CN103632137B - A kind of human eye iris segmentation method - Google Patents
A kind of human eye iris segmentation method Download PDFInfo
- Publication number
- CN103632137B CN103632137B CN201310570892.4A CN201310570892A CN103632137B CN 103632137 B CN103632137 B CN 103632137B CN 201310570892 A CN201310570892 A CN 201310570892A CN 103632137 B CN103632137 B CN 103632137B
- Authority
- CN
- China
- Prior art keywords
- human eye
- image
- iris
- sample
- training set
- 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.)
- Expired - Fee Related
Links
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of human eye iris segmentation method, the method concrete steps: step one: training set eye image sample is positioned human eye canthus one by one;Step 2: training set human eye iris image sample is positioned human eye center one by one;Step 3: use, based on sparse and low-rank decomposition algorithm, training set is carried out batch alignment;Step 4: the method that the training set after batch alignment uses Canny rim detection and Hough transformation combine realizes the iris segmentation to training set eye image;Step 5: the test picture of input is realized human eye iris segmentation.The method uses algorithm based on sparse and low-rank decomposition to training set sample batch alignment, solve the inconsistent problem blocked with human eye eyelash of the brightness in great amount of samples, then use Canny rim detection and Hough transformation that the image and test image that remove brightness flop and occlusion issue are reached the purpose of iris segmentation.The method can be widely used in iris identification field.
Description
Technical field
The present invention relates to image steganalysis field, particularly to a kind of human eye iris segmentation method.
Background technology
We are in the society of an advanced IT application at present, and people are increasing to information requirement, meanwhile, to information
Security requirement more and more higher, and identity recognizing technology be exactly one improve Information Security method, it is increasingly
Many fields are paid much attention to by people.Identification is i.e. some uniqueness characteristic utilizing human body, uses some skill
This this feature is differentiated by art, thus is identified the identity of people.Previous conventional identity recognizing technology be dependent on face,
The features such as fingerprint, hand-type, signature are identified, but these features are all the surfaces of some human bodies, also exist the biggest
Easy change so that rely on these features to carry out identification and can become not to be the most reliable.In recent years, iris knowledge has been risen
Other technology, due to the uniqueness of iris, unchangeable property, can not by the operation uniqueness such as change, be allowed in scientific research and
Industrial circle serves the most important effect.But, due to the special construction of iris, in image acquisition process, we
Iris image purely can not be shot, the iris image generally collected not only comprises iris, also comprises eyes other
Part, such as pupil, eyelid, eyelashes etc., such picture can not be directly identified by iris recognition technology, can only be to iridial part of retina
Divide and identify.Therefore an important pretreatment of iris recognition technology is exactly human eye iris segmentation.
The result of human eye iris segmentation is typically used in iris identification, is the straight of iris identity recognizing technology
Connecing objective for implementation, the accuracy of segmentation badly influences the accuracy of identification, and therefore iris splitting method is critically important, is to protect
Barrier iris identification preprocessing means accurately and one of key technology.
Eye image is mainly directly detected by human eye iris segmentation, is partitioned into iris portion therein.Rainbow
Film image dividing method mainly utilizes iris outer edge to be approximately circular, uses the circle in the modeling of iris outer edge or image
Shape detection method is carried out.Wherein, iris outer edge models and often follows following steps: rim detection and edge modeling.Edge
Detection is typically the imagery exploitation canny after gaussian filtering or sobel operator are carried out rim detection.Then typically by number
Edge in the image of rim detection binaryzation is modeled by the method learned.And the circle detection in image is generally also and first adopts
By rim detection, then the image after rim detection is used the inside and outside circle border of Hough change-detection iris, thus realize iris
Segmentation.
Although these algorithm comparative maturities above-mentioned, but these methods are all also existing identical shortcoming.Owing to normally adopting
Collect to iris sample in there is brightness flop and human eye is the most ciliary blocks, the circle of the inner and outer boundary of iris can be made
Shape is inconspicuous, causes segmentation errors.The iris image being simultaneously partitioned into also exists brightness flop, the upper and lower ciliary screening of human eye
The shortcomings such as gear, sample image not alignment, can cause the biggest impact, make the accuracy of identification drop follow-up iris identification
Low.
Summary of the invention
The present invention provides a kind of human eye iris segmentation method, can remove with batch alignment eye image to be detected
Ciliary in brightness flop in eye image, upper lower eyelid block, set up clear, remove human eye that block, batch alignment
Image pattern model, then these samples are realized iris segmentation.
The present invention solves the technical scheme of the problems referred to above and is characterized mainly in that: it specifically comprises the following steps that
Step one: the training set eye image sample of key words sorting is carried out human eye canthus location one by one, utilizes
Then these angle points are carried out retrieval traversal, select abscissa by the angle point in Harris Corner Detection Algorithm detection eye image
Minimum point is human eye canthus;
Step 2: training set human eye iris image sample is positioned human eye center one by one, first by threshold value to sample to be detected
Picture carries out binaryzation, then utilizes canny edge detection method that binary image is carried out rim detection and obtains binaryzation
Boundary graph, recycling Hough transformation is round to boundary graph detection, is set to the inner edge of human eye iris at the beginning of the circle detected, and by this circle
Center elect human eye center as;
Step 3: utilize the human eye canthus point and human eye center detected in training set, uses and divides based on sparse and low-rank
Training set sample is entered by (the Robust Alignment by Sparse and Low-rank Decomposition) algorithm solved
Row batch alignment;
Step 4: split the training set after batch alignment, first by canny edge detecting technology by people to be detected
Eye iris image is converted into the boundary graph of binaryzation, utilizes Hough transformation to find radius in boundary graph between pupil maximum radius
And a circle between iris maximum radius, the circle drawn is defined as iris external diameter circle, then again continues with utilization suddenly
Husband's conversion finds another circle in the iris external diameter circle region found out, and the circle obtained is defined as iris internal diameter circle, this
Region between iris external diameter circle and iris internal diameter circle just can be set to iris region by sample, thus realizes training set iris
Segmentation;
Step 5: the eye image of input is repeated step one and step 2, find the eye image of input canthus and
Eye image is rotated by human eye center, recycling canthus and the oculocentric line of people, makes this line adjust to horizontal level, finally
Image after adjusting is repeated step 4, it is achieved the human eye iris segmentation to input picture.
The invention has the beneficial effects as follows: of the present invention is a kind of human eye iris segmentation method, and the method is the most certainly
Move and have chosen based on sparse and low-rank decomposition algorithm two datum marks, utilize and realize training based on sparse and low-rank decomposition algorithm
Sample batch alignment, has good denoising effect, can remove the brightness flop in training set, ciliary in upper lower eyelid
Block;By test sample is rotated, test sample is made to align with training sample, the matter of this sample image not only improved
Amount, and the accuracy of follow-up iris identification can be improved.Therefore, during the method can be widely used for iris identification field.
On the basis of technique scheme, the present invention can also do following improvement.
Further, described human eye canthus detection utilizes Harris Corner Detection to carry out.Specifically comprise the following steps that
1). use canny operator to calculate the directional derivative of image, calculate respectively and horizontally and vertically to go up
Directional derivative;
2). calculate the coefficients correlation matrix of each pixel;
3). calculate the angle point value of each pixel;
4). find out maximum in all angle point values;
5). travel through all of angle point value, if the angle point value of detection pixel is more than 0.01 times of maximum angular point value, and
Specifying in neighborhood pixels region is maximum, just this pixel is labeled as angle point;
6). all angle points are traveled through, the leftmost position being in human eye according to human eye canthus, by abscissa
Little angle point is labeled as human eye canthus.
Further, described human eye Spot detection is based on canny rim detection and Hough transformation.Concrete steps are such as
Under:
1). picture to be split is carried out binaryzation based on threshold value, obtains binary image;
2). binary image Gaussian filter is carried out smothing filtering, uses canny operator to calculate filtering image ladder
The amplitude of degree and direction;
3). detect along the gradient direction calculated, the pixel not being local maximum is set to 0, i.e. to gradient side
To carrying out non-maxima suppression, obtain the edge of image only one of which pixel width;
4). choose two threshold values th1 and th2(), non-maxima suppression image is carried out process and obtains
Two width images.The gray value of 1 Grad of the image pixel less than th1 is set to 0, constant more than the pixel value of threshold value;Image 2
The gray value of the Grad pixel less than th2 is set to 0, constant more than the pixel value of threshold value.Image 2 is scanned, when running into
The pixel p of one non-zero gray scale (x, time y), follow the tracks of with p (x, y) is the contour line of starting point, until the terminal q of contour line (x,
y).With q in image 2 (x, y) some s (x, 8 adjacent domains y) that some position is corresponding in image under consideration 1.If at s, (x y) puts
(x, y) exists, then included in image 2, as r (x, y) point non-zero pixels s1 in 8 adjacent domains.With r, (x y) is
Starting point, repeats the scanning to image 2, until all cannot continue in image 1 and image 2.When complete to comprise p (x,
After the link of contour line y), it is labeled as accessing by this contour line.Continue image 2 is scanned, find next profile
Line.Repeat track, until can not find new contour line in image 2;
5). (a, b, r), from the beginning of first pixel of image 2, make a be equal to the horizontal stroke of current pixel to set up parameter space A
Coordinate, b is equal to the vertical coordinate of current pixel, and the initial value of r is set to predefined minima, finds in image 2 through so that (a b) is
The center of circle, r is the non-zero pixel number of radius, be saved in A (a, b, r) in.Then make r add 1, repeat to find step, and preserve phase
The result answered, until r is equal to predefined maximum.Complete currently so that (a, is b) after the circle in the center of circle is searched, and makes that (a b) is
The next point of current pixel, repeats step, until (a b) is last pixel of image, i.e. completes parameter space A
(a, b, assignment r);
6). determine the maximum of parameter space A, corresponding to maximum (a b) is the circular circle detected
The heart, is labeled as human eye center by this point;
7). re-execute step one and two, until all of training set sample has been made human eye canthus and center
Detection.
Further, the batch alignment of described training set sample is based on sparse and low-rank decomposition algorithm.Concrete steps are such as
Under:
1). according toMatrix D is carried out low-rank decomposition and obtains A and E.
Wherein, D represents and belongs to of a sort training set sample matrix to be split, and each column vector of matrix represents a sample, and
Its element is to be arranged in order by each pixel value in this samples pictures to form;After A represents training set batch alignment
Image pattern result, A is a matrix identical with D size, and each column vector represents the column vector in D at same position
Represent sample go dry and alignment after formed result;Represent a weight parameter, it is intended that for(N represents this
The number of sample in class training set).Representing a conversion to sample matrix D, wherein human eye canthus and 2, center are to use
Initialize, E is a matrix identical with A with D size, and each column vector of E represents the row in D at same position
Vector samples pictures in brightness flop and block;0 normal form of representing matrix E;
2). according to A, each column vector being reduced into a sheet by a sheet image, each pixel value of image is successively equal to working as prostatitis
Each element of vector, obtains the image pattern after batch alignment.
Further, the described process that the training set after batch alignment realizes human eye iris segmentation is first to use canny
Rim detection, then use Hough transformation first to detect a radius circle between maximum pupil and maximum iris radius,
It is defined as iris external diameter, further uses Hough change to carry out iris inner diameter measurement in the region in iris external diameter circle.
Accompanying drawing explanation
Fig. 1 is a kind of human eye iris segmentation method general flow chart that the present invention relates to;
Fig. 2 is the human eye canthus detection method flow chart of steps that the present invention relates to;
Fig. 3 is the human eye central point detection method flow chart of steps that the present invention relates to;
Fig. 4 is the use canny rim detection that the present invention relates to and Hough transformation realizes iris segmentation flow chart of steps.
Detailed description of the invention
Being described principle and the feature of the present invention below in conjunction with accompanying drawing, example is served only for explaining the present invention, and
Non-for limiting the scope of the present invention.
Fig. 1 is a kind of human eye iris segmentation method general flow chart that the present invention relates to;Fig. 2 is the people that the present invention relates to
Eye canthus detection method flow chart of steps;Fig. 3 is the human eye central point detection method flow chart of steps that the present invention relates to;Fig. 4 is
The use canny rim detection that the present invention relates to and Hough transformation realize iris segmentation flow chart of steps;By Fig. 1,2,3,4 institute
Showing, a kind of typical enforcement step of the present invention is as follows:
Image to be split is provided by user.
Step one: the training set human eye iris image sample of known classification is carried out location, human eye canthus: utilize Harris
Corner Detection Algorithm detects the angle point in eye image one by one, and these angle points carry out full search traversal, selects abscissa
Little Corner character is position, human eye canthus.Specifically comprise the following steps that
1). use canny operator to calculate the directional derivative of image, calculate respectively and horizontally and vertically to go up
Directional derivativeWith;
2). calculate the coefficients correlation matrix of each pixel, wherein
Represent and Gaussian templateDoing convolution, the variance of this template is 2, and neighborhood window size is 7*7;
3). calculate the angle point value of each pixel;Represent and seek square
Battle arrayDeterminant,Representing matrixMark,Represent weighted value, between 0.04 to 0.06;
4). find out all angle point valuesMiddle maximum;
5). travel through all of angle point value, if the angle point value of detection pixel is more than 0.01, and neighbouring specifying
Being maximum in pixel region, this pixel is just labeled as angle point, the size in neighborhood pixels region is 9*9;
6). all angle points are traveled through, the leftmost position being in human eye according to human eye canthus, by abscissa
Little angle point is labeled as human eye canthus.
Step 2: detect the human eye center of training set sample based on canny rim detection and Hough transformation.Concrete steps are such as
Under:
1). based on threshold value, image to be split being carried out binaryzation, obtains binary image, the size of threshold value is appointed as
20;
2). binary image Gaussian filter is carried out smothing filtering, uses canny operator to calculate filtering image ladder
The amplitude of degree and direction;
3). detect along the gradient direction calculated, the pixel not being local maximum is set to 0, i.e. to gradient side
To carrying out non-maxima suppression, obtain the edge of image only one of which pixel width;
4). choose two threshold values th1 and th2(), non-maxima suppression image is carried out process and obtains
Two width images.The gray value of 1 Grad of the image pixel less than th1 is set to 0, constant more than the pixel value of threshold value;Image 2
The gray value of the Grad pixel less than th2 is set to 0, constant more than the pixel value of threshold value.Image 2 is scanned, when running into
The pixel p of one non-zero gray scale (x, time y), follow the tracks of with p (x, y) is the contour line of starting point, until the terminal q of contour line (x,
y).With q in image 2 (x, y) some s (x, 8 adjacent domains y) that some position is corresponding in image under consideration 1.If at s, (x y) puts
(x, y) exists, then included in image 2, as r (x, y) point non-zero pixels s1 in 8 adjacent domains.With r, (x y) is
Starting point, repeats the scanning to image 2, until all cannot continue in image 1 and image 2.When complete to comprise p (x,
After the link of contour line y), it is labeled as accessing by this contour line.Continue image 2 is scanned, find next profile
Line.Repeat track, until can not find new contour line in image 2;
5). (a, b, r), from the beginning of first pixel of image 2, make a be equal to the horizontal stroke of current pixel to set up parameter space A
Coordinate, b is equal to the vertical coordinate of current pixel, and the initial value of r is set to predefined minima, finds in image 2 through so that (a b) is
The center of circle, r is the non-zero pixel number of radius, be saved in A (a, b, r) in.Then make r add 1, repeat to find step, and preserve phase
The result answered, until r is equal to predefined maximum.Complete currently so that (a, is b) after the circle in the center of circle is searched, and makes that (a b) is
The next point of current pixel, repeats step, until (a b) is last pixel of image, i.e. completes parameter space A
(a, b, assignment r);
6). determine the maximum of parameter space A, corresponding to maximum (a b) is the circular circle detected
The heart, is labeled as human eye center by this point;
7). re-execute step one and two, until all of training set sample has been made human eye canthus and center
Detection.
Step 3: utilize human eye canthus and 2, center as initial transformation reference point, use and divide based on sparse and low-rank
The algorithm (Robust Alignment by Sparse and Low-rank Decomposition, RASL) solved carries out batch
Alignment.Specifically comprise the following steps that
1). according toMatrix D is carried out low-rank decomposition and obtains A and E.D
Representing and belong to of a sort training set sample matrix to be split, each column vector of matrix represents a sample, and its element
It is to be arranged in order by each pixel value in this samples pictures to form.A represents decent to the figure after training set batch alignment
This result, A is a matrix identical with D size, and each column vector represents in D what the column vector at same position represented
Sample go dry and alignment after formed result.Represent a weight parameter, it is intended that for(N represents that this class is trained
Concentrate the number of sample).Representing a conversion to sample matrix D, wherein human eye canthus and 2, center are used to initially
Change, E is a matrix identical with A with D size, and each column vector of E represents the column vector in D at same position
Brightness flop in samples pictures and blocking,0 normal form of representing matrix E;
2). according to A, each column vector is reduced into a sheet by a sheet image, each pixel value of image successively equal to when prostatitis to
Each element of amount, obtains the image pattern after batch alignment.
Step 4: the image after batch alignment is realized human eye iris segmentation.Specifically comprise the following steps that
1). the sample after batch alignment is used the 2 of step 2), 3), 4), 5) obtain the edge graph of binaryzation;
2). determine the maximum of parameter space A, show that (a, b, r) corresponding circle, this circle is outside human eye iris parameter
Footpath;
3). determine iris external diameter, in the region of iris external diameter, repeat the 5 of step 2);
4). determine the maximum of parameter space A, show that (a, b, r) corresponding circle, this circle is in human eye iris parameter
Footpath, is set to iris region by the region between iris external diameter and internal diameter, thus can realize the segmentation of human eye iris;
5). repeat 1), 2), 3), 4), 5), until all of training sample all segments.
Step 5: the eye image to be identified (test sample) of input is repeated step one and step 2, finds input
Eye image is rotated by the canthus of eye image and human eye center, recycling canthus and the oculocentric line of people, makes this line adjust
Whole to horizontal level, so that the eye image of input aligns with training set, finally the image after adjusting is repeated step 4, real
The now human eye iris segmentation to input picture.
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all spirit in the present invention and
Within principle, any modification, equivalent substitution and improvement etc. made, should be included within the scope of the present invention.
Claims (7)
1. a human eye iris segmentation method, it is characterised in that the method specifically comprises the following steps that
Step one: training set eye image sample is carried out human eye canthus location one by one, utilizes Harris Corner Detection Algorithm to examine
Surveying the angle point in eye image, then these angle points carry out retrieval traversal, the point selecting abscissa minimum is human eye canthus;
Step 2: training set human eye iris image sample is positioned human eye center one by one, first by threshold value to samples pictures to be detected
Carry out binaryzation, then utilize canny edge detection method that binary image is carried out rim detection and obtain the border of binaryzation
Figure, recycling Hough transformation is round to boundary graph detection, is set to the inner edge of human eye iris at the beginning of the circle detected, and by this circle
The heart elects human eye center as;
Step 3: utilize the human eye canthus point and human eye center detected in training set, uses based on sparse and low-rank decomposition
(Robust Alignment by Sparse and Low-rank Decomposition) algorithm is to training set human eye iris figure
Decent carries out batch alignment;
Further, described specifically comprise the following steps that based on sparse and low-rank decomposition training set sample batch alignment algorithm
1) basisMatrix D is carried out low-rank decomposition and obtains A and E, wherein, D table
Showing and belong to of a sort training set sample matrix to be split, each column vector of matrix represents a sample, and its element is
It is arranged in order by each pixel value in this samples pictures and forms;A represents the image pattern after training set batch alignment
As a result, A is a matrix identical with D size, and each column vector represents in D the sample that the column vector at same position represents
Originally go dry and alignment after formed result;γ represents a weight parameter, it is intended that for(N represents this class training set
The number of middle sample);τ represents one to sample matrix D conversion, and wherein human eye canthus and 2, center are used to initialize
τ, E are matrixes identical with A with D size, and each column vector of E represents the sample of the column vector in D at same position
Brightness flop in this picture and blocking;0 normal form of | | E | | 0 representing matrix E;
2) according to A, each column vector being reduced into a sheet by a sheet image, each pixel value of image is equal to the every of current column vector successively
One element, obtains the image pattern after batch alignment;
Step 4: the training set human eye iris image after batch alignment is split, first will treat by Canny edge detecting technology
The human eye iris image of detection is converted into the boundary graph of binaryzation, utilizes Hough transformation to find radius in boundary graph between pupil
A circle between maximum radius and iris maximum radius, is defined as the circle drawn iris external diameter circle, the most again continues
Continue and utilize Hough transformation to find another circle in the iris external diameter circle region found out, the circle obtained is defined as in iris
Footpath circle, thus can be set to iris region by the region between iris external diameter circle and iris internal diameter circle, thus realize training
The segmentation of iris in collection image;
Step 5: the eye image of input is repeated step one and step 2, finds canthus and the human eye of the eye image of input
Eye image is rotated by center, recycling canthus and the oculocentric line of people, makes this line adjust to horizontal level, finally exchanges
Image after whole repeats step 4, it is achieved the human eye iris segmentation to input picture.
A kind of human eye iris segmentation method, it is characterised in that the location at human eye canthus is profit
It is image to be in leftmost positional information carry out with human eye canthus.
A kind of human eye iris segmentation method, it is characterised in that the location at human eye canthus be by
Angle point is carried out, and the angle point detected is carried out full search traversal, when the abscissa of angle point to be compared is less than the angle detected
During the abscissa put, current angle point being labeled as position, human eye canthus, traversal is until all angle points have traveled through entirely successively.
A kind of human eye iris segmentation method, it is characterised in that described based on sparse and low
Two datum marks in order decomposition algorithm be calculated by the step one described in claim 1 and step 2 rather than hands
Dynamic choose.
A kind of human eye iris segmentation method, it is characterised in that examine utilizing canny edge
Survey before carrying out human eye iris internal-and external diameter loop truss with Hough transformation, utilize and be trained based on sparse and low-rank decomposition algorithm
The batch alignment of sample, eliminates the ciliary interference blocking factor in brightness flop, upper lower eyelid, reduces sample to be split
Brightness in Ben and block the interference of unfavorable factor, improves the quality of training set sample.
A kind of human eye iris segmentation method, it is characterised in that training set sample is used base
Human eye position in image pattern can be ajusted in sparse and low-rank decomposition algorithm, thus reduce in image acquisition process
The interference that collecting device is inconsistent relative to acquisition target position.
A kind of human eye iris segmentation method, it is characterised in that test sample is revolved
Turn, human eye position in image pattern in test sample can be ajusted, so that test sample is alignd with training set sample, carry
The high quality of test sample.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310570892.4A CN103632137B (en) | 2013-11-15 | 2013-11-15 | A kind of human eye iris segmentation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310570892.4A CN103632137B (en) | 2013-11-15 | 2013-11-15 | A kind of human eye iris segmentation method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103632137A CN103632137A (en) | 2014-03-12 |
CN103632137B true CN103632137B (en) | 2016-08-24 |
Family
ID=50213166
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310570892.4A Expired - Fee Related CN103632137B (en) | 2013-11-15 | 2013-11-15 | A kind of human eye iris segmentation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103632137B (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104573660A (en) * | 2015-01-13 | 2015-04-29 | 青岛大学 | Method for precisely positioning human eyes by SIFT point descriptor |
RU2016138608A (en) | 2016-09-29 | 2018-03-30 | Мэджик Лип, Инк. | NEURAL NETWORK FOR SEGMENTING THE EYE IMAGE AND ASSESSING THE QUALITY OF THE IMAGE |
CN109844809B (en) * | 2017-11-30 | 2022-04-15 | 深圳配天智能技术研究院有限公司 | Image processing method and device and computer readable storage medium |
CN109165306B (en) * | 2018-08-09 | 2021-11-23 | 长沙理工大学 | Image retrieval method based on multitask Hash learning |
CN109376649A (en) * | 2018-10-20 | 2019-02-22 | 张彦龙 | A method of likelihood figure, which is reduced, from eye gray level image calculates the upper lower eyelid of identification |
CN110619273B (en) * | 2019-08-14 | 2023-10-31 | 张杰辉 | Efficient iris recognition method and recognition device |
CN110781745B (en) * | 2019-09-23 | 2022-02-11 | 杭州电子科技大学 | Tail eyelash detection method based on composite window and gradient weighted direction filtering |
CN111144413A (en) * | 2019-12-30 | 2020-05-12 | 福建天晴数码有限公司 | Iris positioning method and computer readable storage medium |
CN111241951B (en) * | 2020-01-03 | 2023-10-31 | 张杰辉 | Iris image processing method and device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1900951A (en) * | 2006-06-02 | 2007-01-24 | 哈尔滨工业大学 | Iris image flexible specification method based on mathematical morphology |
CN101059836A (en) * | 2007-06-01 | 2007-10-24 | 华南理工大学 | Human eye positioning and human eye state recognition method |
CN101901472A (en) * | 2010-07-07 | 2010-12-01 | 清华大学 | Method for aligning non-rigid robust batch images based on matrix rank minimization |
CN101923645A (en) * | 2009-06-09 | 2010-12-22 | 黑龙江大学 | Iris splitting method suitable for low-quality iris image in complex application context |
CN102332098A (en) * | 2011-06-15 | 2012-01-25 | 夏东 | Method for pre-processing iris image |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4498224B2 (en) * | 2005-06-14 | 2010-07-07 | キヤノン株式会社 | Image processing apparatus and method |
-
2013
- 2013-11-15 CN CN201310570892.4A patent/CN103632137B/en not_active Expired - Fee Related
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1900951A (en) * | 2006-06-02 | 2007-01-24 | 哈尔滨工业大学 | Iris image flexible specification method based on mathematical morphology |
CN101059836A (en) * | 2007-06-01 | 2007-10-24 | 华南理工大学 | Human eye positioning and human eye state recognition method |
CN101923645A (en) * | 2009-06-09 | 2010-12-22 | 黑龙江大学 | Iris splitting method suitable for low-quality iris image in complex application context |
CN101901472A (en) * | 2010-07-07 | 2010-12-01 | 清华大学 | Method for aligning non-rigid robust batch images based on matrix rank minimization |
CN102332098A (en) * | 2011-06-15 | 2012-01-25 | 夏东 | Method for pre-processing iris image |
Non-Patent Citations (1)
Title |
---|
基于肤色检测和人眼定位的人脸检测方法;徐锋 等;《计算机***应用》;20101231;第19卷(第2期);177-179 * |
Also Published As
Publication number | Publication date |
---|---|
CN103632137A (en) | 2014-03-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103632137B (en) | A kind of human eye iris segmentation method | |
CN105335725B (en) | A kind of Gait Recognition identity identifying method based on Fusion Features | |
CN102902967B (en) | Method for positioning iris and pupil based on eye structure classification | |
CN104091147B (en) | A kind of near-infrared eyes positioning and eye state identification method | |
CN103218605B (en) | A kind of fast human-eye positioning method based on integral projection and rim detection | |
CN109285179A (en) | A kind of motion target tracking method based on multi-feature fusion | |
CN103093215A (en) | Eye location method and device | |
CN104835175B (en) | Object detection method in a kind of nuclear environment of view-based access control model attention mechanism | |
CN107122737A (en) | A kind of road signs automatic detection recognition methods | |
CN106156712A (en) | A kind of based on the ID (identity number) card No. recognition methods under natural scene and device | |
CN107464252A (en) | A kind of visible ray based on composite character and infrared heterologous image-recognizing method | |
CN105512618B (en) | Video tracing method | |
CN103295016A (en) | Behavior recognition method based on depth and RGB information and multi-scale and multidirectional rank and level characteristics | |
CN104134200A (en) | Mobile scene image splicing method based on improved weighted fusion | |
CN110443128A (en) | One kind being based on SURF characteristic point accurately matched finger vein identification method | |
CN106682641A (en) | Pedestrian identification method based on image with FHOG- LBPH feature | |
CN103136519A (en) | Sight tracking and positioning method based on iris recognition | |
CN103425970A (en) | Human-computer interaction method based on head postures | |
CN109978848A (en) | Method based on hard exudate in multiple light courcess color constancy model inspection eye fundus image | |
CN112329656B (en) | Feature extraction method for human action key frame in video stream | |
CN104268520A (en) | Human motion recognition method based on depth movement trail | |
CN106295532A (en) | A kind of human motion recognition method in video image | |
CN105225216A (en) | Based on the Iris preprocessing algorithm of space apart from circle mark rim detection | |
Li | An Iris recognition algorithm based on coarse and fine location | |
CN103198319A (en) | Method of extraction of corner of blurred image in mine shaft environment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20160824 Termination date: 20171115 |
|
CF01 | Termination of patent right due to non-payment of annual fee |