Methods and systems for quality assessment of iris images

Abstract

This invention relates to a method and system for quality assessment of iris images. The method includes: determining the pupil center and pupil radius in the iris image to be assessed; determining the radius of a fitted circle corresponding to the outer contour of the iris in the iris image to be assessed based on the pupil center and pupil radius, and using this radius as the iris radius; calculating a first score value; calculating a second score value based on the fitted circle; calculating a third score value based on each pixel in the iris image to be assessed; and calculating a comprehensive score value for the iris image to be assessed; and assessing the quality of the iris image to be assessed based on the first score value, the second score value, the third score value, and the comprehensive score value. This invention allows for rapid assessment of the quality of acquired iris images, eliminating low-quality images and ensuring that only high-quality iris images are entered into the database.

Description

Technical Field

[0001] This invention relates to the field of iris technology, and in particular to a method and system for quality assessment of iris images. Background Technology

[0002] Since the beginning of the 21st century, with social development and progress, the demand for security has gradually increased. Iris recognition, as a highly efficient and secure biometric technology, offers advantages over other biometric technologies such as facial recognition, fingerprint recognition, and voice recognition. It boasts characteristics such as accuracy, speed, ease of operation, high reliability, and non-contact detection. Therefore, iris recognition technology has broad prospects in the fields of biometrics and security. Furthermore, the human iris contains the richest textural information, composed of numerous crypts, folds, and pigment spots, making it the most unique structure in the human body. The formation of the iris is determined by genetics. Around eight months after birth, the iris has essentially developed to a sufficient size and entered a relatively stable period. Except for special circumstances caused by external eye damage, the textural structure of the iris hardly changes. Therefore, iris recognition, as a biometric method, is particularly suitable for building large iris feature databases. Other biometric databases, such as facial and fingerprint databases, require regular updates to individual biometric data. Because of the stable and unchanging structure of the human iris texture, the iris feature database hardly needs to be updated after each person completes the iris feature registration and can be used for life, which greatly reduces the maintenance cost of large feature databases.

[0003] However, in the current process of establishing large iris feature databases (such as national and provincial iris databases), inconsistencies in the quality of the final iris images can occur due to issues such as operator errors, different acquisition equipment, inconsistent lighting conditions, and the subject's closed eyes or shaking during the acquisition process. Therefore, it is necessary to provide an algorithm for quickly evaluating the quality of acquired iris images, eliminating low-quality images, and ensuring that only high-quality iris images are entered into the database. Summary of the Invention

[0004] To solve the above-mentioned technical problems, or at least partially solve them, the present invention provides a method and system for quality assessment of iris image acquisition.

[0005] In a first aspect, the present invention provides a method for quality assessment of iris image acquisition, comprising:

[0006] Determine the pupil center and pupil radius in the iris image to be evaluated;

[0007] Based on the pupil center and the pupil radius, determine the radius of the fitted circle corresponding to the outer contour of the iris in the iris image to be evaluated, and use this radius as the iris radius;

[0008] A first score is calculated based on the pupil radius and the iris radius. The first score is used to evaluate the effective area size of the iris in the iris image to be evaluated.

[0009] Based on the fitted circle, a second score value is calculated, which is used to evaluate the size of the occluded area of ​​the iris in the iris image to be evaluated.

[0010] A third score is calculated based on each pixel in the iris image to be evaluated. The third score is used to evaluate the clarity of the iris image to be evaluated.

[0011] The comprehensive score of the iris image to be evaluated is calculated based on the first score, the second score, and the third score.

[0012] The quality of the iris image to be evaluated is assessed based on the first score, the second score, the third score, and the comprehensive score.

[0013] Secondly, the present invention provides a quality assessment system for iris image acquisition, comprising:

[0014] The first determining module is used to determine the pupil center and pupil radius in the iris image to be evaluated;

[0015] The second determining module is used to determine the radius of the fitted circle corresponding to the outer contour of the iris in the iris image to be evaluated based on the pupil center and the pupil radius, and to use this radius as the iris radius.

[0016] The first calculation module is used to calculate a first score value based on the pupil radius and the iris radius. The first score value is used to evaluate the effective area size of the iris in the iris image to be evaluated.

[0017] The second calculation module is used to calculate a second score value based on the fitted circle. The second score value is used to evaluate the size of the occluded area of ​​the iris in the iris image to be evaluated.

[0018] The third calculation module calculates a third score value based on each pixel in the iris image to be evaluated. The third score value is used to evaluate the clarity of the iris image to be evaluated.

[0019] The comprehensive calculation module is used to calculate the comprehensive score of the iris image to be evaluated based on the first score, the second score, and the third score.

[0020] The image evaluation module is used to evaluate the quality of the iris image to be evaluated based on the first score, the second score, the third score, and the comprehensive score.

[0021] The iris image quality assessment method and system provided in this embodiment first determines the pupil center and pupil radius in the iris image to be assessed, and then determines the iris radius. Based on the pupil radius and iris radius, three score values ​​are calculated. A comprehensive score is then calculated based on these three score values. The iris image to be assessed is evaluated from three different dimensions and the comprehensive score to determine whether the iris image is qualified. This invention can quickly assess the quality of acquired iris images, eliminate low-quality images, and ensure that only high-quality iris images are entered into the database. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the method for quality assessment of iris image acquisition in an embodiment of the present invention;

[0025] Figure 2a This is a schematic diagram of an iris image to be evaluated in an embodiment of the present invention;

[0026] Figure 2b For the purposes of this embodiment of the invention Figure 2a The binarized image obtained after binarization processing;

[0027] Figure 2c For the purposes of this embodiment of the invention Figure 2a A schematic diagram of the iris image to be evaluated after light spot filling;

[0028] Figure 3a This is a schematic diagram of the specific process of step S110 in an embodiment of the present invention;

[0029] Figure 3b This is a schematic diagram of a ring-shaped template in an embodiment of the present invention;

[0030] Figure 3c In this embodiment of the invention, the vector from the midpoint o of a shape to a point p on the circle is... gradient at point p A schematic diagram;

[0031] Figure 3dThis is a schematic diagram of a butterfly-shaped template in an embodiment of the present invention;

[0032] Figure 4a This is a ring-shaped region in an embodiment of the present invention;

[0033] Figure 4b For the purposes of this embodiment of the invention Figure 4a A schematic diagram of the rectangular unfolded image obtained after polar coordinate unfolding;

[0034] Figure 5 This is a schematic diagram of an iris image to be evaluated in an embodiment of the present invention, where part of the iris region is obscured;

[0035] Figure 6 This is a flowchart illustrating the method for quality assessment of iris image acquisition in an embodiment of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] In a first aspect, embodiments of the present invention provide a method for quality assessment of iris image acquisition, see [link to previous section]. Figure 1 and 6 The method includes the following steps S100 to S170:

[0038] S110. Determine the pupil center and pupil radius in the iris image to be evaluated;

[0039] In practice, to facilitate the determination of the pupil center and pupil radius, the light spot in the iris image to be evaluated can be filled before performing S110. The light spot area is the area formed by the reflection of the light source in the pupil.

[0040] In other words, the following steps may be performed before executing S110 in the embodiments of the present invention:

[0041] S100, Fill in the light spot area in the iris image to be evaluated.

[0042] In specific implementation, S100 may include the following steps:

[0043] S101. The iris image to be evaluated is binarized to obtain a binarized image, and a connected component containing only white pixels is found in the binarized image.

[0044] For example, targeting Figure 2a The iris image shown has a depth of 8. With a threshold set to 10, pixels with a value greater than or equal to 10 are assigned a value of 1, and pixels with a value less than 10 are assigned a value of 0. Figure 2b The binarized image is shown.

[0045] In practice, in a binary image, in the region where the pupil may appear (for example, in a 640*480 iris image, it is generally a square region with a side length of 200 centered at the midpoint of the image), seed filling or stacking is used to find and mark all connected regions of a preset size, such as 8-adjacent white connected regions. These connected regions are the possible locations of the light spot.

[0046] S102. Find the connected components that satisfy the first preset condition in each connected component, take the connected components that satisfy the first preset condition as the light spot region, and replace all white pixels in the light spot region of the iris image to be evaluated with black pixels to achieve the filling of the light spot region in the iris image to be evaluated.

[0047] The first preset condition includes that the number of black pixels among the surrounding pixels of the connected region is greater than a preset number.

[0048] It is understandable that the light spot is white and the surrounding area is black. Therefore, this characteristic can be used to filter out the light spot region among all connected components. That is, if there are enough black pixels around a connected component, then the connected component is considered to be the light spot region.

[0049] In practical implementation, the process of selecting the light spot area in S102 can specifically include:

[0050] S102a. Calculate the center point of each of the connected components;

[0051] For example, by averaging the x-coordinates of all points in a connected domain and then rounding them up, and by averaging the y-coordinates of all points and then rounding them up, we can obtain the x-coordinates and y-coordinates of the center point of this connected domain.

[0052] S102b: Select a corresponding preset region in each of the connected components in the binarized image. If the number of black pixels in the preset region is greater than the preset number, then the connected component corresponding to the preset region is a spot region. The preset region is a square region centered on the center point of the corresponding connected component and with a side length of a preset side length.

[0053] For example, for a connected component, take the center point of the connected component as the center, select a square region of 23 as the preset region, and count the number of black pixels in this square region in the binarized image. If the number of black pixels is greater than the threshold (e.g., 65), it means that this connected component is a spot region.

[0054] Understandably, when filling in light spots, the light spots in the original image are filled in; for example, after filling in the light spots, the image becomes... Figure 2c As shown in the image, when filling the light spot, you can also set the pixel value of all pixels in a circular area with a radius of 23 centered on the center of the light spot area in the iris image to be evaluated to 0, thereby filling the light spot area.

[0055] Understandably, the iris image to be evaluated can be processed in subsequent steps after the light spot filling is performed.

[0056] For specific implementation, see Figure 3a S110 may specifically include the following steps:

[0057] S1. Determine the range of values ​​for the pupil radius, and take the minimum value in the range as the initial radius;

[0058] S2. Generate a vertical and horizontal annular template corresponding to the current radius; perform two-dimensional convolution processing on the iris image to be evaluated that fills the light spot region with the Sobel operator in the vertical direction and the Sobel operator in the horizontal direction, respectively, to obtain a first gradient image in the vertical direction and a second gradient image in the horizontal direction; convolve the vertical annular template and the first gradient image to obtain a first image; convolve the horizontal annular template and the second gradient image to obtain a second image; sum the first image and the second image to obtain a third image;

[0059] For example, see a circular template. Figure 3b As shown, the square portion represents the template boundary. The template size must be larger than the maximum possible radius of the pupil; for example, in a 640*480 image, the template size is typically 130*130 to 150*150. A ring-shaped region with the template's midpoint as the center, a radius of r, and a thickness of 1 pixel is extracted. All other values ​​are set to 0. Depending on the values ​​of the ring-shaped region, ring-shaped templates can be divided into two types: vertical ring-shaped templates (fv) and horizontal ring-shaped templates (fh).

[0060] The gradient of each pixel in the vertical template fv is the cosine of the angle between the line connecting that pixel to the center of the circle and the vertical direction. The gradient value at each point is Y. op The calculation method is as follows:

[0061]

[0062] Among them, Y op Let the gradient value at this point be taken. Suppose the template size is D*D, and the coordinates of a point are (i, j). Then the x and y values ​​corresponding to this point are:

[0063] x = i - (D - 1) / 2

[0064] y = j - (D - 1) / 2

[0065] Similarly, in the horizontal circular template fh, the gradient of each pixel is the cosine of the angle between the line connecting that pixel to the center and the horizontal direction, and the gradient value at that point is X. op The calculation method is as follows:

[0066]

[0067] Among them, X op The gradient values ​​at this point, x, y, and the coordinates of this point are related as follows:

[0068] x = i - (D - 1) / 2

[0069] y = j - (D - 1) / 2

[0070] Among them, the Sobel operator Gy in the vertical direction:

[0071]

[0072] Among them, the horizontal Sobel operator Gx:

[0073]

[0074] The iris image Src, which fills the light spot region, is subjected to two-dimensional convolution with the Sobel operator Gy in the vertical direction and the Sobel operator Gx in the horizontal direction, respectively, to obtain the first gradient image gv in the vertical direction and the second gradient image gh in the horizontal direction.

[0075]

[0076]

[0077] The horizontal circular template fh is convolved with the second gradient image gh, and the vertical circular template fv is convolved with the first gradient image gv. The convolved images are then summed to obtain the third image gn.

[0078]

[0079] In the third image gn, the point with the largest pixel value can be considered the center of the most circular part of the image, with a radius of r. See also... Figure 3c When a shape approximates a circle, the vector from the midpoint o of the shape to a point p on the circle is... gradient with point p The smaller the included angle θ between the points, the larger cosθ becomes. The sum of the included angles cosθ at each point p on the annulus is:

[0080]

[0081] in, It is divided into the gradient coordinates on the y-axis and the gradient coordinates on the x-axis.

[0082]

[0083]

[0084] As can be seen, the value of each pixel in the third image gn obtained after convolution and summation corresponds to the sum of the angle cosθ between the gradient vector at all points on a circle with that point as the center and radius r and the center vector. Therefore, the pixel with the largest pixel value in the third image gn is the center of the shape that is closest to the circle. To make it more accurate, a butterfly model was also added.

[0085] S3. Generate a butterfly template corresponding to the current radius; perform two-dimensional convolution processing on the butterfly template and the iris image to be evaluated that fills the light spot area to obtain a fourth image;

[0086] In practice, see the structural diagram of the butterfly model. Figure 3d The size of the disc-shaped template should be larger than the maximum value of the pupil radius. For example, when using a 640*480 image, the template size is generally 130*130 to 150*150. A circle with radius r and center at the template's midpoint is extracted. Values ​​inside the circle are set to 1, and all other values ​​are set to 0.

[0087] The fourth image gm is obtained by performing a two-dimensional convolution between the disc template d and the iris image Src to be evaluated:

[0088]

[0089] Each point in the fourth image gm corresponds to the sum of gray values ​​within a circle centered at that point and with radius r. The point with the largest pixel value in the fourth image gm corresponds to the center of the darkest and roundest part, with radius r.

[0090] S4. Summing the third image and the fourth image yields the fifth image corresponding to the current radius;

[0091] For example, summing the third image gn and the fourth image gm yields the fifth image g:

[0092] g = gn + gm

[0093] S5. Find the maximum pixel value and the pixel corresponding to the maximum pixel value in the fifth image corresponding to the current radius, and take the maximum pixel value in the fifth image corresponding to the current radius as the local maximum pixel value corresponding to the current radius;

[0094] S6. Determine whether the current radius is the maximum value within the preset range; if yes, select the maximum pixel value from the local maximum pixel values ​​corresponding to each radius within the preset range, use the maximum pixel value as the overall maximum pixel value, and use the radius corresponding to the overall maximum pixel value as the pupil radius; otherwise, increment the current radius by 1 to update the current radius, and return to S2.

[0095] Understandably, in each iteration, the maximum pixel value in a fifth image g is taken as a local maximum pixel value. By traversing different radii r and finding the local maximum pixel values ​​corresponding to different r in the fifth image g, the maximum value among these local maximum pixel values ​​is selected as the overall maximum pixel value. The radius corresponding to the overall maximum pixel value is taken as the pupil radius, and the pixel corresponding to the overall maximum pixel value is taken as the pupil center.

[0096] As can be seen, in the above process, the iris image to be evaluated and the corresponding annular and butterfly templates of different radii are convolved in two dimensions by traversing the image to be evaluated, so as to find the darkest and roundest region, which is then used as the pupil region.

[0097] S120. Based on the pupil center and the pupil radius, determine the radius of the fitted circle corresponding to the outer contour of the iris in the iris image to be evaluated, and use this radius as the iris radius.

[0098] In specific implementation, S120 may include the following steps:

[0099] S121. Select an annular region in the iris image to be evaluated, and expand the annular region using polar coordinates to obtain a corresponding rectangular expanded image; wherein, the center of the annular region is the pupil center, the inner radius of the annular region is the pupil radius, and the outer radius of the annular region is the preset maximum radius of the iris; the horizontal coordinate of the rectangular expanded image is the angle, and the total coordinate is the radius;

[0100] For example, Figure 4a The annular region shown is expanded in polar coordinates to obtain... Figure 4bThe image shown is a rectangular unfolded image. The maximum radius of the iris can be set as needed; for example, it is typically 190° for a 640*480 image. The horizontal axis of the rectangular unfolded image is the angle θ, and the maximum value of the vertical axis is the difference between the maximum and minimum radii. The radius step size is 1, the angle step size is 360 / pi / pupil radius, and the pupil radius is r. P .

[0101] S122. Perform two-dimensional convolution processing on the preset Gaussian sharpening template and the rectangular unfolded image to obtain the corresponding sixth image;

[0102] For example, a Gaussian sharpening template and a rectangular unfolded image are used for two-dimensional convolution to achieve high-pass filtering, thereby sharpening the image and enhancing texture gradients. The Gaussian sharpening template used is shown below:

[0103]

[0104] S123. Generate a corresponding vertical gradient image based on the sixth image, denoted as the seventh image; the horizontal coordinate of the seventh image is the angle, the vertical coordinate is the radius, and the pixel value corresponding to each pixel point is the vertical gradient value under the corresponding angle and radius.

[0105] For the sixth image, the vertical gradient value G corresponding to each pixel in the vertical gradient image can be calculated as follows: n (r):

[0106]

[0107] n = 1, 2, ..., N

[0108] Among them, G n (r) represents the pixel (r, θ) in the seventh image. n The pixel value of G n (r) represents the vertical gradient value, I(r,θ) n ) to the pixel (r, θ) in the sixth image n The pixel value at ().

[0109] S124. Set the initial value of the column number n to 2, and perform the following steps: S01. Calculate the vertical gradient value of each pixel in the nth column of the seventh image and sum the maximum accumulated gradient value among the multiple accumulated gradient values ​​in the (n-1)th column to obtain the accumulated gradient value of each pixel in the nth column; S02. Increment n by 1 to update n; if the updated n is greater than N, then form an accumulated gradient image based on the accumulated gradient values ​​corresponding to each pixel in each column, which is denoted as the eighth image; if n is less than or equal to N, then return to S01; N is the total number of columns of the seventh image;

[0110] Wherein, the accumulated gradient value of each pixel in the first column is the vertical gradient value; the multiple accumulated gradient values ​​corresponding to each pixel in the first row and nth column include: the accumulated gradient value of the pixel in the first row and n-1th column and the accumulated gradient value of the pixel in the second row and n-1th column; the multiple accumulated gradient values ​​corresponding to each pixel in the last row and nth column include: the accumulated gradient value of the pixel in the last row and n-1th column and the accumulated gradient value of the pixel in the second-to-last row and n-1th column; the multiple accumulated gradient values ​​corresponding to each pixel in the rth row and nth column (excluding the pixels in the first row, last row, and first column) include the accumulated gradient values ​​of the pixels in the (n-1th column and r-1th row), the accumulated gradient values ​​of the pixels in the (n-1th column and rth row), and the accumulated gradient values ​​of the pixels in the (n-1th column and r+1th row).

[0111] As can be seen, S124 is a forward process from left to right, in which the accumulated gradient value is calculated by accumulating column by column. Each column corresponds to an angle θ, and each row corresponds to a radius r. For example, the accumulated gradient value of the pixel in the nth column and rth row can be calculated using the following method:

[0112]

[0113] As can be seen, the cumulative gradient value of each pixel in the eighth image is the sum of the vertical gradient value of that pixel and the [C] of the previous column. n-1 (r-1) C n-1 (r) C n-1 The summation is obtained by taking the maximum of the three values ​​[r+1]. For some special locations, such as pixels in the first row, last row, or first column, appropriate adjustments can be made. For example, the accumulated gradient value C for each pixel in the first column of the eighth image. n (r) is its vertical gradient value G n (r), the cumulative gradient value of the pixels in the first row can be expressed by the following formula:

[0114]

[0115] The accumulated gradient value of the pixels in the last row can be represented by the following formula:

[0116]

[0117] Using the method described above, the accumulated gradient value corresponding to each pixel in the eighth image can be obtained.

[0118] S125. Set the initial value of the column number n to N, and perform the following steps: S11. Select the optimal radius value in the nth column of the eighth image; S12. Select the optimal radius value corresponding to the (n-1)th column from the multiple radius values ​​corresponding to the optimal radius value in the (n-1)th column; S13. Decrement n by one to update the column number n. If n is not 0, return to S11. If n is 0, output the optimal radius value corresponding to each column in the eighth image and the angle corresponding to each column.

[0119] The optimal radius value in a column is the radius value corresponding to the pixel with the largest accumulated gradient value in that column; the optimal radius value in the first row and nth column corresponds to multiple radius values ​​in the (n-1)th column, including the optimal radius value in the nth column and the optimal radius value in the nth column + 1; the optimal radius value in the last row and nth column corresponds to multiple radius values ​​in the (n-1)th column, including the optimal radius value in the nth column and the optimal radius value in the nth column - 1; the optimal radius value in the nth column (excluding the first and last rows) corresponds to multiple radius values ​​in the (n-1)th column, including the optimal radius value in the nth column - 1, the optimal radius value in the nth column, and the optimal radius value in the nth column + 1.

[0120] In step S126, the optimal radius value in each column is selected by comparing values ​​column by column from right to left. Each column in the eighth image corresponds to an angle θ, and each row corresponds to a radius r. For example, the optimal radius value in the nth column is r'. n The optimal radius value r' corresponding to the (n-1)th column n-1 It can be expressed as follows:

[0121]

[0122] For some special positions, r∈[r'] n -1,r' n ,r' n +1] is used for deletion. For example, for the first row, r∈[r' n ,r' n +1], for the last row r∈[r' n -1,r' n ].

[0123] In the program implementation, due to the gradient accumulation matrix C n (r) is obtained by accumulating column by column from left to right, therefore the initial value of the backtracking process is the gradient accumulation matrix C. n (r) The radius R corresponding to the maximum value in the rightmost column, i.e.

[0124] r' N =arg max[C N (r)]

[0125] Then through r' N The value finds its maximum value within the following range:

[0126] [C n-1 (r' N -1),C n-1 (r' N ),C n-1 (r' N +1)]

[0127] The radius corresponding to the maximum value found is the optimal radius value r' in column N-1. N-1 The calculation proceeds column by column from right to left until the first column ends. In addition, to ensure the contours remain close during the backtracking process, the following constraints need to be added.

[0128] |r' n -r' N |≤n,n≤N

[0129] After the calculation, a series of corresponding radii r' will be obtained. n With angle θ n ,Right now:

[0130] (r' n ,θ n ), n=1,2,.....N

[0131] Among them, (r' n ,θ n ) represents the polar coordinates of the iris outline. By converting these coordinates back to rectangular coordinates, we can obtain the coordinates of the precise iris outline.

[0132] Understandably, the above process is the Viterbi algorithm. Using the Viterbi algorithm, the path with the highest gradient on the polar coordinate unfolded vertical gradient image is used to plan the boundary of the iris. That is, finding the optimal path in the polar coordinate unfolded gradient image through this algorithm is equivalent to finding a set of optimal radii and corresponding angles.

[0133] (R1,R2.........R n ...R N )

[0134] (θ1,θ2.........θ n ..........θ N )

[0135] S126. Transform the optimal radius value and the angle corresponding to each column into a rectangular coordinate system to obtain multiple coordinate points; perform circle fitting on the iris contour based on the multiple coordinate points to obtain the fitted circle corresponding to the outer contour of the iris; determine the iris radius and the iris center based on the fitted circle.

[0136] Understandably, after restoring each optimal radius value and each angle to a Cartesian coordinate system, a circle is fitted based on each coordinate point. The fitted circle is then used as the outer contour of the iris, and the iris center and iris radius are determined based on the fitted circle.

[0137] Understandably, circle fitting is a fitting method based on line segments and the completeness of the circle. For example, multiple coordinate points obtained after restoring the coordinate system are H. n (x n ,y n The coordinates are N, and the center coordinates and radius of the fitted circle are obtained by performing circle fitting using these coordinates.

[0138] The fitting steps are as follows:

[0139] 1. Calculate the average value m of the x-coordinates of all coordinate points. x The average value m of the vertical axis y .

[0140]

[0141]

[0142] 2. Calculate a series of variance values:

[0143]

[0144]

[0145]

[0146]

[0147]

[0148]

[0149]

[0150] 3. Calculate parameter x c With y c :

[0151] x c =0.5·[S xy ·(S yyy+S xxy )-S yy ·(S xxx +S xyy )] / (S xy ·S xy -S xx ·S yy )

[0152] y c =0.5·[S xy ·(S xxx +S xyy )-S xx ·(S yyy +S xxy )] / (S xy ·S xy -S xx ·S yy )

[0153] 4. Calculate the center c of the iris. I With radius r I :

[0154] c I (x c +m x y c +m y )

[0155]

[0156] Understandably, once the iris center and iris radius are calculated, the relevant score can be calculated based on these parameters.

[0157] S130. Calculate a first score value based on the pupil radius and the iris radius. The first score value is used to evaluate the effective area size of the iris in the iris image to be evaluated.

[0158] In practical implementation, the first formula can be used to calculate the first score value, and the first formula may include:

[0159]

[0160] In the formula, Score V For the first score value, r P Let r be the pupil radius. I Let be the radius of the iris.

[0161] As can be seen, the ratio of the first score to the pupil radius to the iris radius is such that the larger the ratio, the smaller the effective area of ​​the iris and the smaller the first score. Therefore, the lower the ratio, the higher the first score. The first score is between 0 and 100, and generally 36-42 is the passing score. If the score is lower than this threshold, the image is not acceptable.

[0162] Understandably, if the score is too low, you can increase the ambient light when taking the photo to reduce the size of the pupils.

[0163] S140. Calculate a second score value based on the fitted circle. The second score value is used to evaluate the size of the occluded area of ​​the iris in the iris image to be evaluated.

[0164] In practical implementation, the second formula can be used to calculate the second score value, and the second formula includes:

[0165]

[0166] In the formula, Score o S is the second score value. r S is the total number of pixels within the effective area of ​​the iris. v The total number of pixels in the fitted circle is denoted as , and the effective area is the area in the fitted circle excluding the occluded areas.

[0167] like Figure 5 As shown, the dark area at the edge of the fitted circle represents the area where the iris is obscured. The area of ​​the region can be represented by the number of pixels in the region. First, the number of pixels in the entire fitted circle is counted as the area s of the fitted annulus. v Then, the total number of pixels in the region after excluding the occluded area is counted as the area s of the actual effective region of the iris. r .

[0168] The second score ranges from 0 to 100, with 30-36 being the passing score. If the score is below this threshold, the iris image is considered unqualified.

[0169] In an iris image to be evaluated, eyelids, eyelashes, and other areas often obscure the iris. The lower the initial score, the more severe the iris obstruction.

[0170] S150. Calculate a third score value based on each pixel in the iris image to be evaluated. The third score value is used to evaluate the clarity of the iris image to be evaluated.

[0171] In specific implementation, S150 may include:

[0172] S151. A two-dimensional convolution process is performed using a preset Gaussian template and the iris image to be evaluated to obtain a ninth image. The pixel values ​​of all pixels in the ninth image are added together to obtain the total energy corresponding to the iris image to be evaluated.

[0173] For example, the following Gaussian template can be used:

[0174]

[0175] By performing a two-dimensional convolution with the original image using the aforementioned 8*8 Gaussian template, bandpass filtering of the source image is achieved, preserving the useful bands that clearly reflect iris texture. After convolution, the pixel values ​​of each pixel in the resulting ninth image are summed to obtain the total energy P of the image within the useful bands. Let the pixel value of each pixel after convolution be:

[0176] I(x n ,y m )

[0177] The total energy P is calculated as follows:

[0178]

[0179] S152. The third score value is calculated using the third formula, which includes:

[0180] Score D =P 2 / (P 2 +F 2 )·100

[0181] In the formula, Score D The third score value is P, the total energy is F, and the target energy preset value for 50% focus is F.

[0182] That is, the total energy P is mapped to a fraction from 0 to 100, and in a 640*480 iris image, F is generally taken as 2.4·10. 7 The general resolution threshold is set at 66-72; images below this threshold are considered unacceptable.

[0183] During iris image capture, low camera resolution, incorrect focusing, and camera shake can result in images with insufficient sharpness, making it impossible to locate the iris area and identify iris texture. This algorithm uses a two-dimensional convolution between the image and a bandpass template to obtain parameters that effectively reflect the image's texture sharpness, thus scoring the image's sharpness. The higher the score, the sharper the image.

[0184] S160. Calculate the comprehensive score of the iris image to be evaluated based on the first score, the second score, and the third score.

[0185] That is, the comprehensive score is obtained by weighted averaging the scores from the three dimensions:

[0186] Score=ω V Score V +ω o Score o +ω D Score D

[0187] Where ω V ,ω o ,ω D The weights for the first, second, and third scores are respectively set, and can be configured according to the actual situation, typically 0.3, 0.35, and 0.35 respectively. The overall score is usually set between 78 and 82 points; a total score below this threshold is considered unqualified.

[0188] S170. The quality of the iris image to be evaluated is assessed based on the first score, the second score, the third score, and the comprehensive score.

[0189] In specific implementation, S170 may specifically include: if at least one of the first score value, the second score value, the third score value and the comprehensive score value corresponding to the iris image to be evaluated is less than the corresponding lower limit value, then the iris image to be evaluated is unqualified.

[0190] As can be seen, a veto system is used for an iris image to be evaluated. If any score in any dimension is lower than its corresponding threshold, it is considered unqualified. The comprehensive score is calculated by weighted average of the scores in each dimension. If the scores in all three dimensions of the image exceed the threshold, but the comprehensive score is lower than the corresponding threshold, it is still considered unqualified.

[0191] The iris image quality assessment method and system provided in this embodiment first determines the pupil center and pupil radius in the iris image to be assessed, and then determines the iris radius. Based on the pupil radius and iris radius, three score values ​​are calculated. A comprehensive score is then calculated based on these three score values. The iris image to be assessed is evaluated from three different dimensions and the comprehensive score to determine whether the iris image is qualified. This invention can quickly assess the quality of acquired iris images, eliminate low-quality images, and ensure that only high-quality iris images are entered into the database.

[0192] Secondly, embodiments of the present invention provide a quality assessment system for iris image acquisition, comprising:

[0193] The first determining module is used to determine the pupil center and pupil radius in the iris image to be evaluated;

[0194] The second determining module is used to determine the radius of the fitted circle corresponding to the outer contour of the iris in the iris image to be evaluated based on the pupil center and the pupil radius, and to use this radius as the iris radius.

[0195] The first calculation module is used to calculate a first score value based on the pupil radius and the iris radius. The first score value is used to evaluate the effective area size of the iris in the iris image to be evaluated.

[0196] The second calculation module is used to calculate a second score value based on the fitted circle. The second score value is used to evaluate the size of the occluded area of ​​the iris in the iris image to be evaluated.

[0197] The third calculation module calculates a third score value based on each pixel in the iris image to be evaluated. The third score value is used to evaluate the clarity of the iris image to be evaluated.

[0198] The comprehensive calculation module is used to calculate the comprehensive score of the iris image to be evaluated based on the first score, the second score, and the third score.

[0199] The image evaluation module is used to evaluate the quality of the iris image to be evaluated based on the first score, the second score, the third score, and the comprehensive score.

[0200] It is understood that the system provided in this aspect corresponds to the method provided in the first aspect, and examples, implementation methods, beneficial effects, etc. can be referred to the corresponding parts of the first aspect, which will not be repeated here.

[0201] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0202] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0203] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as RON / RAN, magnetic disk, optical disk), and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0204] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.

Claims

1. A method for quality assessment of iris images, characterized in that, include: Determine the pupil center and pupil radius in the iris image to be evaluated; Based on the pupil center and the pupil radius, determine the radius of the fitted circle corresponding to the outer contour of the iris in the iris image to be evaluated, and use this radius as the iris radius; A first score is calculated based on the pupil radius and the iris radius. The first score is used to evaluate the effective area size of the iris in the iris image to be evaluated. Based on the fitted circle, a second score value is calculated, which is used to evaluate the size of the occluded area of ​​the iris in the iris image to be evaluated. A third score is calculated based on each pixel in the iris image to be evaluated. The third score is used to evaluate the clarity of the iris image to be evaluated. The comprehensive score of the iris image to be evaluated is calculated based on the first score, the second score, and the third score. The quality of the iris image to be evaluated is assessed based on the first score, the second score, the third score, and the comprehensive score. Determining the pupil center and pupil radius in the iris image to be evaluated includes: S1. Determine the range of values ​​for the pupil radius, and take the minimum value in the range as the initial radius; S2. Generate a vertical and horizontal annular template corresponding to the current radius; perform two-dimensional convolution processing on the iris image to be evaluated that fills the light spot region with the Sobel operator in the vertical direction and the Sobel operator in the horizontal direction, respectively, to obtain a first gradient image in the vertical direction and a second gradient image in the horizontal direction; convolve the vertical annular template and the first gradient image to obtain a first image; convolve the horizontal annular template and the second gradient image to obtain a second image; sum the first image and the second image to obtain a third image; S3. Generate a butterfly template corresponding to the current radius; perform two-dimensional convolution processing on the butterfly template and the iris image to be evaluated that fills the light spot area to obtain a fourth image; S4. Summing the third image and the fourth image yields the fifth image corresponding to the current radius; S5. Find the maximum pixel value and the pixel corresponding to the maximum pixel value in the fifth image corresponding to the current radius, and take the maximum pixel value in the fifth image corresponding to the current radius as the local maximum pixel value corresponding to the current radius; S6. Determine whether the current radius is the maximum value within the preset range; If so, the maximum pixel value is selected from the local maximum pixel values ​​corresponding to each radius within the preset range, and this maximum pixel value is taken as the overall maximum pixel value. The radius corresponding to the overall maximum pixel value is taken as the pupil radius, and the pixel corresponding to the overall maximum pixel value is taken as the pupil center. Otherwise, increment the current radius by 1 to update the current radius, and return to S2.

2. The method according to claim 1, characterized in that, Before determining the pupil center and pupil radius in the iris image to be evaluated, the method further includes: filling the light spot area in the iris image to be evaluated; The step of filling the light spot region in the iris image to be evaluated includes: The iris image to be evaluated is binarized to obtain a binarized image, and a connected component containing only white pixels is found in the binarized image. In each connected component, a connected component that satisfies a first preset condition is found. The connected component that satisfies the first preset condition is taken as the spot region. All white pixels in the spot region of the iris image to be evaluated are replaced with black pixels to fill the spot region of the iris image to be evaluated. The first preset condition includes that the number of black pixels in the surrounding pixels of the connected component is greater than a preset number.

3. The method according to claim 2, characterized in that, The step of searching for connected components that satisfy a first preset condition in each connected component, and using the connected components that satisfy the first preset condition as the light spot region, includes: Calculate the center point of each of the connected components; In each of the connected components in the binarized image, a corresponding preset region is selected. If the number of black pixels in the preset region is greater than the preset number, then the connected component corresponding to the preset region is the spot region. The preset region is a square region centered at the center point of the corresponding connected component and with a side length of a preset side length.

4. The method according to claim 1, characterized in that, The step of determining the radius of the fitted circle corresponding to the outer contour of the iris in the iris image to be evaluated based on the pupil center and the pupil radius, and using this radius as the iris radius, includes: A ring-shaped region is selected in the iris image to be evaluated, and the ring-shaped region is expanded in polar coordinates to obtain a corresponding rectangular expanded image; wherein, the center of the ring-shaped region is the center of the pupil, the inner radius of the ring-shaped region is the pupil radius, and the outer radius of the ring-shaped region is the preset maximum radius of the iris; the horizontal coordinate of the rectangular expanded image is the angle, and the total coordinate is the radius; The preset Gaussian sharpening template and the expanded rectangular image are subjected to two-dimensional convolution processing to obtain the corresponding sixth image; Based on the sixth image, a corresponding vertical gradient image is generated, denoted as the seventh image; the horizontal axis of the seventh image is the angle, the vertical axis is the radius, and the pixel value of each pixel point is the vertical gradient value under the corresponding angle and radius. The initial value of the column number n is set to 2, and the following steps are performed: S01, calculate the vertical gradient value of each pixel in the nth column of the seventh image and sum the maximum accumulated gradient value among the multiple accumulated gradient values ​​in the (n-1)th column to obtain the accumulated gradient value of each pixel in the nth column; S02, add 1 to n to update n; if the updated n is greater than N, then form an accumulated gradient image based on the accumulated gradient values ​​corresponding to each pixel in each column, denoted as the eighth image; if n is less than or equal to N, then return to S01; N is the total number of columns of the seventh image; where the accumulated gradient value of each pixel in the 1st column is the vertical gradient value; the accumulated gradient value of each pixel in the 1st row and nth column is the vertical gradient value; The plurality of accumulated gradient values ​​include: the accumulated gradient values ​​of the pixels in the first row and (n-1)th column and the accumulated gradient values ​​of the pixels in the second row and (n-1)th column; the plurality of accumulated gradient values ​​corresponding to each pixel in the last row and (n-1)th column include: the accumulated gradient values ​​of the pixels in the last row and (n-1)th column and the accumulated gradient values ​​of the pixels in the second-to-last row and (n-1)th column; the plurality of accumulated gradient values ​​corresponding to each pixel in the r-th row and (n-1)th column (excluding the pixels in the first row, the last row, and the first column) include: the accumulated gradient values ​​of the pixels in the (n-1)th column and (r-1)th row, the accumulated gradient values ​​of the pixels in the (n-1)th column and (r-1)th row; The initial value of the column number n is set to N, and the following steps are performed: S11, select the optimal radius value in the nth column of the eighth image; S12, select the optimal radius value corresponding to the (n-1)th column from the multiple radius values ​​corresponding to the optimal radius value in the nth column; S13, decrement n by one to update the column number n. If n is not 0, return to S11. If n is 0, output the optimal radius value corresponding to each column of the eighth image and the angle corresponding to each column. Among them, the optimal radius value in a column is the one with the largest accumulated gradient value in that column. The radius value corresponding to the pixel; the optimal radius value in the first row and nth column corresponds to multiple radius values ​​in the (n-1)th column, including the optimal radius value in the nth column and the optimal radius value in the nth column + 1; the optimal radius value in the last row and nth column corresponds to multiple radius values ​​in the (n-1)th column, including the optimal radius value in the nth column and the optimal radius value in the nth column - 1; the optimal radius value in the nth column (excluding the first and last rows) corresponds to multiple radius values ​​in the (n-1)th column, including the optimal radius value in the nth column - 1, the optimal radius value in the nth column, and the optimal radius value in the nth column + 1; The optimal radius value and the angle corresponding to each column are transformed into a rectangular coordinate system to obtain multiple coordinate points; the iris contour is fitted with a circle based on the multiple coordinate points to obtain the fitted circle corresponding to the outer contour of the iris; the iris radius and the iris center are determined based on the fitted circle.

5. The method according to claim 1, wherein The calculation of the first score value based on the pupil radius and the iris radius includes: The first score is calculated using a first formula, which includes: In the formula, Score V For the first score value, r P Let r be the pupil radius. I Let be the radius of the iris.

6. The method according to claim 1, characterized in that, The step of calculating the second score value based on the fitted circle includes: calculating the second score value using a second formula, wherein the second formula includes: In the formula, Score o For the second score, s r s is the total number of pixels within the effective area of ​​the iris. v The total number of pixels in the fitted circle is denoted as , and the effective area is the area in the fitted circle excluding the occluded areas.

7. The method according to claim 1, characterized in that, The step of calculating the third score value based on each pixel in the iris image to be evaluated includes: A second-dimensional convolution process is performed using a preset Gaussian template and the iris image to be evaluated to obtain a ninth image. The pixel values ​​of all pixels in the ninth image are summed to obtain the total energy corresponding to the iris image to be evaluated. The third score is calculated using a third formula, which includes: Score D =P 2 / (P 2 +F 2 )·100 In the formula, Score D The third score value is P, the total energy is F, and the target energy preset value for 50% focus is F.

8. The method according to claim 1, characterized in that, The process of evaluating the quality of the iris image to be evaluated based on the first score, the second score, the third score, and the comprehensive score includes: If at least one of the first score, the second score, the third score, and the comprehensive score corresponding to the iris image to be evaluated is less than the corresponding lower limit, then the iris image to be evaluated is unqualified.

9. A system for quality assessment of iris image acquisition using the method described in claim 1, characterized in that, include: The first determining module is used to determine the pupil center and pupil radius in the iris image to be evaluated; The second determining module is used to determine the radius of the fitted circle corresponding to the outer contour of the iris in the iris image to be evaluated based on the pupil center and the pupil radius, and to use this radius as the iris radius. The first calculation module is used to calculate a first score value based on the pupil radius and the iris radius. The first score value is used to evaluate the effective area size of the iris in the iris image to be evaluated. The second calculation module is used to calculate a second score value based on the fitted circle. The second score value is used to evaluate the size of the occluded area of ​​the iris in the iris image to be evaluated. The third calculation module calculates a third score value based on each pixel in the iris image to be evaluated. The third score value is used to evaluate the clarity of the iris image to be evaluated. The comprehensive calculation module is used to calculate the comprehensive score of the iris image to be evaluated based on the first score, the second score, and the third score. The image evaluation module is used to evaluate the quality of the iris image to be evaluated based on the first score, the second score, the third score, and the comprehensive score.

Images