CN112017132A - Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix - Google Patents

Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix Download PDF

Info

Publication number
CN112017132A
CN112017132A CN202011131164.XA CN202011131164A CN112017132A CN 112017132 A CN112017132 A CN 112017132A CN 202011131164 A CN202011131164 A CN 202011131164A CN 112017132 A CN112017132 A CN 112017132A
Authority
CN
China
Prior art keywords
vein
image
scale
enhancement
directions
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.)
Pending
Application number
CN202011131164.XA
Other languages
Chinese (zh)
Inventor
韩静
赵壮
张毅
陈霄宇
郭恩来
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Zhipu Photoelectric Technology Co ltd
Original Assignee
Nanjing Zhipu Photoelectric Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing Zhipu Photoelectric Technology Co ltd filed Critical Nanjing Zhipu Photoelectric Technology Co ltd
Priority to CN202011131164.XA priority Critical patent/CN112017132A/en
Publication of CN112017132A publication Critical patent/CN112017132A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G06T5/70—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
    • 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
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
    • 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
    • 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

Abstract

The invention discloses a vein image enhancement method based on a maximum curvature method and a multi-scale Hessian matrix, and belongs to the technical field of acquisition and processing of superficial vein distribution images. Firstly, determining vein central lines in four directions, setting the range and the iteration step length of a scale factor, then initializing the scale factor, then performing Hessian enhancement to obtain an enhanced image, and then iterating the scale factor until the scale factor exceeds the maximum value of the range; then, the enhancement results in all directions are compared, and the maximum value is reserved; and finally, adding the maximum correspondences in the four directions to obtain a final enhancement result. The vein image enhancement method based on the maximum curvature method and the multi-scale Hessian matrix has the advantages that the extracted vein structure is complete and coherent, the details are rich, and the condition that the enhancement effect of broken lines and weak veins in the single-scale method enhancement result is not obvious is improved.

Description

Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix
Technical Field
The invention relates to a vein image enhancement method based on a maximum curvature method and a multi-scale Hessian matrix, and belongs to the technical field of acquisition and processing of superficial vein distribution images.
Background
Acquisition and processing of superficial vein distribution images are one of important research subjects in the field of biomedical imaging, and have important application values in the fields of medical diagnosis and treatment, biological feature identification and the like. For example, in the field of clinical medicine, the feature extraction of superficial veins is helpful to improve the success rate of venipuncture, assist diagnosis and treatment of vascular diseases, and the like; the superficial veins are characterized by universality and uniqueness, are difficult to forge or alter, and are more user-friendly and safe for biometric identification as a living feature. However, due to the difference of factors such as skin pigment, the thickness of blood vessels, the positions and depths of blood vessels, the thickness of fat and the like, the vein conditions are different among different individuals, the vein detection capability is different, and particularly, the vein detection of special people such as infants, old people and obese patients is more difficult.
Due to different absorption capacities of biological tissues to infrared light, the strong absorption capacity of hemoglobin in vein blood vessels to infrared light enables a vein image which is clearer than the vein distribution situation of visible light imaging to be obtained through near infrared imaging. However, near infrared imaging has the defects of low image contrast, low signal-to-noise ratio, blurred vein texture edge, poor visual effect and the like, and is not enough to be directly used for imaging or auxiliary diagnosis and treatment and the like. In order to improve the definition of the vein image and suppress noise, so as to facilitate the identification of human eyes or machines, the acquired vein image needs to be enhanced.
Therefore, a vein image enhancement method based on the maximum curvature method and the multi-scale Hessian matrix is needed.
Disclosure of Invention
In order to solve the technical problems, the invention discloses a vein image enhancement method based on a maximum curvature method and a multi-scale Hessian matrix, which has the following specific technical scheme: comprises that
Step 1, detecting vein central lines in four directions of vein image
Figure 93337DEST_PATH_IMAGE001
；
Step 2, setting scale factors
Figure 526855DEST_PATH_IMAGE002
Range of (1)
Figure 229231DEST_PATH_IMAGE003
And iteration step size
Figure 918839DEST_PATH_IMAGE004
；
Step 3, initializing scale factors
Figure 787700DEST_PATH_IMAGE005
；
Step 4, pair
Figure 652888DEST_PATH_IMAGE006
Respectively carrying out Hessian enhancement to obtain enhanced images in four directions
Figure 713117DEST_PATH_IMAGE007
；
Step 5, iterating the scale factor
Figure 657064DEST_PATH_IMAGE008
Jump to step 4 until
Figure 578884DEST_PATH_IMAGE009
And step 6 is carried out;
step 6, comparing the enhancement result graphs in all directions, and keeping the maximum value to obtain
Figure 547846DEST_PATH_IMAGE010
；
And 7, adding the maximum responses in the four directions to obtain a final enhancement result.
Further, with respect to vein centerlines
Figure 529708DEST_PATH_IMAGE011
，
To obtain a more accurate center point position, multiple directions of the vein image are detected, and representative angles of 0 °, 45 °, 90 ° and 135 ° are selected, as shown in fig. 1. And respectively carrying out curvature calculation on all cross section curves in the 4 directions of the vein image to obtain points of all local curvature maximum values, and forming a vein central point position detection result graph in the 4 directions.
By using
Figure 960952DEST_PATH_IMAGE012
Representing pixel points on a vein image
Figure 732467DEST_PATH_IMAGE013
Is determined by the gray-scale value of (a),
Figure 572247DEST_PATH_IMAGE014
then the image is represented
Figure 725011DEST_PATH_IMAGE015
A certain position
Figure 377972DEST_PATH_IMAGE016
The cross section curve of the upper part has a mapping function
Figure 156441DEST_PATH_IMAGE017
So that
Figure 585148DEST_PATH_IMAGE018
。
Curvature
Figure 659546DEST_PATH_IMAGE019
The formula of (c) is shown as follows:
Figure 563917DEST_PATH_IMAGE020
（1）
cross-sectional curve of vein image
Figure 631230DEST_PATH_IMAGE021
Is concave upwards, then its curvature
Figure 399597DEST_PATH_IMAGE022
The value is positive and the lowest point of the fovea is the center point of the vein. One curve is recorded with
Figure 346693DEST_PATH_IMAGE023
Each concave region can be calculated to obtain a local curvature maximum value point, namely a vein central point, and the position of the vein central point is recorded as
Figure 285830DEST_PATH_IMAGE024
Wherein, in the step (A),
Figure 704304DEST_PATH_IMAGE025
. The width of the concave region can be regarded as the width of the vein and is set as the center point of the vein
Figure 45287DEST_PATH_IMAGE024
Within the neighborhood range of (2), the number of continuous pixels with positive curvature is recorded as
Figure 304230DEST_PATH_IMAGE026
。
Figure 652035DEST_PATH_IMAGE026
The larger the center point
Figure 77462DEST_PATH_IMAGE024
The greater the width of the vein in which it is located. To indicate the detected centre point
Figure 335268DEST_PATH_IMAGE027
The probability of the vein region located at each central point is provided with a score
Figure 93008DEST_PATH_IMAGE028
Definition of
Figure 757470DEST_PATH_IMAGE029
. Projecting a score to
Figure 32594DEST_PATH_IMAGE030
Plane, as shown in the following formula:
Figure 331857DEST_PATH_IMAGE031
（2）
therein, a point
Figure 293122DEST_PATH_IMAGE032
Is composed of
Figure 756464DEST_PATH_IMAGE033
And (4) defining.
Respectively carrying out the above calculation on cross section curves of the vein image in four directions of horizontal, vertical, 135 degrees and 45 degrees to obtain the preliminarily detected vein centerline images in the four directions
Figure 772962DEST_PATH_IMAGE034
As shown in fig. 2.
Further, in order to obtain a clear and accurate vein structure and suppress noise, a Hessian matrix is adopted to enhance the vein center images in four directions extracted by the maximum curvature method in the last step. The vein vessel is a linear tubular structure, and the vessel enhancement can be regarded as a filtering process for finding a linear structure. The Hessian matrix is sensitive to the structure, is often used for extracting linear structure features, and can be applied to two-dimensional and three-dimensional vein images.
The basic idea of the Hessian matrix is to extract the main direction of the local second order structural decomposition of the image. The eigenvalue and eigenvector of the Hessian matrix can be used to describe the local structure of the image, the eigenvector corresponding to the eigenvalue with the smallest absolute value represents the direction with the smallest curvature, i.e. along the eigen direction of the image, and the eigenvector corresponding to the eigenvalue with the largest absolute value represents the direction with the largest curvature, i.e. perpendicular to the eigen direction of the image.
A binary Hessian matrix is defined as follows:
Figure 254628DEST_PATH_IMAGE035
（3）
wherein the content of the first and second substances,
Figure 432799DEST_PATH_IMAGE036
as an image
Figure 113662DEST_PATH_IMAGE037
In that
Figure 153425DEST_PATH_IMAGE038
The second partial derivative in the direction of the direction,
Figure 99384DEST_PATH_IMAGE039
is composed of
Figure 199189DEST_PATH_IMAGE040
In that
Figure 371545DEST_PATH_IMAGE041
The second partial derivative in the direction of the direction,
Figure 979112DEST_PATH_IMAGE042
representative image
Figure 546622DEST_PATH_IMAGE043
In that
Figure 332176DEST_PATH_IMAGE038
Direction and
Figure 241095DEST_PATH_IMAGE044
the mixed partial derivative in direction is shown as follows:
Figure 137506DEST_PATH_IMAGE045
（4）
Figure 825102DEST_PATH_IMAGE046
（5）
Figure 827562DEST_PATH_IMAGE047
（6）
since the second-order partial derivative is sensitive to noise, the Hessian matrix is first subjected to gaussian smoothing, which is considered as the convolution of the image with the second-order gaussian derivative, as shown in the following formula:
Figure 646613DEST_PATH_IMAGE048
（7）
Figure 35132DEST_PATH_IMAGE049
（8）
Figure 934823DEST_PATH_IMAGE050
（9）
wherein, in order
Figure 62179DEST_PATH_IMAGE051
Two-dimensional Gaussian function of scale
Figure 915997DEST_PATH_IMAGE052
And the second order partial derivative thereof is shown as follows:
Figure 154212DEST_PATH_IMAGE053
（10）
Figure 49355DEST_PATH_IMAGE054
（11）
Figure 98345DEST_PATH_IMAGE055
（12）
Figure 642721DEST_PATH_IMAGE056
（13）
for vein central line images detected in four directions
Figure 684626DEST_PATH_IMAGE057
Hessian matrix enhancement is performed respectively. In the horizontal direction thereof
Figure 840801DEST_PATH_IMAGE058
For example, will
Figure 670479DEST_PATH_IMAGE059
Respectively carrying out convolution operation with three second-order partial derivatives of the Gaussian function to obtain second-order partial derivatives at three different angles in horizontal, vertical and diagonal directions
Figure 13736DEST_PATH_IMAGE060
. For Hessian matrix
Figure 344485DEST_PATH_IMAGE061
The absolute value of the characteristic value is searched and sequenced to obtain the characteristic value
Figure 214221DEST_PATH_IMAGE062
And
Figure 854281DEST_PATH_IMAGE063
from the feature values, a vein similarity function can be constructed.
Figure 435566DEST_PATH_IMAGE064
The above pixel points pass through the vein similarity function to obtain response output
Figure 819274DEST_PATH_IMAGE065
。
Further, in general, pixels in an image can be classified into three categories:
1. background, the gray distribution is more uniform, and the curvature in each direction is smaller;
2. isolated points, with greater curvature in each direction;
3. the blood vessel has a large curvature in the radial direction and a small curvature in the axial direction.
Table 1 summarizes the correspondence between eigenvalues of the Hessian matrix and structural shapes in the two-dimensional case,
Figure 543517DEST_PATH_IMAGE066
and
Figure 167527DEST_PATH_IMAGE067
respectively representing the magnitude and magnitude of the characteristic value, the signs + and-then representing the positive and negative of the characteristic value, setting
Figure 157480DEST_PATH_IMAGE068
。
TABLE 1 relationship table of characteristic values of Hessian matrix of various possible shape structures under two-dimensional conditions
Figure 531829DEST_PATH_IMAGE069
Further, the vein similarity function is defined as follows:
Figure 963377DEST_PATH_IMAGE070
（14）
wherein the content of the first and second substances,
Figure 7556DEST_PATH_IMAGE071
can be used to distinguish structural shapes.
Figure 812701DEST_PATH_IMAGE072
The sensitivity for regulating and distinguishing block region and strip region, c has influence on the smooth degree of filtered image, S is a second-order structure, and has
Figure 787479DEST_PATH_IMAGE073
。
Figure 987779DEST_PATH_IMAGE074
And
Figure 937280DEST_PATH_IMAGE075
set to 0.5 and 450, respectively.
Vein center line image for other directions
Figure 416672DEST_PATH_IMAGE076
Respectively performing the same operation to obtain
Figure 368710DEST_PATH_IMAGE077
. The Hessian matrix enhancement results are shown in fig. 3.
Further, the maximum responses in the four directions are added to obtain the final enhancement result.
Vein central line image of four directions enhanced by Hessian matrix
Figure 679DEST_PATH_IMAGE078
And
Figure 635929DEST_PATH_IMAGE079
it is necessary to merge and filter out noise to achieve the final vein structure extraction. Every pixel point on every vein center image
Figure 87770DEST_PATH_IMAGE080
Comparing with two adjacent pixels on the left and right sides of the pixel, and connecting the pixels if the values of the pixel and the left and right pixels are larger; if the pixel point value is smaller, the two side pixelsIf the value of the point is larger, only marking the pixel point; if the pixel point is larger and the values of the pixel points on the two sides are smaller, the point is marked as noise. This process can be represented by the following formula:
Figure 843498DEST_PATH_IMAGE081
（15）
the same operation is done in the other three directions as shown in the following equation:
Figure 579242DEST_PATH_IMAGE082
（16）
Figure 136125DEST_PATH_IMAGE083
（17）
Figure 498099DEST_PATH_IMAGE084
（18）
finally, the four directional images are compared point by point
Figure 618371DEST_PATH_IMAGE085
And
Figure 21670DEST_PATH_IMAGE086
and taking the maximum value of the values of the same pixel point as a final vein feature image, wherein the maximum value is shown as the following formula:
Figure 437870DEST_PATH_IMAGE087
。
has the advantages that: the vein image enhancement method based on the maximum curvature method and the multi-scale Hessian matrix has the advantages that the extracted vein structure is complete and coherent, the details are rich, and the condition that the enhancement effect of broken lines and weak veins in the single-scale method enhancement result is not obvious is improved.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic diagram of the direction of curvature calculation of a cross-sectional curve;
figure 3 vein image vein centerline image in four directions: (a) original drawing; (b) level of
Figure 51254DEST_PATH_IMAGE089
(ii) a (c) Is perpendicular to
Figure 725949DEST_PATH_IMAGE091
； (d) 135°
Figure 734487DEST_PATH_IMAGE093
；(e) 45°
Figure 898753DEST_PATH_IMAGE095
；
FIG. 4 is a diagram of a Hessian enhancement result of a vein central line image in four directions of a vein image; (a) original drawing; (b) level of
Figure 405957DEST_PATH_IMAGE097
(ii) a (c) Is perpendicular to
Figure 946660DEST_PATH_IMAGE099
； (d) 135°
Figure 809705DEST_PATH_IMAGE101
；(e) 45°
Figure 82554DEST_PATH_IMAGE103
；
FIG. 5 EMC and EMCs enhancement results compare: (a) original drawing; (b) EMC enhancement results; (c) (iii) EMCs enhancement results; (d) filtering burrs by Gaussian filtering after EMCs are enhanced;
FIG. 6 contrast results of different algorithm enhancements under the same lighting conditions: (a) original drawing; (b) CLAHE; (c) retinex; (d) EMC; (e) EMC-C; (f) EMC-fast; (g) EMCs;
FIG. 7 contrast the enhancement effect of different algorithms under different lighting conditions: (a) original drawing; (b) CLAHE; (c) retinex; (d) EMC; (e) EMC-fast; (f) EMCs.
Detailed Description
The present invention is further illustrated by the following figures and specific examples, which are to be understood as illustrative only and not as limiting the scope of the invention, which is to be given the full breadth of the appended claims and any and all equivalent modifications thereof which may occur to those skilled in the art upon reading the present specification.
The vein Enhancement Method (EMC) combining the maximum curvature method and the single-scale Hessian matrix enhancement can well extract main vein structures, but thin and weak deep veins are lost. To obtain more details of the vein, this section proposes to fuse Hessian enhancement features of different scales of the vein centerline.
The vein diameters in the vein images are different, and the vein structure extracted by adopting single-scale Hessian enhancement only has a part of veins with relatively consistent diameters, as shown in FIG. 4, the detected main veins are relatively obvious and thick veins on the original vein image. Some of the deeper and weaker veins are not enhanced or even completely unextracted. For better enhancement results, different size scales are set
Figure 404951DEST_PATH_IMAGE104
Matching veins of different diameters. Determining responses at different scales
Figure 172181DEST_PATH_IMAGE105
Then, the maximum response is obtained as shown in the following formula:
Figure 139000DEST_PATH_IMAGE106
the specific process is as follows:
1. detecting vein center lines in four directions of vein image
Figure 832019DEST_PATH_IMAGE107
；
2. Setting scale factors
Figure 143177DEST_PATH_IMAGE108
Range of (1)
Figure 494524DEST_PATH_IMAGE109
And iteration step size
Figure 768379DEST_PATH_IMAGE110
；
3. Initializing scale factors
Figure 71447DEST_PATH_IMAGE111
；
4. To pair
Figure 509381DEST_PATH_IMAGE112
Respectively carrying out Hessian enhancement to obtain enhanced images in four directions
Figure 867681DEST_PATH_IMAGE113
；
5. Iterative scale factor
Figure 58360DEST_PATH_IMAGE114
Jump to step 4 until
Figure 843913DEST_PATH_IMAGE115
And step 6 is carried out;
6. then comparing the enhancement result graphs in all directions, and keeping the maximum value to obtain
Figure 129664DEST_PATH_IMAGE116
And
Figure 275343DEST_PATH_IMAGE117
；
7. and adding the maximum responses in the four directions to obtain a final enhancement result.
Compared with the EMC method, the enhancement result of the vein Enhancement Method (EMCs) combining the maximum curvature method and the multi-scale Hessian matrix enhancement is shown in FIG. 5, the contrast of the enhancement effect of the EMCs is higher, the vein structure is more consistent, and the broken line part in the EMC algorithm enhancement result is enhanced in the EMCs algorithm. The weak veins which are not obvious in the original vein image have better enhancement effect in the EMCs algorithm.
Results and analysis of the experiments
The invention improves a vein Enhancement Method (EMC) combining a maximum curvature method and Hessian matrix enhancement in the prior art, and provides a CUDA acceleration algorithm (EMC-C) of the EMC, a structure optimization algorithm (EMC-fast) based on the EMC-C and an EMCs algorithm of multi-scale Hessian matrix enhancement.
(1) Different algorithms enhance result comparison
The vein enhancement results of the different algorithms are shown for example in fig. 6 and 7:
as can be seen from fig. 6, the CLAHE algorithm highlights the vein structure to some extent, improving the image contrast, but at the same time the noise is also enhanced. The Retinex algorithm results in an overall gray value improvement, but does not enhance the venous structures well. Most veins are extracted by EMC and EMC-C algorithms, the contrast of an EMC-fast enhancement result is higher, but the three algorithms have weaker enhancement capability on the thin and weak veins, and the extracted vein structure has a fracture condition. The EMCs algorithm provided for realizing more accurate vein image enhancement further improves the enhancement effect, has higher image contrast, more complete vein structure, few broken lines and richer details, and can highlight weak veins with unobvious enhancement effect in other methods.
Fig. 7 shows the enhancement effect of each algorithm under different lighting conditions. Therefore, most algorithms including CLAHE, EMC and EMC-fast have poor enhancement effect on the vein images acquired under the dark light condition, and the Retinex algorithm has better enhancement effect on the images under the dark light than under the bright light condition, but enhances the noise. The enhancement effect of the EMCs algorithm is not influenced by the illumination condition, and images with high contrast and rich details can be enhanced.
(2) Enhanced image quality assessment
Since the vein image lacks a corresponding clear reference image, several commonly used non-reference image quality evaluation functions are used to analyze the enhancement effect.
(2.1) gradient function Brenner
Brenner is a simple gradient evaluation function, which is the sum of the squares of the differences in gray values of any two adjacent pixels on the image, as shown by the following equation:
Figure 336840DEST_PATH_IMAGE118
(2.2) Gray-level differential product function SMD2
The gray difference product function takes the gray change as an evaluation criterion, and accumulates the product of the gray difference between each pixel and two adjacent pixels, as shown in the following formula:
Figure DEST_PATH_IMAGE119
(2.3) standard deviation STD
The standard deviation describes the degree of dispersion of the pixel gray scale value from the mean value on the image, and the larger the standard deviation, the more dispersed the image gray scale is, the better the image quality is, as shown in the following formula:
Figure 371923DEST_PATH_IMAGE120
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE121
is the size of the image size and,
Figure 518871DEST_PATH_IMAGE122
is the image mean.
The enhancement performance indicators for the different algorithms are shown in table 2.
TABLE 2 different algorithms enhancement effect index
Figure DEST_PATH_IMAGE123
As can be seen from Table 2, the enhancement results of the method (EMCs) combining the maximum curvature method and the multi-scale Hessian matrix and the EMC method (EMC-fast) subjected to CUDA acceleration and structure optimization are superior to other methods in three indexes of Brenner, SMD2 and STD, and are greatly improved relative to the original vein image. Among them, the indexes of EMCs are the best. Compared with EMC, EMC-fast and EMCs are both more advantageous in vein image enhancement, and the enhancement result indexes are about 1.5 times and 1.95 times of EMC respectively.
In summary, in order to enhance more vein structure details and realize the extraction of fine and deep veins, the invention proposes to combine a multiscale Hessian matrix with a maximum curvature method on the basis of the original algorithm and set the scale
Figure 828760DEST_PATH_IMAGE104
Different Hessian matrixes are matched and enhanced with vein vessels with different diameters, and then multi-scale enhancement results are fused. Experiments prove that compared with other traditional image enhancement methods and the original vein image enhancement algorithm based on the maximum curvature method and the Hessian matrix, the maximum curvature method and the multi-scale Hessian combined method can obtain the enhancement effect with higher contrast, the extracted vein structure is complete and coherent, the details are rich, and the condition that the enhancement effect of broken lines and weak veins in the single-scale method enhancement result is not obvious is improved.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.
Furthermore, it should be understood that although the present description refers to embodiments, not every embodiment may contain only a single embodiment, and such description is for clarity only, and those skilled in the art should integrate the description, and the embodiments may be combined as appropriate to form other embodiments understood by those skilled in the art.

Claims (6)

1. A vein image enhancement method based on a maximum curvature method and a multi-scale Hessian matrix is characterized in that: comprises that
Step 1, detecting vein central lines in four directions of vein image
Figure 49436DEST_PATH_IMAGE001
；
Step 2, setting scale factors
Figure 583317DEST_PATH_IMAGE002
Range of (1)
Figure 296189DEST_PATH_IMAGE003
And iteration step size
Figure 737666DEST_PATH_IMAGE004
；
Step 3, initializing scale factors
Figure 367230DEST_PATH_IMAGE005
；
Step 4, pair
Figure 72012DEST_PATH_IMAGE006
Respectively carrying out Hessian enhancement to obtain enhanced images in four directions
Figure 272180DEST_PATH_IMAGE007
；
Step 5, iterating the scale factor
Figure 969878DEST_PATH_IMAGE008
Jump to step 4 until
Figure 204681DEST_PATH_IMAGE009
And step 6 is carried out;
step 6, comparing the enhancement result graphs in all directions, and keeping the maximum value to obtain
Figure 345944DEST_PATH_IMAGE010
；
And 7, adding the maximum responses in the four directions to obtain a final enhancement result.
2. The vein image enhancement method based on the maximum curvature method and the multi-scale Hessian matrix according to claim 1, wherein: vein center line in the step 1
Figure 548255DEST_PATH_IMAGE011
The obtaining step comprises:
step 1.1: detecting a plurality of directions of the vein image, and selecting representative angles of 0 degree, 45 degrees, 90 degrees and 135 degrees;
step 1.2: respectively carrying out curvature calculation on all cross section curves in the 4 directions of the vein image to obtain points of all local curvature maximum values, and forming a vein central point position detection result graph in the 4 directions;
step 1.3: by using
Figure 65955DEST_PATH_IMAGE012
Representing pixel points on a vein image
Figure 155265DEST_PATH_IMAGE013
Is determined by the gray-scale value of (a),
Figure 185538DEST_PATH_IMAGE014
then the image is represented
Figure 360298DEST_PATH_IMAGE015
A certain position
Figure 150531DEST_PATH_IMAGE016
The cross section curve of the upper part has a mapping function
Figure 343615DEST_PATH_IMAGE017
So that
Figure 295522DEST_PATH_IMAGE018
(ii) a Curvature
Figure 206846DEST_PATH_IMAGE019
The formula of (c) is shown as follows:
Figure 800769DEST_PATH_IMAGE020
；
step 1.4: cross-sectional curve of vein image
Figure 864671DEST_PATH_IMAGE021
Is concave upwards, then its curvature
Figure 767905DEST_PATH_IMAGE022
The value is positive, and the lowest point of the concave area is the central point of the vein;
Figure 182837DEST_PATH_IMAGE023
one curve is recorded with
Figure 314873DEST_PATH_IMAGE024
Each concave region can be calculated to obtain a local curvature maximum value point, namely a vein central point, and the position of the vein central point is recorded as
Figure 482549DEST_PATH_IMAGE025
Wherein, in the step (A),
Figure 41837DEST_PATH_IMAGE026
(ii) a The width of the concave region is the width of the vein and is set as the center point of the vein
Figure 678486DEST_PATH_IMAGE027
Within the neighborhood range of (2), the number of continuous pixels with positive curvature is recorded as
Figure 863480DEST_PATH_IMAGE028
；
Figure 901974DEST_PATH_IMAGE028
The larger the center point
Figure 897743DEST_PATH_IMAGE027
The greater the width of the vein in which it is located; to indicate the detected centre point
Figure 5376DEST_PATH_IMAGE029
The probability of the vein region located at each central point is provided with a score
Figure 479214DEST_PATH_IMAGE030
Definition of the fraction
Figure 372215DEST_PATH_IMAGE030
Projected to the target
Figure 788153DEST_PATH_IMAGE031
Plane, as shown in the following formula:
Figure 133814DEST_PATH_IMAGE032
(ii) a Therein, a point
Figure 411343DEST_PATH_IMAGE033
Is composed of
Figure 673697DEST_PATH_IMAGE034
Defining;
step 1.5, cross section curves of the vein image in four directions of horizontal 0 degrees, vertical 90 degrees, 135 degrees and 45 degrees are respectively calculated in steps 1.1-1.4, and a vein central line image preliminarily detected in the four directions is obtained
Figure 745690DEST_PATH_IMAGE035
。
3. The vein image enhancement method based on the maximum curvature method and the multi-scale Hessian matrix according to claim 1, wherein: the Hessian enhancement in the step 3 comprises the following steps:
step 3.1: will be provided with
Figure 578647DEST_PATH_IMAGE036
Respectively carrying out convolution operation with three second-order partial derivatives of the Gaussian function to obtain second-order partial derivatives at three different angles in horizontal, vertical and diagonal directions
Figure 643555DEST_PATH_IMAGE037
Step 3.2: for Hessian matrix
Figure 487711DEST_PATH_IMAGE038
The absolute value of the characteristic value is searched and sequenced to obtain the characteristic value
Figure 996184DEST_PATH_IMAGE039
And
Figure 565705DEST_PATH_IMAGE040
constructing a vein similarity function from the characteristic values;
step 3.3:
Figure 919457DEST_PATH_IMAGE041
the above pixel points pass through the vein similarity function to obtain response output
Figure 641557DEST_PATH_IMAGE042
；
Step 3.4: vein center line image for other directions
Figure 304619DEST_PATH_IMAGE043
Respectively carrying out the operations of the steps 3.1 to 3.3 to obtain
Figure 846590DEST_PATH_IMAGE044
。
4. The vein image enhancement method based on the maximum curvature method and the multi-scale Hessian matrix according to claim 3, wherein: obtaining the characteristic value in the step 3.2
Figure 253301DEST_PATH_IMAGE045
And
Figure 829907DEST_PATH_IMAGE046
comparing the corresponding relation between the characteristic value of the Hessian matrix and the structural shape under the two-dimensional condition; definition H and L represent the magnitude and magnitude of the characteristic value, respectively, and the definition symbols + and-represent the positive and negative of the characteristic value, respectively
Figure 680182DEST_PATH_IMAGE047
(ii) a When the linear structure is bright, the characteristic value L corresponds to
Figure 693138DEST_PATH_IMAGE048
Characteristic value H-corresponding
Figure 654272DEST_PATH_IMAGE049
(ii) a When the linear structure is dark, the characteristic value corresponds to
Figure 334652DEST_PATH_IMAGE050
The characteristic value H + corresponds to
Figure 621408DEST_PATH_IMAGE051
(ii) a When the lumpy structure is bright, the characteristic value H-corresponds to
Figure 137971DEST_PATH_IMAGE052
Characteristic value H-corresponding
Figure 886484DEST_PATH_IMAGE053
(ii) a When the lump-shaped structure is dark, the characteristic value H + corresponds to
Figure 437682DEST_PATH_IMAGE054
Characteristic value H-corresponding
Figure 629760DEST_PATH_IMAGE055
。
5. The vein image enhancement method based on the maximum curvature method and the multi-scale Hessian matrix according to claim 4, wherein: the vein similarity function in step 3.2 is defined as follows:
Figure 882887DEST_PATH_IMAGE056
wherein the content of the first and second substances,
Figure 185823DEST_PATH_IMAGE057
can be used to distinguish structural shapes;
Figure 309637DEST_PATH_IMAGE058
used for adjusting the sensitivity of distinguishing the block area and the strip area, C has influence on the smoothness of the filtered image,
Figure 938196DEST_PATH_IMAGE059
is a second-order structure, having
Figure 429351DEST_PATH_IMAGE060
；
Figure 519667DEST_PATH_IMAGE061
And
Figure 779878DEST_PATH_IMAGE062
set to 0.5 and 450, respectively.
6. The vein image enhancement method based on the maximum curvature method and the multi-scale Hessian matrix according to claim 1, wherein: the step of gaussian convolution in step 3.1 is to define a binary Hessian matrix as follows:
Figure 297447DEST_PATH_IMAGE063
wherein the content of the first and second substances,
Figure 275899DEST_PATH_IMAGE064
as an image
Figure 920638DEST_PATH_IMAGE065
In that
Figure 284623DEST_PATH_IMAGE066
The second partial derivative of the direction is,
Figure 723826DEST_PATH_IMAGE067
is composed of
Figure 455152DEST_PATH_IMAGE068
In that
Figure 152850DEST_PATH_IMAGE069
The second partial derivative in the direction of the direction,
Figure 856495DEST_PATH_IMAGE070
representative image
Figure 981446DEST_PATH_IMAGE065
In that
Figure 668910DEST_PATH_IMAGE066
Direction and
Figure 904719DEST_PATH_IMAGE069
the mixed partial derivatives in direction, I (x, y), represent the two-dimensional image as shown in the following equation:
Figure 994029DEST_PATH_IMAGE071
since the second partial derivative is sensitive to noise, when the Hessian matrix is calculated, gaussian smoothing, i.e. convolution of the image with the second derivative of gaussian, is performed as shown in the following equation:
Figure 40614DEST_PATH_IMAGE072
wherein, in order
Figure 464642DEST_PATH_IMAGE073
Two-dimensional Gaussian function of scale
Figure 254874DEST_PATH_IMAGE074
And the second order partial derivative thereof is shown as follows:
Figure 198691DEST_PATH_IMAGE075
。
CN202011131164.XA 2020-10-21 2020-10-21 Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix Pending CN112017132A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011131164.XA CN112017132A (en) 2020-10-21 2020-10-21 Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011131164.XA CN112017132A (en) 2020-10-21 2020-10-21 Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix

Publications (1)

Publication Number Publication Date
CN112017132A true CN112017132A (en) 2020-12-01

Family

ID=73527469

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011131164.XA Pending CN112017132A (en) 2020-10-21 2020-10-21 Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix

Country Status (1)

Country Link
CN (1) CN112017132A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113256518A (en) * 2021-05-20 2021-08-13 上海理工大学 Structured light image enhancement method for intraoral 3D reconstruction

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113256518A (en) * 2021-05-20 2021-08-13 上海理工大学 Structured light image enhancement method for intraoral 3D reconstruction
CN113256518B (en) * 2021-05-20 2022-07-29 上海理工大学 Structured light image enhancement method for intraoral 3D reconstruction

Similar Documents

Publication Publication Date Title
Zhao et al. Retinal vessels segmentation based on level set and region growing
Zhao et al. Automated vessel segmentation using infinite perimeter active contour model with hybrid region information with application to retinal images
Mary et al. Retinal fundus image analysis for diagnosis of glaucoma: a comprehensive survey
Lupascu et al. FABC: retinal vessel segmentation using AdaBoost
Lam et al. A novel vessel segmentation algorithm for pathological retina images based on the divergence of vector fields
Mendonça et al. Automatic localization of the optic disc by combining vascular and intensity information
CN108986106A (en) Retinal vessel automatic division method towards glaucoma clinical diagnosis
Al-Shamlan et al. Feature extraction values for breast cancer mammography images
EP3724848B1 (en) Image processing method for glaucoma detection and computer program products thereof
Jaafar et al. Automated detection of red lesions from digital colour fundus photographs
Jony et al. Detection of lung cancer from CT scan images using GLCM and SVM
de Souza Jr et al. Automatic frontal sinus recognition in computed tomography images for person identification
Yang et al. Algorithm for detecting micro-aneurysms in low-resolution color retinal images
Pock et al. Multiscale medialness for robust segmentation of 3d tubular structures
Rodrigues et al. Retinal vessel segmentation using parallel grayscale skeletonization algorithm and mathematical morphology
Abdallah et al. An automated vessel segmentation of retinal images using multiscale vesselness
CN112233789A (en) Regional feature fusion type hypertensive retinopathy classification method
CN108665474B (en) B-COSFIRE-based retinal vessel segmentation method for fundus image
CN112017132A (en) Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix
Liu et al. Retinal vessel segmentation using densely connected convolution neural network with colorful fundus images
CN111292285B (en) Automatic screening method for diabetes mellitus based on naive Bayes and support vector machine
Salem et al. Unsupervised segmentation of retinal blood vessels using a single parameter vesselness measure
CN110390678A (en) A kind of organization type dividing method of colorectal cancer IHC dye image
KG et al. An approach to the detection of retinoblastoma based on apriori algorithm
Isavand Rahmani et al. Retinal blood vessel segmentation using gabor filter and morphological reconstruction

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20201201

RJ01 Rejection of invention patent application after publication