint extraction module 122: Glint position determination module
123: Glint correction module 200: Adjacent sum binarization unit
210: first binarization module 220: second binarization module
300: pupil processing unit 310: N-1 calculation module
320: first Nth calculation module 330: first determination module
340: second judgment module 400: pupil detection part

Claims (36)

In the image information of the eye part,
Removing glints appearing in the image information;
A neighboring summation binarization step of comparing the image information from which the glint has been removed with a preset reference gray level value to correct the image information;
Determining a pupil region using the gravity center of the image information by the adjacent summation binarization step; And
And summing the pupil and iris boundary points using the center of gravity in the determined pupil region.
The method according to claim 1,
Wherein removing the glint comprises:
Determining a gray level value of a pupil region adjacent to the glint in the image information; correcting the glint using the gray level value; and removing the glint present in the pupil with the corrected glint Further comprising the steps of: (a) extracting a pupil using adjacent binomial binarization;
The method of claim 2,
The step of verifying the gray level value comprises:
Dividing the image information into a plurality of pixels, setting a reference pixel within the glint to determine whether a gray level value of a surrounding pixel surrounding the reference pixel is included in a gray level value range of the pupil area, Wherein the average value of the gray level values of the surrounding pixels included in the gray level value range of the pupil region is obtained and the gray level value of the pixel is confirmed.
The method of claim 3,
Wherein the reference pixel is the center of gravity of the glint.
The method of claim 3,
Setting a left neighboring neighboring pixel of the reference pixel as a new reference pixel if the gray level value of the neighboring pixel is out of a gray level value range of the pupil region,
And sets the upper neighboring neighboring pixel as a new reference pixel if the gray level value of the left neighboring neighboring pixel is out of the gray level value range of the pupil region,
Setting a right neighboring neighboring pixel as a new reference pixel when the gray level value of the upper neighboring neighboring pixel is out of the gray level value range of the pupil region,
And if the gray level value of the right neighboring neighboring pixel deviates from the gray level value range of the pupil region, the lower neighboring neighboring pixel is set as a new reference pixel.
The method of claim 2,
Wherein the step of correcting the glint comprises:
Wherein the determined gray level value of the adjacent pupil region is applied to the inside of the glint.
The method of claim 6,
The calibration process may include:
The correction process is traced to extract a pixel corresponding to the remaining glint, the remaining glint is set as a new reference pixel, and a gray level value range of the pupil area of surrounding pixels surrounding the new reference pixel Wherein the gray level value of the new reference pixel is corrected to the gray level value of the neighboring pixels included in the neighboring sum binarization.
The method of claim 7,
The calibration process may include:
And the correction process is completed when the gray level value of the surrounding pixels surrounding all the pixels existing in the glint is included in the gray level value range of the pupil region by tracing the path where the correction process is performed The pupil extraction method using adjacent sum binarization.
The method of claim 6,
Wherein the step of correcting the glint comprises:
If the glint is present at the boundary between the pupil and the iris,
Before applying the gray level value of the identified neighboring pupil region to the inside of the glint,
A first point and a second point among a plurality of points meeting the pupil region among a plurality of straight lines are determined and then the first point and the second point are connected to each other And correcting the glint region close to the pupil region to the pupil region.
The method of claim 9,
Wherein the first point is a direction in which the plurality of straight lines are rotated in one direction,
And a pixel located on the pupil side when the two points are detected while one point is detected after detecting the points included in the gray level value range in the pupil region,
The second point is that the plurality of straight lines rotate in the other direction,
Wherein the pupil region is a pixel located at a pupil side at a time point when one point included in the gray level value range within the pupil region is continuously sensed and is no longer sensed.
The method of claim 2,
Wherein removing the glints present in the pores comprises:
And the corrected glint is merged into a glint appearing in the image information.
The method according to claim 1,
Wherein the neighboring summation binarization step comprises:
The method comprising the steps of: dividing the image information into a plurality of pixels, setting each of the divided pixels as a main pixel, comparing a sum of a gray level value of the main pixel and surrounding pixels surrounding the main pixel, A first binarization step of expanding a bright region of the image information by correcting a gray level value of a main pixel,
A first binarization step of dividing the image information corrected by the first binarization step into a plurality of pixels, setting each of the divided pixels as a main pixel, setting a sum of the gray level values of the main pixel and surrounding pixels surrounding the main pixel, And a second secondary binarization step of expanding a dark region of the image information by comparing a set second reference value and correcting a gray level value of the main pixel.
The method according to claim 1,
Wherein determining the pupil region comprises:
Calculating an N-1 imaginary center of gravity of the filtered image according to criteria set along rows and columns in the profile information of the rows and columns of the adjacent summed binarized image;
Reversing the neighboring summation binarized image and adopting only the pixels located within the circle estimated with the N-1 imaginary center of gravity as the center of the circle in the reversed image, Calculating a virtual center of gravity; And
And wherein the pupil region is determined when the Nth imaginary center of gravity satisfies the following equation.

(X _n And Y _n The coordinates of the Nth virtual gravity center, and a and b are constants)
14. The method of claim 13,
The step of calculating the (N-1) imaginary center of gravity comprises:
Reversing the adjacent summed image up and down;
A horizontal image extracting step of consecutively contiguous dark area pixels along a row of the reversed image;
A vertical image extracting step of consecutively contiguous dark area pixels along the column of the reversed image; And
And the N-1 imaginary center of gravity is calculated using the coordinate values of the pixels constituting the superimposed image after extracting the superimposed image of the horizontal image and the vertical image, Pore extraction method.
14. The method of claim 13,
The step of calculating the Nth imaginary center of gravity comprises:
Calculating an (N-1) th left pixel and an (N-1) th right pixel located at both ends of a group of pixels that meet at the top when extending vertically along the (N-1) imaginary center of gravity;
Calculating an N-1 virtual circle center (X _{n -1,} Y _{n -1} ) determined by an X coordinate average value of the pixels corresponding to the both ends and a Y coordinate of the (N-1) imaginary center of gravity;
Calculating a circle having a radius (r _{n -1} ) between a distance between the (N-1) _th imaginary circle center and the (N-1) _th left pixel or the (N-1) _th right pixel;
Wherein the Nth imaginary center of gravity is calculated using the coordinates of the pixels constituting the adopted dark region after adopting pixels in the dark region located in the circle and calculating the Nth imaginary center of gravity using the neighboring summation binarization .
14. The method of claim 13,
Wherein the equation satisfying that the pupil region is determined satisfies the following equation.

(r _{n -1} - r _n ) / r _{n -1} 5/100
(r _n The center of gravity of the Nth virtual gravity is the center of the circle, and the radius of this circle)
18. The method of claim 16,
The r _n Wherein:
Calculating an Nth left pixel and an Nth right pixel located at both ends of a group of pixels that meet at the top when extending vertically along the Nth imaginary center of gravity;
Calculating an X-coordinate average value of pixels corresponding to the both ends and an N- _th virtual circle center (X _n, Y _n ) determined as a Y-coordinate of the Nth virtual gravity center;
Wherein the distance between the Nth imaginary circle center and the Nth left pixel or the Nth right pixel is calculated as a radius r _n .
The method according to claim 1,
Wherein the step of summing the boundary points of the pupil and iris comprises:
Wherein pupil extraction is carried out by using a minimum circular method using points determined by checking the points where the light and shade are changed while rotating in one direction about the center of gravity of the determined pupil region, Way.
In image information obtained by photographing an eye part of an individual,
A glint extracting unit for removing the glint using a gray level value of a pupil area adjacent to the glint in the photographed image information;
An adjacent summation binarizer for comparing the image information composed of a plurality of pixels from which the glint is removed with a set reference value to expand a dark region and a bright region of the image information;
A pupil processor for determining a pupil region using the center of gravity of the binarized image information; And
And a pupil detection unit for extracting a pupil by summing the boundary points of the pupil and the iris using the center of gravity in the determined pupil region.
The method of claim 19,
Wherein the glint extracting unit comprises:
A data base for storing the photographed image information;
Extracting a glint from the photographed image information, binarizing the image information, and determining whether the glint corresponds to an inner region of a pupil of the image information region represented by a gray level value, An iris processing unit for correcting the glint to a gray level value range of the pupil to determine the pupil region and extracting the pupil region to determine an iris region if the lint is in the pupil region; And
Further comprising a comparison unit for determining whether or not the image information previously stored in the database unit matches the iris region determined by the glint processing unit.
The method of claim 20,
The iris processing unit
Further comprising a glint extraction module for dividing the image information into a plurality of pixels and judging the glint value if the gray level value of the pixel is equal to or greater than a reference gray level value.
The method of claim 20,
The iris processing unit
And a glint position determination module for determining whether the glint corresponds to an inner region of the pupil of the image information region by masking the extracted glint into the image information expressed in binary form and expressed in black and white And the pupil extraction control device.
The method of claim 20,
The iris processing unit
Dividing the image information into a plurality of pixels, setting a reference pixel within the glint to determine whether a gray level value of a surrounding pixel surrounding the reference pixel is included in a gray level value range of the pupil area, Further comprising a glint correction module for correcting the average value of the gray level value of the surrounding pixels included in the gray level value range of the pupil area to the gray level value of the reference pixel.
24. The method of claim 23,
Wherein the reference pixel in the glint correction module is a center of gravity of the glint.
27. The method of claim 24,
Setting a left neighboring neighboring pixel of the reference pixel as a new reference pixel if the gray level value of the neighboring pixel is out of a gray level value range of the pupil region,
And sets the upper neighboring neighboring pixel as a new reference pixel if the gray level value of the left neighboring neighboring pixel is out of the gray level value range of the pupil region,
Setting a right neighboring neighboring pixel as a new reference pixel when the gray level value of the upper neighboring neighboring pixel is out of the gray level value range of the pupil region,
And sets the lower adjacent pixel as a new reference pixel when the gray level value of the right neighboring peripheral pixel is out of the gray level value range of the pupil region.
26. The method of claim 25,
The calibration process may include:
A correction process is performed to track a path corresponding to the remaining glint, and the remaining glit is set as a new reference pixel, and then a gray level value range of the pupil area of surrounding pixels surrounding the new reference pixel And corrects the gray level value of the new reference pixel with the gray level value average value of the surrounding pixels included in the pupil extraction control signal.
27. The method of claim 26,
The calibration process may include:
And the correction process is completed when the gray level value of the surrounding pixels surrounding all the pixels existing in the glint is included in the gray level value range of the pupil region by tracing the path where the correction process is performed The pupil extraction control device.
24. The method of claim 23,
If the glint is present at the boundary between the pupil and the iris,
A first point and a second point among a plurality of points meeting the pupil region among a plurality of straight lines are determined and then the first point and the second point are connected to each other And corrects the glint region close to the pupil region to the pupil region.
29. The method of claim 28,
Wherein the first point is a direction in which the plurality of straight lines are rotated in one direction,
And a pixel located on the pupil side when the two points are detected while one point is detected after detecting the points included in the gray level value range in the pupil region,
The second point is that the plurality of straight lines rotate in the other direction,
Wherein the pupil extraction control device is a pixel located at a pupil side at a time point when one point included in the gray level value range within the pupil region is continuously sensed and is no longer sensed.
The method of claim 19,
Wherein the adjacent summation binarization unit comprises:
The method comprising the steps of: dividing the image information into a plurality of pixels, setting each of the divided pixels as a main pixel, comparing a sum of a gray level value of the main pixel and surrounding pixels surrounding the main pixel, A first binarization module for expanding the bright region of the image information by correcting the gray level value of the main pixel,
A first binarization step of dividing the image information corrected by the first binarization step into a plurality of pixels, setting each of the divided pixels as a main pixel, setting a sum of the gray level values of the main pixel and surrounding pixels surrounding the main pixel, And a second binarization module which compares the second reference value with the set second reference value and corrects the gray level value of the main pixel to expand the dark region of the image information.
The method of claim 19,
Wherein the pupil-
An N-1 calculation module for calculating an N-1 imaginary center of gravity of an image filtered according to a reference set along a row and a column in profile information of the row and column of the adjacent summed binarized image;
Reversing the neighboring summation binarized image and adopting only the pixels located within the circle estimated with the N-1 imaginary center of gravity as the center of the circle in the reversed image, A nth calculation module for calculating a virtual center of gravity,
Further comprising a first difference determining module for determining whether the Nth imaginary center of gravity satisfies the following equation to determine the pupil area.

(X _n And Y _n The coordinates of the Nth virtual gravity center, and a and b are constants)
32. The method of claim 31,
Wherein the (N-1)
A vertical image comprising vertically adjacent pixels along a row of the reversed image and a horizontal image consisting of pixels of successively adjacent dark regions along the column; Extracts the superimposed image of the horizontal image and the vertical image, and calculates the N-1 imaginary center of gravity using the coordinate values of the pixels constituting the superimposed image, Device.
The method of claim 19,
Wherein the Nth calculation module comprises:
1) -th left pixel and an (N-1) -th right pixel positioned at both ends of a group of pixels that meet at the uppermost end when vertically extending along the (N-1) imaginary center of gravity, 1 imaginary circle center (X _{n -1,} Y _{n -1} ) determined by a coordinate average value and a Y coordinate of the (N-1) imaginary center of gravity, (R _{n -1} ) between the (N-1) _th left pixel and the (N-1) _th right pixel, and then adopts only the pixels in the dark region located inside the circle, And calculates the Nth imaginary center of gravity using the coordinates of the constituent pixels.
32. The method of claim 31,
Wherein the equation satisfying that the pupil area is determined further includes a second determination module that determines whether or not the following expression is satisfied.

(r _{n -1} - r _n ) / r _{n -1} 5/100
(r _n The center of gravity of the Nth virtual gravity is the center of the circle, and the radius of this circle)
35. The method of claim 34,
The second-order judging module comprises:
The r _n An Nth left pixel and an Nth right pixel located at both ends of a group of pixels that meet at the uppermost end when vertically extending along the Nth imaginary center of gravity are calculated, (X _n, Y _n ) determined based on the average value and the Y coordinate of the center of the N _th gravity _, and then calculates the center of the N virtual circle center (X _n, Y _n ) between the Nth virtual circle center and the Nth left pixel or the Nth right center pixel And the distance is calculated as a radius (r _n ).
The method of claim 19,
Wherein the pupil detection unit comprises:
The pupil is summed by summing the boundary points of the pupil and the iris, and the points where the light and shade are changed while rotating in one direction about the center of gravity of the determined pupil region are checked, and pupils are extracted through the minimum circular method using these points The pupil extraction control device.

Images