An iris image positioning method

Abstract

The application provides an iris image positioning method and relates to the technical field of iris recognition. The method converts a collected color iris image into a gray image, carries out smoothing filtering on the gray image to eliminate noise influence, then binarizes the image and eliminates noise through an adaptive threshold, positions a pupil center in a manner of calculating a centroid, determines a pupil radius through the pupil center, positions an accurate pupil contour, and searches for an outer contour of the iris with the pupil center as a reference, so that image noise can be avoided, the calculation amount of edge detection can be greatly reduced, the problem that iris positioning and recognition fail due to more eyelid and eyelash shielding can be solved, and the iris texture information part in the human eye image can be more accurately and quickly segmented.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to an iris image localization method. Background Technology

[0002] Iris recognition technology is widely recognized in the industry as one of the most secure biometric technologies currently available. It is widely used in scenarios requiring precise identity verification, such as customs, prisons, medical institutions, and banks. Compared to fingerprint and facial recognition, iris recognition technology has significant advantages. First, iris recognition technology is highly stable. The shape of a person's iris is fixed after the age of one and remains almost unchanged throughout life. Furthermore, the size of the iris changes with the size of the pupil, making iris recognition suitable for liveness detection. Second, iris recognition technology is non-invasive. The iris is an externally visible organ, and collection and verification do not require physical contact, making it safer and more hygienic. Finally, the iris is unique; no two irises are exactly alike in nature, and even the irises of a person's left and right eyes, or even twins, are different.

[0003] The foundation of iris recognition technology lies in iris localization and feature extraction from iris images, with iris localization being particularly crucial as it provides accurate iris texture images for recognition. Iris localization technology relies on edge detection algorithms. However, due to interference from noise such as eyelids, eyelashes, and light spots, conventional edge detection methods lag behind in accuracy and real-time performance, failing to meet current performance requirements for iris localization. Furthermore, eyelids can partially obscure the iris image, leading to missing iris information, and high-frequency noise such as eyelashes can cause misjudgments during the circle detection stage, missing the true iris boundary.

[0004] During iris localization, interference from noise such as reflected light spots from supplementary lighting, eyelashes, and eyelids is common. Because the iris acquisition environment varies, it is necessary to dynamically set the binarization threshold to better remove noise and locate the iris contour. At the same time, it is also necessary to dynamically limit the search range of edge detection to reduce computational complexity and improve real-time performance.

[0005] Numerous methods for iris image detection have been developed, offering high accuracy and stability. However, in scenarios with complex noise, these traditional algorithms perform poorly in edge detection. Further improvements are needed to enhance the performance of these edge detection algorithms. Summary of the Invention

[0006] The purpose of this invention is to provide an iris image localization method to improve the accuracy of iris image localization and accelerate the localization speed.

[0007] The technical solution to achieve the objective of this invention is: an iris image localization method, the specific steps of which are as follows:

[0008] Step 1: Obtain an image of the iris of the eye and convert it into a grayscale image of the iris;

[0009] Step 2: Smooth the image to remove noise such as eyelashes and eyelids to obtain a denoised iris image. Fill the internal holes such as reflected light spots in the denoised iris image to obtain an adaptive binary image.

[0010] Step 3: Locate the precise pupil center coordinates using the adaptive binary image, and calculate the pupil radius based on the center to obtain the pupil outline circle;

[0011] Step 4: Extract the outer contour of the iris from the denoised iris image based on the pupil center, and obtain the accurate outer contour circle of the iris by fitting the outer contour of the iris. The iris can be separated using the accurate pupil contour circle and the accurate outer contour circle of the iris.

[0012] Optionally, the step of converting the iris image into an iris grayscale image includes:

[0013] Obtain each pixel n of the iris image of the eye. i,j The RGB intensity value of each pixel n i,j The RGB values ​​are converted to grayscale values ​​using a conversion formula to obtain the grayscale image of the iris;

[0014] The formula for converting RGB values ​​to grayscale values ​​is:

[0015] Gray(n i,j )=0.212671×R(n i,j )+0.715160×G(n i,j )+0.072169×B(n i,j )

[0016] In the formula, Gray(n) i,j ) represents the pixel n in the grayscale image of the iris. i,j The grayscale value, i represents the row number of the pixel, j represents the column number of the pixel, R(n i,j ) represents the number of pixels n i,j The red intensity value, G(n) i,j ) represents the number of pixels n i,j The green intensity value, B(n) i,j ) represents the number of pixels n i,j The blue intensity value.

[0017] Optionally, the steps of smoothing the image to remove noise such as eyelashes and eyelids to obtain a denoised iris image, and binarizing the denoised iris image to fill in internal holes such as reflected light spots to obtain an adaptive binary image include:

[0018] Anisotropic smoothing filtering is applied to the grayscale image of the iris to remove noise such as eyelids and eyelashes, resulting in a denoised iris image. Integral projection is used to coarsely locate the pupil center in the denoised iris image to determine a rough pupil center. Based on the center, an adaptive thresholding method is used to binarize the denoised iris image and separate the pupil.

[0019] Optionally, the specific steps for anisotropic smoothing filtering are as follows:

[0020] When processing pixels, first obtain the divergence of the pixel in four directions. The calculation formula is as follows:

[0021]

[0022] In the formula, I x,y I represents the intensity of the pixel at (x,y). x,y-1 This represents the intensity of the pixel at (x, y-1). This represents the divergence of pixel I in the north direction at (x,y). Let I represent the divergence of pixel I at (x,y) in the south direction. Let represent the divergence of pixel I at (x,y) in the east direction. This represents the divergence of pixel I in the west direction at (x,y);

[0023] The thermal conductivity of a pixel in four directions is calculated using the following formula:

[0024]

[0025] In the formula, cN x,y Let cS be the thermal conductivity at (x,y) in the north direction. x,y Let E be the thermal conductivity at (x,y) in the south direction. x,y Let cW be the thermal conductivity at (x,y) in the east direction. x,y Let be the thermal conductivity at (x,y) in the west direction. Let I be the divergence in the north direction of pixel I. Let I be the divergence in the south direction. Let I be the divergence in the east direction of pixel I. Let be the divergence of pixel I in the west direction, and k be the weighting coefficient;

[0026] Combining the two sets of equations above, the iterative formula for anisotropic smoothing filtering can be obtained as follows:

[0027]

[0028] In the formula I t+1 This represents the pixel intensity after iteration t+1, where t is the iteration number, λ is the smoothing control, and cN is the pixel intensity.x,y Let cS be the thermal conductivity at (x,y) in the north direction. x,y Let E be the thermal conductivity at (x,y) in the south direction. x,y Let cW be the thermal conductivity at (x,y) in the east direction. x,y Let be the thermal conductivity at (x,y) in the west direction. Let I be the divergence of pixel I in the north direction after t iterations. Let I be the divergence of pixel I in the south direction after t iterations. Let I be the divergence of pixel I in the east direction after t iterations. Let I be the divergence of pixel I in the west direction after t iterations. After processing all pixels according to the formula, the denoised iris image can be obtained.

[0029] Optionally, the specific steps for using integral projection to coarsely locate the pupil center in the denoised iris image, determine a rough pupil center, and then binarize the denoised iris image using an adaptive thresholding method based on the center are as follows:

[0030] First, the horizontal and vertical integral projections of the grayscale image are used to locate a point within the pupil region. The formula for calculating this point is:

[0031]

[0032] In the formula, roughx represents the x-axis coordinate of the positioning point, roughy represents the y-axis coordinate of the positioning point, m is the height of the image, n is the width of the image, m1 = 0.25m, m2 = 0.75m, n1 = 0.4n, n2 = 0.6n, I(:,y) represents the gray level of all pixels with equal y-coordinates, and I(x,:) represents the gray level of all pixels with equal x-coordinates. After locating the point, the adaptive binarization threshold can be calculated using the following formula:

[0033]

[0034] thresh represents the binarization threshold, roughx represents the x-axis coordinate of the positioning point, roughy represents the y-axis coordinate of the positioning point, a is the defined search range offset value, and I(x,y) is the pixel intensity at (x,y). After calculating the adaptive binarization threshold using the above formula, the image can be binarized using this value to obtain an adaptive binary image.

[0035] Optionally, the steps of finding and locating the precise pupil center coordinates using the adaptive binary image, and calculating the pupil radius based on the center to obtain the pupil outline circle, include:

[0036] The adaptive binary image now only contains the pupil region and scattered noise. Opening operations are used to fill these noisy openings. Then, the area of ​​the remaining connected regions is calculated; the largest connected region corresponds to the pupil region. Precise center location is then performed based on the pupil region. The specific formula for calculating the center position is as follows:

[0037]

[0038] In the formula, m is the image height, n is the image width, and BW(x,y) is the binarized image at (x,y), calculated from x. o and y o The coordinates of the center o are used as a reference. The maximum distance between the pupil center and a non-zero point is searched every 15 degrees along the circumference of the pupil region. The average value of these maximum distances is then calculated, which is the pupil radius. After obtaining the pupil radius and the center, the pupil outline can be obtained.

[0039] Optionally, the steps of iris outer contour extraction and outer contour circle fitting include:

[0040] Based on the ratio of the pupil circle radius to the iris circle radius, the search range of the iris outer contour is limited. The iris outer contour is obtained by using an edge detection algorithm. The iris outer contour is then used for circle fitting to obtain an accurate iris outer contour circle.

[0041] Optionally, the steps of using the Viterbi algorithm for edge detection to search for iris edges include:

[0042] Dynamic programming is used to search for the point with the maximum gradient of gray values ​​in all angular directions within the range of 0° to 360°. The algorithm then detects the shortest path accumulated between these gradient points and forcibly closes the path. The calculation formula is as follows:

[0043]

[0044] In the formula, k represents the time scale, λ is the gradient search weight, and when k = K, Γ(x) k Γ(x0) = 0, which is the globally optimal complete contour found. max )=∞,max≠0,(x k+1 ,x k ) represents a path found between time k+1 and time k, ε k Indicates in (x k+1 ,x k The algorithm searches for the maximum gradient within the path range. Once a matching maximum is found, the corresponding path Γ is saved. All paths are then calculated through time iteration. After the iteration ends, the minimum matching value of the path is output, which is Γ(x). kThis is the globally optimal path solution.

[0045] Optionally, the specific formula for performing circle fitting on the searched contour using the least squares method is as follows:

[0046]

[0047]

[0048] In the formula, x i Let y be the x-coordinate of the i-th iris contour data point. i Let x0 be the ordinate of the i-th iris contour data point, x0 and y0 be the x and y coordinates of the center of the fitting circle, and R represent the radius of the fitting circle. A, B, and C are intermediate variables for calculating x0, y0, and R. After obtaining the radius and center position of the outer contour circle of the iris, the iris image can be segmented from the pupil contour.

[0049] Compared with the prior art, the significant advantages of this invention are:

[0050] (1) The present invention uses an adaptive threshold method to locate the pupil and eliminate noise, which can adapt to images in various iris databases and improve the accuracy of iris localization;

[0051] (2) This invention reduces the computational load of edge detection by limiting the search range of the outer contour of the iris, which greatly speeds up the calculation of the outer contour of the iris.

[0052] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0053] Figure 1 This is a grayscale image of the iris according to an embodiment of the present invention.

[0054] Figure 2 This is a smoothing filter noise reduction diagram according to an embodiment of the present invention.

[0055] Figure 3 This is an adaptive binary image according to an embodiment of the present invention.

[0056] Figure 4 This is a diagram of the opening of the filling spot in an embodiment of the present invention.

[0057] Figure 5 Image showing the iris localization result.

[0058] Figure 6 This is a flowchart of the present invention. Detailed Implementation

[0059] The following describes the embodiments of the present invention through specific examples and in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. In one embodiment, an iris image localization method, such as... Figure 6 As shown, it includes the following steps:

[0060] Step 1: Obtain an image of the iris and convert it to a grayscale image. The specific method is as follows:

[0061] Obtain each pixel n of the iris image of the eye. i,j The RGB intensity value of each pixel n i,j The RGB values ​​are converted to grayscale values ​​using a conversion formula to obtain the grayscale image of the iris;

[0062] In this embodiment, the formula for converting RGB values ​​to grayscale values ​​is:

[0063] Gray(n i,j )=0.212671×R(n i,j )+0.715160×G(n i,j )+0.072169×B(n i,j )

[0064] In the formula, Gray(n) i,j ) represents the pixel n in the grayscale image of the iris. i,j The grayscale value, i represents the row number of the pixel, j represents the column number of the pixel, R(n i,j ) represents the number of pixels n i,j The red intensity value, G(n) i,j ) represents the number of pixels n i,j The green intensity value, B(n) i,j ) represents the number of pixels n i,j The blue intensity value, the conversion result is as follows Figure 1 As shown.

[0065] Step 2: Smooth the image to remove noise such as eyelashes and eyelids to obtain a denoised iris image. Binarize the denoised iris image, filling in internal holes such as reflected light spots to obtain an adaptive binary image. The specific method is as follows:

[0066] Anisotropic smoothing is used to filter and reduce noise in the image. When processing pixels, the divergence of the pixel in four directions is first obtained, and the calculation formula is as follows:

[0067]

[0068] In the formula, I x,y I represents the intensity of the pixel at (x,y). x,y-1This represents the intensity of the pixel at (x, y-1). This represents the divergence of pixel I in the north direction at (x,y). Let I represent the divergence of pixel I at (x,y) in the south direction. Let represent the divergence of pixel I at (x,y) in the east direction. This represents the divergence of pixel I in the west direction at (x,y);

[0069] The thermal conductivity of a pixel in four directions is calculated using the following formula:

[0070]

[0071] In the formula, cN x,y Let cS be the thermal conductivity at (x,y) in the north direction. x,y Let E be the thermal conductivity at (x,y) in the south direction. x,y Let cW be the thermal conductivity at (x,y) in the east direction. x,y Let be the thermal conductivity at (x,y) in the west direction. Let I be the divergence in the north direction of pixel I. Let I be the divergence in the south direction. Let I be the divergence in the east direction of pixel I. Let be the divergence of pixel I in the west direction, and k be the weighting coefficient;

[0072] Combining the two sets of equations above, the iterative formula for anisotropic smoothing filtering can be obtained as follows:

[0073]

[0074] In the formula I t+1 This represents the pixel intensity after iteration t+1, where t is the iteration number, λ is the smoothing control, and cN is the pixel intensity. x,y Let cS be the thermal conductivity at (x,y) in the north direction. x,y Let E be the thermal conductivity at (x,y) in the south direction. x,y Let cW be the thermal conductivity at (x,y) in the east direction. x,y Let be the thermal conductivity at (x,y) in the west direction. Let I be the divergence of pixel I in the north direction after t iterations. Let I be the divergence of pixel I in the south direction after t iterations. Let I be the divergence of pixel I in the east direction after t iterations. Let I be the divergence of pixel I in the west direction after t iterations. The denoised iris image can be obtained by processing all pixels according to the formula, as shown below. Figure 2 As shown;

[0075] First, the horizontal and vertical integral projections of the grayscale image are used to locate a point within the pupil region. The formula for calculating this point is:

[0076]

[0077] In the formula, roughx represents the x-axis coordinate of the positioning point, roughy represents the y-axis coordinate of the positioning point, m is the height of the image, n is the width of the image, m1 = 0.25m, m2 = 0.75m, n1 = 0.4n, n2 = 0.6n, I(:,y) represents the gray level of all pixels with equal y coordinates, and I(x,:) represents the gray level of all pixels with equal x coordinates.

[0078]

[0079] thresh represents the binarization threshold, roughx represents the x-axis coordinate of the localization point, roughy represents the y-axis coordinate of the localization point, 'a' is the defined search range offset value, and I(x,y) is the pixel intensity at (x,y). An adaptive binary map is calculated, and the result is as follows: Figure 3 As shown;

[0080] Step 3: Find the center of the circle in the adaptive binary image, locate the precise coordinates of the pupil center, and calculate the pupil radius based on the center to obtain the pupil outline circle;

[0081] The adaptive binary image now only contains the pupil region and scattered noise. An opening operation is used to fill these noisy openings. The result of the opening operation is as follows: Figure 4 As shown, the area of ​​the remaining connected regions is then calculated. The largest connected region corresponds to the pupil region. Based on the pupil region, the precise center of the circle is located. The specific formula for calculating the center position is as follows:

[0082]

[0083] In the formula, m is the image height, n is the image width, and BW(x,y) is the binarized image at (x,y), calculated from x. o and y o These are the coordinates of the center o of the circle;

[0084] Using the pupil center as a reference, search for the maximum distance between the pupil center and a non-zero point position every 15 degrees along the circumference of the pupil region. Then calculate the average value of these maximum distances, which is the pupil radius. After obtaining the pupil radius and the center of the circle, the pupil outline can be obtained.

[0085] Step 4: The steps for iris outer contour extraction and outer contour circle fitting include:

[0086] Based on the ratio of the pupil radius to the iris radius, the search range for the outer contour of the iris is defined as follows:

[0087] R out >1.5×R in &&R out <3.3×R in

[0088] R out R represents the radius of the outer contour of the iris. in This represents the radius of the pupil outline. Within this range, the Viterbi algorithm is used for edge detection to search for the iris edge. The calculation formula is as follows:

[0089]

[0090] In the formula, k represents the time scale, λ is the gradient search weight, and when k = K, Γ(x) k Γ(x0) = 0, which is the globally optimal complete contour found. max )=∞,max≠0,(x k+1 ,x k ) represents a path found between time k+1 and time k, ε k Indicates in (x k+1 ,x k The algorithm searches for the maximum gradient within the path range. Once a matching maximum is found, the corresponding path Γ is saved. All paths are then calculated through time iteration. After the iteration ends, the minimum matching value of the path is output, which is Γ(x). k This is the globally optimal path solution;

[0091] The specific formula for performing circle fitting on the searched contour using the least squares method is as follows:

[0092]

[0093]

[0094] In the formula, x i Let y be the x-coordinate of the i-th iris contour data point. i Let x0 be the ordinate of the i-th iris contour data point, x0 and y0 be the x and y coordinates of the center of the fitted circle, and R represent the radius of the fitted circle. A, B, and C are intermediate variables for calculating x0, y0, and R. The results are as follows: Figure 5 As shown.

[0095] This invention uses an automated thresholding method to separate the pupil and reflected light spot in the iris image, and is adaptive to various iris databases, which can improve the accuracy of iris localization.

[0096] This invention reduces the computational load of edge detection by limiting the search range of the iris outer contour, thus significantly accelerating the iris outer contour search speed.

Claims

1. An iris image localization method, characterized in that, Includes the following steps: Step 1: Obtain an image of the iris of the eye and convert it into a grayscale image of the iris; Step 2: Smooth the grayscale image of the iris to remove noise, and obtain a denoised iris image. Then, binarize the denoised iris image and fill the inner hole to obtain an adaptive binary image. Step 3: Locate the precise pupil center coordinates using the adaptive binary image, and calculate the pupil radius based on the center to obtain the pupil outline circle; Step 4: Extract the outer contour of the iris image based on the pupil center, and obtain a precise outer contour circle by fitting the outer contour. Separate the iris using the precise pupil contour circle and the precise outer contour circle. The steps of iris outer contour extraction and outer contour circle fitting include: Based on the ratio of the pupil circle radius to the iris circle radius, the search range of the iris outer contour is limited. The iris outer contour is obtained by using an edge detection algorithm. The iris outer contour is then used for circle fitting to obtain an accurate iris outer contour circle. The steps for obtaining the outer contour of the iris using an edge detection algorithm include: The Viterbi algorithm searches for the points with the largest gradient gray values ​​in all angular directions within the range of 0° to 360°. It then uses dynamic programming to detect the shortest paths accumulated between these gradient points and forcibly closes the paths. The Viterbi algorithm's calculation formula is as follows: In the formula, k represents the time scale, and λ is the gradient search weight. When k=K, It is the globally optimal complete contour found through the search. , max≠0 This represents a path found between time k+1 and time k. Indicates in The maximum gradient value is searched within the path range. Once a matching maximum value is found, the corresponding path is saved. Furthermore, it calculates all paths through time iteration, and outputs the minimum coincidence value of the paths after the iteration is complete. This is the globally optimal path solution.

2. The iris image localization method according to claim 1, characterized in that, The steps to convert an image of the iris of the eye into a grayscale image of the iris include: Obtain each pixel n of the iris image of the eye. i,j The RGB intensity value of each pixel n i,j The RGB values ​​are converted to grayscale values ​​using a conversion formula to obtain the grayscale image of the iris; The formula for converting RGB values ​​to grayscale values ​​is: In the formula, For the pixel n in the grayscale image of the iris i,j The grayscale value, where i represents the row number of the pixel and j represents the column number of the pixel. For pixel n i,j The red intensity value, For pixel n i,j The green intensity value, For pixel n i,j The blue intensity value.

3. The iris image localization method according to claim 1, characterized in that, The steps of smoothing and removing noise from the iris grayscale image to obtain a denoised iris image, and filling the inner holes of the denoised iris image to obtain an adaptive binary image include: Anisotropic smoothing filtering is applied to the grayscale image of the iris to remove eyelid and eyelash noise, resulting in a denoised iris image. Integral projection is used to locate the pupil region in the denoised iris image, and a point within the pupil is determined and set as the coarse center. Based on the coarse center, an adaptive thresholding method is used to binarize the denoised iris image and separate the pupil.

4. The iris image localization method according to claim 3, characterized in that, The specific steps of anisotropic smoothing filtering are as follows: The divergence of a pixel in four directions is calculated using the following formula: In the formula, This represents the intensity of the pixel at (x,y). This represents the intensity of the pixel at (x, y-1). This represents the divergence of pixel I in the north direction at (x,y). Let I represent the divergence of pixel I at (x,y) in the south direction. Let represent the divergence of pixel I at (x,y) in the east direction. This represents the divergence of pixel I in the west direction at (x,y); The thermal conductivity of a pixel in four directions is calculated using the following formula: In the formula, Let be the thermal conductivity at (x,y) in the north direction. Let be the thermal conductivity at (x,y) in the south direction. Let be the thermal conductivity at (x,y) in the east direction. Let be the thermal conductivity at (x,y) in the west direction. Let I be the divergence in the north direction of pixel I. Let I be the divergence in the south direction. Let I be the divergence in the east direction of pixel I. Let be the divergence of pixel I in the west direction, and k be the weighting coefficient; Combining the two sets of equations above, the iterative formula for anisotropic smoothing filtering is obtained as follows: In the formula This represents the pixel intensity after iteration t+1, where t is the iteration number and λ is the smoothing control. Let be the thermal conductivity at (x,y) in the north direction. Let be the thermal conductivity at (x,y) in the south direction. Let be the thermal conductivity at (x,y) in the east direction. Let be the thermal conductivity at (x,y) in the west direction. Let I be the divergence of pixel I in the north direction after t iterations. Let I be the divergence of pixel I in the south direction after t iterations. Let I be the divergence of pixel I in the east direction after t iterations. Let I be the divergence of pixel I in the west direction after t iterations. After processing all pixels according to the formula, the denoised iris image can be obtained.

5. The iris image localization method according to claim 3, characterized in that, The specific steps for locating the pupil region in the denoised iris image using integral projection are as follows: The point within the pupil region is located using the horizontal and vertical integral projections of the grayscale image. The formula for calculating this point is: In the formula, This represents the x-axis coordinate of the positioning point. The y-coordinates of the positioning points are represented by m, the height of the image is m, and the width of the image is n. m1 = 0.25m, m2 = 0.75m, n1 = 0.4n, n2 = 0.6n. This represents the grayscale value of all pixels with the same y-coordinate. This represents the grayscale value of all pixels with the same x-coordinate. The adaptive binarization threshold is calculated using the following formula: Indicates the binarization threshold. This represents the x-axis coordinate of the positioning point. This represents the y-coordinate of the positioning point, and 'a' is the defined search range offset value. Let (x,y) be the pixel intensity at (x,y). The image is binarized using an adaptive binarization threshold to obtain an adaptive binary image.

6. The iris image localization method according to claim 1, characterized in that, The steps of finding and locating the precise pupil center coordinates using the adaptive binary image, and calculating the pupil radius based on the center to obtain the pupil outline circle include: Use opening operations to fill noise openings; Calculate the area of ​​the remaining connected regions. The largest connected region corresponds to the pupil region. Based on the pupil region, accurately locate the center of the circle. The specific formula for calculating the center position is as follows: In the formula, m is the image height and n is the image width. For the binarized image at (x,y), calculate and These are the coordinates of the center o of the circle; Using the center of the pupil as a reference, search for the maximum distance between the center of the pupil and a non-zero point position every 15 degrees along the circumference of the pupil region; Find the average of all the maximum distances. This average is the pupil radius. Once you have the pupil radius and the center of the circle, you can get the pupil outline.

7. The iris image localization method according to claim 1, characterized in that, The specific formula for performing circle fitting on the extracted iris outer contour using the least squares method is as follows: In the formula, x i Let y be the x-coordinate of the i-th iris contour data point. i Let x0 be the ordinate of the i-th iris contour data point, x0 and y0 be the x and y coordinates of the center of the fitting circle, R represent the radius of the fitting circle, and A, B, and C are intermediate variables for calculating x0, y0, and R.

Images