CN109145748B - Vein line extraction algorithm of visible light vein imaging image - Google Patents
Vein line extraction algorithm of visible light vein imaging image Download PDFInfo
- Publication number
- CN109145748B CN109145748B CN201810810652.XA CN201810810652A CN109145748B CN 109145748 B CN109145748 B CN 109145748B CN 201810810652 A CN201810810652 A CN 201810810652A CN 109145748 B CN109145748 B CN 109145748B
- Authority
- CN
- China
- Prior art keywords
- vein
- point
- seed
- line
- tracking
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 210000003462 vein Anatomy 0.000 title claims abstract description 131
- 238000000605 extraction Methods 0.000 title claims abstract description 34
- 238000003384 imaging method Methods 0.000 title claims abstract description 22
- 238000001914 filtration Methods 0.000 claims abstract description 23
- 238000000034 method Methods 0.000 claims description 52
- 239000013598 vector Substances 0.000 claims description 27
- 230000004044 response Effects 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000005316 response function Methods 0.000 claims description 6
- 238000012804 iterative process Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 2
- 238000012217 deletion Methods 0.000 abstract description 3
- 230000037430 deletion Effects 0.000 abstract description 3
- 230000008447 perception Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 9
- 210000004204 blood vessel Anatomy 0.000 description 8
- 230000003044 adaptive effect Effects 0.000 description 7
- 238000012800 visualization Methods 0.000 description 7
- 239000008280 blood Substances 0.000 description 3
- 210000004369 blood Anatomy 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000000670 limiting effect Effects 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000002829 reductive effect Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 241001409201 Tabebuia chrysantha Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000002583 angiography Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000013060 biological fluid Substances 0.000 description 1
- 238000004883 computer application Methods 0.000 description 1
- 210000004351 coronary vessel Anatomy 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 210000000245 forearm Anatomy 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000003333 near-infrared imaging Methods 0.000 description 1
- 230000036961 partial effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000002207 retinal effect Effects 0.000 description 1
- 210000001210 retinal vessel Anatomy 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000002792 vascular Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/14—Vascular patterns
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a vein line extraction algorithm of a visible light vein imaging image, which belongs to the technical field of information perception and recognition, and the prior art has no vein extraction method specially aiming at the image, and particularly, the vein line extraction algorithm has better continuity and less noise by the steps of imaging image denoising and enhancement, Gabor filtering, tracking seed point determination, vein line tracking, Frangi filtering and vein line deletion; the invention provides a vein line extraction algorithm based on line tracking aiming at a visible light imaging image, and can obtain an ideal vein line.
Description
Technical Field
The invention belongs to the technical field of information perception and recognition, and particularly relates to a vein line extraction algorithm for a visible light vein imaging image.
Background
The biological characteristics have the advantages of difficult forgetting, good anti-counterfeiting performance, difficult theft, portability, availability at any time and any place and the like, so the biological characteristic identification technology for identifying the identity by utilizing the physiological and behavior characteristics of a person has universality, stability and uniqueness, can overcome the defects of the traditional authentication method, and becomes a popular research direction at home and abroad at present. Vein recognition based on near-infrared images is a novel biological feature recognition technology, wherein extraction of vein features is a very critical step, and the quality of feature extraction directly determines the final recognition effect. Compared with other characteristics, the vein line characteristics reflect the overall venation mode of the veins better, and have better discrimination.
In recent years, there are many researches on a near-infrared image vein extraction algorithm: miura et al (Miura N, Nagasaka A, Miyatake T. feature extraction of finger blood patterns based on predicted line tracking and its application to personal identification [ J ]. Machine Vision and Applications,2004,15(4):194 and 203.) propose a repeated line tracking method for the first time to obtain finger vein lines. Lajevrdi et al (Lajevrdi S M, Arakala A, Davis S, et al. hand in the authentication using biological mapping [ J ]. IET Biometrics,2014,3(4):302 and 313.) divide the hand back near infrared vein image by using a maximum curvature method, and finally obtain the refined vein line by a morphological method. Zhang et al (Hengyi Zhang, Chaoying Tang, Adams Kong, Noah Craft, "Matching vector Patterns from Color Images for formal interrogation", IEEE International Conference on biometry: Theory, Applications and Systems, pp.77-84, Washington d.c., USA, 2012) enhanced arm near-infrared Images using a limit contrast adaptive histogram equalization method, Gabor filtering the enhanced Images, and finally binarizing the Gabor Images using Otsu method to extract veins. Kangwenxiong and the like (Kangwenxiong, Li Hua Song, Deng Fei. vein gray scale image topological feature extraction method. Chinese science: information science 2011,41:324 one 337.) extract vein trunks by using a directional field extreme image, then obtain vein branches by a maximum neighborhood point tracking algorithm, and further extract all veins. The extraction of blood vessels in retinal and coronary angiographic images is similar to the extraction of veins. Yin et al (Yin Y, Adel M, Bourennane S.automatic segmentation and measurement of vascular in biological fluids using a physiological formulation [ J ]. Computational and chemical methods in a blood vessel, 2013,2013.) use Bayesian maximum posterior criteria to determine vessel boundary points and segment retinal vessels. And (B) extracting blood vessels from coronary artery angiography images by using a shortest path algorithm and a backtracking cumulative method [ J ] computer application research, 2016,33(2): 611-. Hoover et al (Hoover A D, Kouznetsova V, gold baum M. locating blood vessels in reliable images by blood with threshold value and combining the region characteristics of the matched filtered images J. IEEE Transactions on Medical imaging 2000,19(3):203 + 210.) extract blood vessels using a threshold value decreasing search algorithm.
The traditional vein identification has to rely on near infrared imaging equipment, the cost is high, the use is inconvenient, Tang et al (Chaoying Tang, Adams Wai Kin Kong, N.C. Unco-observing blood patterns from color skin images for the purpose of information analysis [ C ]. IEEE International Conference on Computer Vision and Pattern Recognition,2011: 665-. However, the visualization result obtained from the visible light image often contains a lot of noise, and compared with the near-infrared image, it is more difficult to accurately extract the vein line, and at present, there is no vein extraction method specially for the visualization image.
Disclosure of Invention
The invention discloses a vein line extraction algorithm of a visible light vein imaging image aiming at the problems in the prior art, and a vein extraction method specially aiming at the image in the prior art. The invention provides a vein line extraction algorithm based on line tracking aiming at a visible light imaging image, and an ideal vein line can be obtained.
The invention is realized by the following steps:
a vein line extraction algorithm of a visible light vein imaging image is characterized by comprising the following steps:
the method comprises the following steps: carrying out preliminary denoising and enhancement on the visible light imaging result; a large amount of noise exists in a visible light imaging result, and for the situation, a down-sampling method is used for preliminary denoising. Aiming at the problem of unclear vein edges, the vein edge processing method adopts guide filtering with retained edge characteristics to process the vein edges, and the algorithm can overcome the defects that the Gaussian filtering cannot distinguish background noise and edge pixels and easily damages edge structures. But the contrast of the result is reduced compared with the original image, and the preliminary denoising is carried out by using a down-sampling method; and enhancing the filtering result by adopting a contrast-limiting self-adaptive histogram equalization method.
Step two: extracting vein information in the preprocessed image by adopting a real part of a Gabor filter; the frequency and direction expression of the Gabor filter is similar to that of the human visual system, and the Gabor filter has good characteristics in the aspect of extracting local space and frequency domain information of a target, so that the Gabor filter is widely applied to many fields of image processing, such as texture extraction, edge detection and the like, and obtains ideal results.
Step three: determining tracking seed points, and solving the initial tracking direction of the seed points tracked by the vein lines by using local gradient information;
step four: tracking the vein line, wherein the tracking of the vein line is an iterative process, and after the seed points of the image are determined, tracking is performed twice in opposite directions from each seed point until the tracking is terminated when a stop condition is met;
step five: frangi filtering, namely processing the image by using a Frangi filter and a multi-scale linear filter, segmenting the vein line, and finally deleting the vein line according to the segmented vein line and the filtered image;
step six: and deleting the vein lines, and processing the Frangi filtered image by adopting multi-scale linear filtering.
Further, the first step specifically comprises: carrying out preliminary denoising by using a down-sampling method; and enhancing the filtering result by adopting a contrast-limiting self-adaptive histogram equalization method.
Further, the third step is specifically:
3.1, let the modulus and direction of the gradient of a pixel with coordinates (i, j) on the image be GijAnd thetaijSetting (i)0,j0) For a particular point to be tracked, point (i) has the largest projection value in the direction parallel to it due to the vector0,j0) M of (A)The maximum value is obtained by the sum of the projection of the gradient vectors of all the points in the N fields in the main gradient direction of the point;
3.2, point (i)0,j0) Theta for the direction angle of the main gradientgAnd expressing, the functional expression of the gradient projection sum in the neighborhood is as follows:
wherein G isijAnd thetaijRespectively the mode and direction of the gradient of pixel (i, j); m and N represent the size of the neighborhood; derivative the formula (1) and let F' (θ)g) When theta is equal to 0gThe expression of (a) is as follows:
3.3, the gradient vector in the image always points in the direction of the fastest change in gray scale. Thus, point (i)0,j0) The main gradient direction of (i) is perpendicular to the vessel direction at that point0,j0) Direction angle of vein
The estimation can be made using the following equation:
point (i)0,j0) The vein direction of (a) can be described by the following unit vector:
further, in the third step, a series of initial seed points of the image are determined, then the veins are tracked starting from the series of seed points, and finally vein lines are obtained by combining the tracking results of all the seed points. In Gabor-enhanced visible light visualization images, veins are often not connected together, which makes it difficult to track all veins starting from one seed point. Therefore, firstly, a series of seed points are determined, veins are tracked from the seed points, and finally vein lines are obtained by combining the tracking results of all the seed points.
Further, the step of determining the initial seed point of the image is as follows: firstly, dividing an image into a plurality of grid areas by a series of horizontal lines and vertical lines with equal intervals; then searching the gray maximum value points on the lines row by row and column by column along the grid lines, and taking the points as candidate initial seed points
Further, the noise in the image will cause the algorithm to detect a large number of wrong seed points, and in order to solve the problem, the seed points are screened by using two thresholds, namely a global threshold and a local threshold, and the steps are as follows:
firstly, obtaining a threshold value T by using an Otsu method, and then passing through the threshold value TgCalculating to obtain a global threshold T as T/2g(ii) a Next, a local threshold T is setlThe expression is as follows:
Tl＝μseed+ασseed (5)
wherein, museedAnd σseedRespectively representing the gray level mean value and the standard deviation in the neighborhood of the seed point; α is a tuning parameter; to obtain TgAnd TlThen, the initial seed points which are both larger than the two threshold values are kept, and the rest seed points are deleted.
Further, the seed points obtained by the above method are sometimes not on the vein central line, and even deviate from the central line a lot, which affects the final tracking result, so that the initial seed points need to be corrected, and the specific steps of correcting the initial seed points are as follows:
let p be the initial seed point and,
is a direction vector of p, pcIs the corrected seed point of pThe coordinates thereof satisfy the following conditions:
wherein, (x, y) is pixel coordinates; i is a pixel gray value; spIs a cross point p and is related to the direction vector
A collection of pixels in a vertical cross-section.
Further, the fourth step is specifically:
let P0Is a seed point whose initial direction vector
The candidate point P of the next position is obtained by the formula (4)1Comprises the following steps:
wherein, P0And P1Is the respective coordinate, s is the tracking step length;
4.2, obtaining a candidate point P1Then, the true center point P can be determined by equation (6)1', while P can be substituted1The tracking direction of' is adjusted to:
thus, the iterative process of vein centerline tracking can be represented by:
4.3, when the tracking process encounters a bifurcation point, some problems arise if the true center point is still obtained using equation (6). To avoid this problem as much as possible, the center point is selected in such a way that:
wherein ρ (x, y) is an angle between the tracking direction vector of the point and the previous point; seThe gray maximum value point is a set of the gray maximum value points on the point section;
4.4, vein center line l1When extended to D, will be in contact with l3Encounter, a condition of two lines co-existence is likely to occur, and1and l2Similar problems are encountered, and in order to prevent such problems, detection points that are less than a threshold distance from the detected set of centerline points are deleted
4.5, the stopping condition of the tracking algorithm is as follows:
1) the absolute value of the gray difference between the newly detected central point and the last central point is greater than a given threshold;
2) the newly detected gray value of the central point is smaller;
3) the included angle between the newly detected central point and the tracking direction of the last central point is greater than a given threshold value;
4) the distance between the center line being tracked and the already tracked center line is less than a given threshold.
Further, the fifth step is specifically:
and constructing a filter response function by using the eigenvalue of the Hessian matrix, wherein the response function of the Frangi filter is as follows:
wherein, sigma is the scale of the Gaussian function;
wherein λ1And λ2(|λ1|≥|λ2I) is the eigenvalue of the Hessian matrix, beta and c are the size parameters;
5.2, detecting veins by using a multi-scale filter, and adopting the maximum filter response value under n scales as the final filter output:
further, the sixth step is specifically:
the Frangi filter has weak response in a shadow area, but partial shadow is detected, and for the situation, the Frangi filtered image is processed by adopting multi-scale straight line filtering. Then, curve segment segmentation is carried out on the original vein line, and the vein curve segments meeting the following requirements are deleted:
in the formula, length (l)k) The length of the curve of the k-th line is shown,
representing the sum of the lengths of the segments of the k-th curve whose filter response is greater than a certain threshold, α is a constant between 0 and 1.
The beneficial effects of the invention and the prior art are as follows:
1) the invention provides a vein line extraction algorithm based on line tracking aiming at visible light imaging images, fills the blank of a vein extraction method specially aiming at the images in the prior art, and vein lines obtained by adopting the algorithm can obtain higher matching rate;
2) compared with the prior art, the algorithm provided by the invention has the advantages that the algorithm is compared with the method for extracting the central line, the method comprises a local self-adaptive threshold method, a repeated line tracking method and a method for equalizing by adopting a contrast-limited self-adaptive histogram, and more complete and continuous venous lines can be extracted by the method provided by the invention.
Drawings
FIG. 1 is a schematic diagram illustrating seed point calibration according to an embodiment of the present invention;
FIG. 2 is a diagram illustrating a selection result of a seed point according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a venous line tracking in an embodiment of the present invention;
FIG. 4 is a graph of venous line post-processing results;
FIG. 5 compares the method of the present invention with prior art vein line extraction results;
FIG. 6 is a CMC graph of the matching results of the method of the present invention and the prior art using various algorithms;
figure 7 the present invention is a venous blood vessel in an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and effects of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be noted that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Visualization image denoising and enhancement
A large amount of noise exists in a visible light imaging result, and for the situation, a down-sampling method is used for preliminary denoising. Aiming at the problem of unclear vein edges, the vein edge processing method adopts guide filtering with retained edge characteristics to process the vein edges, and the algorithm can overcome the defects that the Gaussian filtering cannot distinguish background noise and edge pixels and easily damages edge structures. However, the contrast of the result is reduced compared with the original image, and the filtering result is enhanced by adopting a contrast-limited adaptive histogram equalization method.
2, Gabor filtering
The invention utilizes the real parts of 16 Gabor filters with different directions and scales to extract vein information in the preprocessed image.
3, tracking seed point determination
And solving the initial tracking direction 0 of the vein line tracking seed point by using the local gradient information. Let the modulus and direction of the gradient of a pixel with coordinates (i, j) on the image be GijAnd thetaij. Since the vector has the maximum projection value in the direction parallel thereto, point (i)0,j0) The sum of the projections of the gradient vectors of all points in the M x N domain at that point in the main gradient direction should be maximized. The M × N domain is a rectangular region, and the size of the neighborhood is set to 5 × 5 in this embodiment. If point (i)0,j0) Theta for the direction angle of the main gradientgAnd expressing, the functional expression of the gradient projection sum in the neighborhood is as follows:
derivative the above equation and let F' (theta)g) When theta is equal to 0gThe expression of (a) is as follows:
the gradient vectors in the image always point in the direction of the fastest change in gray scale. Thus, point (i)0,j0) The main gradient direction of (d) should be perpendicular to the vessel direction at that point. Point (i)0,j0) Direction angle gamma of the veini0j0The estimation can be made using the following equation:
finally, point (i)0,j0) The vein direction of (a) can be described by the following unit vector:
as shown in figure 7 of the drawings,
vessel direction vectors at k-1 iterations, vectors in the figure
Angle of direction of
(i0，j0) Is the current blood vessel center point, theta in the figureg60 °, the main gradient direction obtained by equation (2), and the current blood vessel direction
Vertically; the current blood vessel direction angle is obtained from the formula (3)
Point (i) according to equation (4)0,j0) The vein direction of (a) can be described by the following unit vector:
in Gabor-enhanced visible light visualization images, veins are often not connected together, which makes it difficult to track all veins starting from one seed point. Therefore, firstly, a series of seed points are determined, veins are tracked from the seed points, and finally vein lines are obtained by combining the tracking results of all the seed points. The initial seed point of the image is determined in three steps. Firstly, dividing an image into grid areas by a series of horizontal lines and vertical lines with equal intervals, then searching gray maximum value points on the lines row by row along the grid lines, and taking the points as candidate initial seed points. In the development of imagesThe algorithm detects a large number of wrong seed points, and in order to solve the problem, the seed points are screened by adopting two thresholds, namely a global threshold and a local threshold. Firstly, obtaining a threshold value T by using an Otsu method, and then passing through the threshold value TgCalculating to obtain a global threshold T as T/2gThen setting a local threshold TlThe expression is as follows:
Tl＝μseed+ασseed (5)
in the formula, museedAnd σseedRespectively representing the mean and standard deviation of the gray scale in the neighborhood of the seed point.
To obtain TgAnd TlThen, the initial seed points which are both larger than the two threshold values are kept, and the rest seed points are deleted.
The seed points obtained by the above method are sometimes not on the vein central line and may even deviate from the central line a lot, which affects the final tracking result, so that the seed points need to be corrected. As shown in fig. 1, p is the initial seed point,
is a direction vector of p, pcP is a seed point after correction, and the coordinates of the seed point satisfy the following conditions:
in the formula, SpIs a cross point p and is related to the direction vector
The position and the direction of the seed point obtained by screening the set of the pixel points in the vertical cross section by using the method are (x, y) pixel coordinates; i is a pixel gray value; spIs a cross point p and is related to the direction vector
A collection of pixels in a vertical cross-section.
As shown in fig. 2, fig. 2 shows the positions and directions of the seed points obtained by the screening method. 4, venous line tracking
The tracking of the vein line is an iterative process, after the seed points of the image are determined, tracking is performed twice in opposite directions starting from each seed point until the tracking is terminated when a stop condition is satisfied. Fig. 3 is a schematic diagram of a vein line tracking according to an embodiment of the present invention, as shown in fig. 3. FIG. 3(a) is a schematic diagram of seed point tracking, P0Is a seed point whose initial direction vector
The candidate point P of the next position is obtained by the formula (4)1Comprises the following steps:
in the formula, P0And P1S is the tracking step for the respective coordinate. Obtain a candidate point P1Then, the true center point P can be determined by equation (6)1', while P can be substituted1The tracking direction of' is adjusted to:
thus, the iterative process of vein centerline tracking can be represented by:
when the tracking process encounters a bifurcation point, some problems arise if the true center point is still obtained using equation (6). FIG. 3(b) is a schematic diagram showing the tracing of the venous bifurcation, when the center line l is located1When extending from point A to point B, if the point isC satisfies the condition of equation (6), then the tracking path will become A → B → C, which is clearly wrong and the correct center point should be D. To avoid this problem as much as possible, the center point is selected in such a way that:
where p (x, y) is the angle of the tracking direction vector of the point to the previous point, e.g.
SeIs the set of gray-scale maxima points on the point profile. As can be seen from FIG. 3(b), the vein centerline l1When extended to D, will be in contact with l3Encounter, a condition of two lines co-existence is likely to occur, and1and l2Similar problems are encountered in that, in order to prevent such problems from occurring, detection points that are less than a certain threshold distance from the detected set of centerline points are deleted.
The stopping conditions of the tracking algorithm are: (1) the absolute value of the gray difference between the newly detected central point and the previous central point is larger; (2) the newly detected gray value of the central point is smaller; (3) the included angle between the newly detected central point and the tracking direction of the previous central point is larger; (4) the distance between the centre point being tracked and the centre line already tracked is small.
5, Frangi filtering
The Gabor filtering may cause false detection of the visible light image, such as the venous line post-processing result diagram shown in fig. 4. Fig. 4(a) shows the visible light visualization image after preprocessing, the line shadow existing in the lower right area in fig. 4(a) is actually the arm contour, and as shown in the Gabor filtering energy diagram of fig. 4(b), the response value of the Gabor filter is large at this point and is very different from the filtering result at the real vein, which may cause the false extraction of the vein line, as shown in the original vein line diagram of fig. 4(c), the lowest line is actually the arm contour. The method utilizes the Frangi filter and the multi-scale linear filter to process the image, segments the vein line, and finally deletes the vein line according to the segmented vein line and the filtered image. Fig. 4(d) is a Frangi-filtered image.
The Frangi filter constructs a filter response function by using the eigenvalue of the Hessian matrix, and the response function of the Frangi filter is as follows:
wherein, sigma is the scale of the Gaussian function;
wherein λ1And λ2(|λ1|≥|λ2|) is the eigenvalue of the Hessian matrix. Beta and c are size parameters. Because the width sizes of veins in an image are inconsistent, in order to be able to estimate possible vein structures as accurately as possible, a multi-scale filter is used to detect veins, and the maximum filter response value at n scales is used as the final filter output:
fig. 4(e) is the result of multi-scale straight line filtering.
6, vein line deletion
In the case that the response of the Frangi filter in the shadow area is weak, but a part of the shadow is still detected, the Frangi filtered image is processed by adopting multi-scale straight line filtering (as shown in fig. 4 (d)), and the method can reserve a bright linear area and a dark part which is dark and has a bright area around the bright linear area in the image, inhibit the isolated dark linear area, and repair a part of broken veins in the filtered image. Then, curve segment segmentation is carried out on the original vein line, and the vein curve segments meeting the following requirements are deleted:
in the formula, length (l)k) The length of the curve of the k-th line is shown,
representing the sum of the lengths of the segments of the k-th curve whose filter response is greater than a certain threshold, α is a constant between 0 and 1. Fig. 4(f) is a vein line graph after deletion, and it can be seen that the erroneously extracted arm contour has been deleted.
Fig. 5 shows the comparison of the method of the present invention with the venous line extraction result of the prior art, and 5(a) and 5(b) respectively represent the visible light image and the visible light visualization image, and the comparison of the algorithm of the present invention with the line extraction method of the prior art, including the local adaptive threshold method, the repetitive line tracking method and the Otsu threshold method shown in fig. 5(c) to 5(e), results are shown in fig. 5. Through comparison, the following results can be found: although a part of vein lines can be extracted by using the local adaptive threshold method as shown in fig. 5(c), the final result contains a large number of wrong vein lines, which will have a serious influence on subsequent matching; the venous line obtained by the repeated line tracking method shown in fig. 5(d) has poor continuity and integrity, and contains much noise; and the venous line can be extracted completely and continuously by using the method and the Otsu threshold value method, for example, the extraction results of the algorithm of the invention shown in 5(e) and 5(f) are better in continuity and less in noise by using the method.
The embodiment also performs extraction and matching experiments on a forearm visible light image database of 150 persons, each person acquires two images, the angles and the illumination conditions of the two-time shooting are different, and the time interval is 10 days on average. And taking the image acquired in the first period as a reference image and the image acquired in the second period as an image to be matched. The local adaptive threshold and the repeated line are tracked respectively by adopting a vein line matching algorithm for limiting contrast adaptive histogram equalization, a matching experiment is carried out by adopting an extraction algorithm for limiting contrast adaptive histogram equalization and a vein line extracted by the algorithm of the invention, and a Cumulative Matching Characteristic (CMC) curve is generated, as shown in FIG. 6. It can be seen that the Rank-1 recognition rate of the algorithm is the highest. Therefore, compared with the other three methods, the vein line obtained by the algorithm can obtain higher matching rate.
Claims (8)
1. A vein line extraction algorithm of a visible light vein imaging image is characterized by comprising the following steps:
the method comprises the following steps: carrying out preliminary denoising and enhancement on the visible light imaging result;
step two: extracting vein information in the preprocessed image by adopting a real part of a Gabor filter;
step three: determining tracking seed points, and solving the initial tracking direction of the seed points tracked by the vein lines by using local gradient information; the third step is specifically as follows:
3.1, let the modulus and direction of the gradient of a pixel with coordinates (i, j) on the image be GijAnd thetaijSetting (i)0，j0) To prepare for a particular point to track, point (i)0，j0) The sum of the projections of the gradient vectors of all points in the M x N domain at the point in the main gradient direction should obtain the maximum value;
3.3, Point (i)0，j0) Theta for the direction angle of the main gradientgAnd expressing, the functional expression of the gradient projection sum in the neighborhood is as follows:
wherein G isijAnd thetaijRespectively the mode and direction of the gradient of pixel (i, j); m and N represent the size of the neighborhood; derivative the formula (1) and let F' (θ)g) When theta is equal to 0gThe expression of (a) is as follows:
3.3, Point (i)0，j0) The main gradient direction of (i) is perpendicular to the vessel direction at that point0，j0) Direction angle of vein
The estimation can be made using the following equation:
point (i)0，j0) The vein direction of (a) can be described by the following unit vector:
step four: tracking the vein line, after determining the seed points of the image, starting tracking twice in opposite directions from each seed point until the tracking is terminated when a stopping condition is met; the fourth step is specifically as follows:
let P0Is a seed point whose initial direction vector
The candidate point P of the next position is obtained by the formula (4)1Comprises the following steps:
wherein, P0And P1Is the respective coordinate, s is the tracking step length;
4.2, obtaining a candidate point P1Then, the true center point P can be determined by equation (6)1', while P can be substituted1The tracking direction of' is adjusted to:
thus, the iterative process of vein centerline tracking can be represented by:
4.3, when the tracking process encounters a bifurcation point, the selection mode of the central point is changed into the following mode:
wherein ρ (x, y) is an angle between the tracking direction vector of the point and the previous point; seThe gray maximum value point is a set of the gray maximum value points on the point section;
4.4, when two lines coexist on the vein central line, deleting the detection points with the distance smaller than a certain threshold value from the detected central line point set;
4.5, the stopping condition of the tracking algorithm is as follows:
1) the absolute value of the gray difference between the newly detected central point and the last central point is greater than a given threshold;
2) the newly detected gray value of the central point is smaller;
3) the included angle between the newly detected central point and the tracking direction of the last central point is greater than a given threshold value;
4) the distance between the center line being tracked and the already tracked center line is less than a given threshold;
step five: frangi filtering, namely processing the image by using a Frangi filter and a multi-scale linear filter, segmenting the vein line, and finally deleting the vein line according to the segmented vein line and the filtered image;
step six: and deleting the vein lines, and processing the Frangi filtered image by adopting multi-scale linear filtering.
2. The algorithm for extracting the vein line in the visible light vein imaging image according to claim 1, wherein the first step is specifically as follows: carrying out preliminary denoising by using a down-sampling method; and enhancing the filtering result by adopting a contrast-limiting self-adaptive histogram equalization method.
3. The algorithm for extracting the vein line from the visible light vein imaging image according to claim 1, wherein in the third step, a series of initial seed points of the image are determined, then the vein is tracked starting from the series of seed points, and finally the vein line is obtained by combining the tracking results of all the seed points.
4. The visible light vein imaging image vein line extraction algorithm according to claim 3, wherein the step of determining the initial seed point of the image comprises: firstly, dividing an image into a plurality of grid areas by a series of horizontal lines and vertical lines with equal intervals; and then searching the gray maximum value points on the lines row by row and column by column along the grid lines, and taking the points as candidate initial seed points.
5. The algorithm for extracting the vein line of the visible light vein imaging image according to claim 4, wherein the seed points are screened by using two thresholds, namely a global threshold and a local threshold, and the steps are as follows:
firstly, obtaining a threshold value T by using an Otsu method, and then passing through the threshold value TgCalculating to obtain a global threshold T as T/2g(ii) a Next, a local threshold T is setlThe expression is as follows:
Tl＝μseed+ασseed (5)
wherein, museedAnd σseedRespectively representing the gray level mean value and the standard deviation in the neighborhood of the seed point; α is a tuning parameter;
to obtain TgAnd TlThen, the same shall applyThe initial seed points that are larger than the two thresholds are retained, and the remaining seed points are deleted.
6. The visible light vein imaging image vein line extraction algorithm according to claim 5, wherein the initial seed point is corrected by the following specific steps:
let p be the initial seed point and,
is a direction vector of p, pcP is a seed point after correction, and the coordinates of the seed point satisfy the following conditions:
wherein, (x, y) is pixel coordinates; i is a pixel gray value; spIs a cross point p and is related to the direction vector
A collection of pixels in a vertical cross-section.
7. The vein line extraction algorithm of the visible light vein imaging image according to claim 1, wherein the fifth step is specifically:
and constructing a filter response function by using the eigenvalue of the Hessian matrix, wherein the response function of the Frangi filter is as follows:
wherein, sigma is the scale of the Gaussian function;
wherein λ1And λ2(|λ1|≥|λ2I) is the eigenvalue of the Hessian matrix, beta andc is a size parameter;
5.2, detecting veins by using a multi-scale filter, and adopting the maximum filter response value under n scales as the final filter output:
8. the vein line extraction algorithm of the visible light vein imaging image according to claim 1, wherein the sixth step is specifically:
carrying out curve segment segmentation on the original vein line, and deleting the vein curve segments meeting the following requirements:
in the formula, length (l)k) The length of the curve of the k-th line is shown,
representing the sum of the lengths of the segments of the k-th curve whose filter response is greater than a certain threshold, α is a constant between 0 and 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810810652.XA CN109145748B (en) | 2018-07-23 | 2018-07-23 | Vein line extraction algorithm of visible light vein imaging image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810810652.XA CN109145748B (en) | 2018-07-23 | 2018-07-23 | Vein line extraction algorithm of visible light vein imaging image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109145748A CN109145748A (en) | 2019-01-04 |
| CN109145748B true CN109145748B (en) | 2021-10-26 |
Family
ID=64801338
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810810652.XA Active CN109145748B (en) | 2018-07-23 | 2018-07-23 | Vein line extraction algorithm of visible light vein imaging image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109145748B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800711B (en) * | 2019-01-21 | 2020-12-22 | 数坤（北京）网络科技有限公司 | Vein removal method based on morphological recognition |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107749049A (en) * | 2017-09-07 | 2018-03-02 | 广州中国科学院先进技术研究所 | A kind of vein distribution display method and device |
-
2018
- 2018-07-23 CN CN201810810652.XA patent/CN109145748B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107749049A (en) * | 2017-09-07 | 2018-03-02 | 广州中国科学院先进技术研究所 | A kind of vein distribution display method and device |
Non-Patent Citations (4)
| Title |
|---|
| 一种基于近红外-可见光映射的快捷静脉提取算法;唐超颖;《东南大学学报（自然科学版）》;20160131;第46卷(第1期);全文 * |
| 基于Gabor滤波和形态学变换的手臂静脉线提取;彭晓光等;《计算技术与自动化》;20180331;第37卷(第1期);全文 * |
| 基于贝叶斯理论的手臂静脉线跟踪方法;高昊昇等;《光学学报》;20180227;第38卷(第2期);全文 * |
| 基于链码表示的手臂静脉特征提取与匹配;赵珊等;《光学学报》;20160531;第36卷(第5期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109145748A (en) | 2019-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108154519B (en) | Method and device for segmenting blood vessels in fundus image and storage medium | |
| Chutatape et al. | Retinal blood vessel detection and tracking by matched Gaussian and Kalman filters | |
| Wu et al. | On the adaptive detection of blood vessels in retinal images | |
| Amer et al. | Edge detection methods | |
| Yin et al. | Automatic segmentation and measurement of vasculature in retinal fundus images using probabilistic formulation | |
| CN110188614B (en) | NLM filtering finger vein denoising method based on skin crack segmentation | |
| Xia et al. | Plant leaf detection using modified active shape models | |
| CN109190460B (en) | Hand-shaped arm vein fusion identification method based on cumulative matching and equal error rate | |
| Zhang et al. | Improving performance of distribution tracking through background mismatch | |
| Thalji et al. | Iris Recognition using robust algorithm for eyelid, eyelash and shadow avoiding | |
| Xu et al. | Edge detection algorithm of medical image based on Canny operator | |
| Wang et al. | A fast method for automated detection of blood vessels in retinal images | |
| Kauba et al. | Advanced variants of feature level fusion for finger vein recognition | |
| CN112926516B (en) | Robust finger vein image region-of-interest extraction method | |
| Wieclawek et al. | Watershed based intelligent scissors | |
| CN109145748B (en) | Vein line extraction algorithm of visible light vein imaging image | |
| CN112734816B (en) | Heterologous image registration method based on CSS-Delaunay | |
| Sutanty et al. | Retinal blood vessel segmentation and bifurcation detection using combined filters | |
| Fang et al. | A new fingerprint thinning algorithm | |
| Ahamed et al. | Automated system for retinal vessel segmentation | |
| Ong et al. | Retina verification using a combined points and edges approach | |
| Jiu et al. | Biometric identification through detection of retinal vasculature | |
| Vlachos et al. | Vein segmentation in infrared images using compound enhancing and crisp clustering | |
| Cappelli et al. | Benchmarking local orientation extraction in fingerprint recognition | |
| Herry et al. | Segmentation and landmark identification in infrared images of the human body |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |