[go: up one dir, main page]

CN102999751A - Scale-invariant feature transform (SIFT) feature based method for identifying eyebrows - Google Patents

Scale-invariant feature transform (SIFT) feature based method for identifying eyebrows Download PDF

Info

Publication number
CN102999751A
CN102999751A CN201310003415XA CN201310003415A CN102999751A CN 102999751 A CN102999751 A CN 102999751A CN 201310003415X A CN201310003415X A CN 201310003415XA CN 201310003415 A CN201310003415 A CN 201310003415A CN 102999751 A CN102999751 A CN 102999751A
Authority
CN
China
Prior art keywords
eyebrow
similarity
region
sub
regions
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
CN201310003415XA
Other languages
Chinese (zh)
Inventor
曹杰
许野平
方亮
刘辰飞
张传峰
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.)
SHANDONG SYNTHESIS ELECTRONIC TECHNOLOGY Co Ltd
Original Assignee
SHANDONG SYNTHESIS ELECTRONIC 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 SHANDONG SYNTHESIS ELECTRONIC TECHNOLOGY Co Ltd filed Critical SHANDONG SYNTHESIS ELECTRONIC TECHNOLOGY Co Ltd
Priority to CN201310003415XA priority Critical patent/CN102999751A/en
Publication of CN102999751A publication Critical patent/CN102999751A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Image Analysis (AREA)

Abstract

The invention discloses an SIFT feature based method for identifying eyebrows. The method comprises the steps of a). collecting face pictures; b). selecting eyebrow regions A; c). dividing eyebrow regions into M subregions; d). obtaining SIFT feature matrixes A<j>(m<j>, w) of subregions; e). obtaining SIFT feature matrixes A1<j>(m<j>, w) and A2<j>(n<j>, w) of eyebrow regions of two pictures, and calculating similarities S<j> of corresponding subregions; f). conducting statistics on probability distribution that whether eyebrows belong to the same person or not; and g). obtaining P(same) according to a Bayes formula to judge whether two comparison pictures originate from the same person. The method for identifying eyebrows is characterized in that rotation and scale changes of pictures are kept unchanged. SIFT features are used for identifying eyebrows, the influence of illumination and posture changes can be effectively reduced, and no human is involved.

Description

Eyebrow identification method based on SIFT features
Technical Field
The invention relates to an eyebrow identification method based on SIFT (scale invariant feature transform) features, and relates to the field of digital image processing application.
Background
Eyebrows are important features in face recognition, and compared with other features of a face, eyebrows have better stability and difference, but are generally influenced by factors such as illumination, posture and the like.
The patent "identity authentication method based on eyebrow identification" (publication number: 1645406) uses the difference of RGB color components of each pixel of eyebrow as the identification basis. The method is greatly influenced by illumination and posture, and particularly under the condition of sidelight, the recognition effect is poor.
The patent eyebrow image identification method based on subregion matching (publication number: 101901353A) manually selects a pure eyebrow image of each user as a template of each user, uses the eyebrow image to be identified to perform convolution operation with the stored eyebrow template in sequence, and identifies after obtaining similarity. Although the method can reduce partial influences of illumination and posture, the process of manually selecting the pure eyebrow image is very complicated and tedious.
Disclosure of Invention
In order to overcome the defects of the technical problems, the invention provides the eyebrow identification method based on the SIFT features, which can effectively reduce the influence of illumination and posture change.
The eyebrow identification method based on SIFT features is characterized by comprising the following steps of: a) collect the face photo, compile the volume asThe face photo library of (1), whereinIs the number of people in the face library,
Figure 201310003415X100002DEST_PATH_IMAGE006
is shown as
Figure 201310003415X100002DEST_PATH_IMAGE008
Face picture of person under different shooting conditions, wherein 0 <
Figure 73162DEST_PATH_IMAGE008
≤(ii) a b) Selecting an eyebrow area
Figure 201310003415X100002DEST_PATH_IMAGE010
Selecting a face photo
Figure 201310003415X100002DEST_PATH_IMAGE012
Of the eyebrow area
Figure 583089DEST_PATH_IMAGE010
Using the face image as a region for calculating the similarity of the face image; c) dividing the area of eyebrow into sub-areasIs divided into
Figure 201310003415X100002DEST_PATH_IMAGE014
Sub-regions, for sub-regions
Figure 201310003415X100002DEST_PATH_IMAGE016
Represents; the intersection between different subregions may or may not be empty;signFirst of the region
Figure 201310003415X100002DEST_PATH_IMAGE018
Sub-region, 0 <
Figure 445948DEST_PATH_IMAGE018
≤
Figure 644848DEST_PATH_IMAGE014
(ii) a d) Acquiring an SIFT feature matrix, and acquiring an eyebrow region by using an SIFT algorithm
Figure 251410DEST_PATH_IMAGE010
Is/are as followsSIFT feature matrix of individual sub-regions
Figure 201310003415X100002DEST_PATH_IMAGE020
，The first representing the area AThe number of feature points extracted by the sub-regions,
Figure 201310003415X100002DEST_PATH_IMAGE024
the dimension of SIFT feature matrix is represented;
Figure 362214DEST_PATH_IMAGE020
indicating the number of lines
Figure 823282DEST_PATH_IMAGE022
The number of rows is
Figure 190810DEST_PATH_IMAGE024
A matrix of (a); e) calculating the similarity of the two face photos, and aiming at the two face photos
Figure 201310003415X100002DEST_PATH_IMAGE026
Andrespectively obtaining SIFT feature matrixes according to the steps b), c) and d)
Figure 201310003415X100002DEST_PATH_IMAGE030
And
Figure 201310003415X100002DEST_PATH_IMAGE032
(ii) a Calculating a matrix of corresponding sub-regions
Figure 744020DEST_PATH_IMAGE030
And
Figure 956826DEST_PATH_IMAGE032
the similarity between any two lines and all the similarity values form a matrix
Figure 201310003415X100002DEST_PATH_IMAGE034
And define，
Figure 201310003415X100002DEST_PATH_IMAGE038
Presentation photograph
Figure 965409DEST_PATH_IMAGE026
And
Figure 612160DEST_PATH_IMAGE028
similarity corresponding to jth sub-region, and its size is taken as matrixMaximum value of all elements in; f) counting the probability distribution of similarity to make two pictures of eyebrow regionThe similarity of the sub-regions is a sample, and the similarity of each sub-region of the eyebrow is counted
Figure 170071DEST_PATH_IMAGE038
Probability distribution when the eyebrows of the same person are the same and when the eyebrows of the different persons are not the same; g) calculating the similarity probability of the eyebrow regions of the two compared images
Figure 201310003415X100002DEST_PATH_IMAGE040
Calculating two images to be compared according to steps b), c), d) and e)
Figure 754767DEST_PATH_IMAGE014
Similarity of individual sub-regions、
Figure 705406DEST_PATH_IMAGE042
、…、
Figure 24129DEST_PATH_IMAGE038
Obtaining by probability distribution in step f
Figure 201310003415X100002DEST_PATH_IMAGE044
And，indicates to correspond to
Figure 201310003415X100002DEST_PATH_IMAGE048
When the sub-regions are eyebrows of the same person, the similarity isA probability value of (d);
Figure 448441DEST_PATH_IMAGE046
indicates to correspond toWhen the sub-regions are not eyebrows of the same person, the similarity is
Figure 341628DEST_PATH_IMAGE038
0 < (r) >, a probability value of
Figure 766662DEST_PATH_IMAGE018
≤
Figure 629576DEST_PATH_IMAGE014
(ii) a Obtaining the probability that the two comparison images are the eyebrows of the same person under each similarity according to a Bayes formula:
Figure 201310003415X100002DEST_PATH_IMAGE050
Figure 201310003415X100002DEST_PATH_IMAGE052
；
and judging whether the two contrast images come from the same person or not according to the probability value of the obtained two contrast images being the eyebrows of the same person.
In step e), the corresponding sub-regions are the eyebrow regions of the two images according to the same sub-region division method, and only the corresponding parts obtained have the meaning and need of calculating the similarity. In this step, since
Figure 117250DEST_PATH_IMAGE036
Calculating
Figure 116430DEST_PATH_IMAGE030
Andmatrix for similarity acquisition between any two rows
Figure 127822DEST_PATH_IMAGE034
In (2), no matter how the elements are arranged, the similarity is not influenced
Figure 529722DEST_PATH_IMAGE038
The magnitude of the value. In step f), the similarity is taken as a sample to obtainProbability distribution when the eyebrows of the same person are the same and when the eyebrows of the different persons are not the same; thus, in step g), the probability distribution function obtained in step f) can be used to calculate
Figure 838661DEST_PATH_IMAGE044
And
Figure 676167DEST_PATH_IMAGE046
the numerical value of (c). A threshold value can be set in the step g), and when the similarity probability value of the eyebrows of the same person of the two acquired comparison images is greater than the set threshold value, the two images are considered to be from the same person; and if the difference is smaller than the set threshold value, the two images are not considered to be the images of the same person.
SIF-based method of the inventionMethod for identifying eyebrows with T characteristics, eyebrow area selected in step b)
Figure 179961DEST_PATH_IMAGE010
Setting selected eyebrow regions for the left eyebrow region, the right eyebrow region or the whole eyebrow regionRespectively has a width and a height of
Figure 201310003415X100002DEST_PATH_IMAGE054
、Which comprises the following steps: b-1) positioning the pupil position, and positioning the face photo by using the pupil positioning method
Figure 219034DEST_PATH_IMAGE012
The pupil position of (1) is the connecting line between two pupils
Figure 201310003415X100002DEST_PATH_IMAGE058
The axes establish a plane rectangular coordinate system, and the coordinates of the left pupil and the right pupil are respectively set as、(ii) a b-2) calculating the interpupillary distance according to a formulaCalculating the distance between two pupils; b-3) if region
Figure 164076DEST_PATH_IMAGE010
When the area is the left eyebrow area, the selected eyebrow areaWidth of (2)
Figure 201310003415X100002DEST_PATH_IMAGE066
Height of，
Figure 611423DEST_PATH_IMAGE010
The coordinates of the center point of the region are(ii) a If region
Figure 346161DEST_PATH_IMAGE010
When the right eyebrow area is selected, the selected eyebrow area
Figure 408794DEST_PATH_IMAGE010
Width of (2)
Figure 21434DEST_PATH_IMAGE066
Height of
Figure 32115DEST_PATH_IMAGE068
，
Figure 937754DEST_PATH_IMAGE010
The coordinates of the center point of the region are
Figure 201310003415X100002DEST_PATH_IMAGE072
(ii) a If region
Figure 220837DEST_PATH_IMAGE010
When the eyebrow area is all, the selected eyebrow area
Figure 932441DEST_PATH_IMAGE010
Width of (2)
Figure 201310003415X100002DEST_PATH_IMAGE074
Height of
Figure 118002DEST_PATH_IMAGE068
，The coordinates of the center point of the region are
Figure 201310003415X100002DEST_PATH_IMAGE076
(ii) a Wherein,
Figure 201310003415X100002DEST_PATH_IMAGE078
、
Figure 201310003415X100002DEST_PATH_IMAGE080
、
Figure 201310003415X100002DEST_PATH_IMAGE082
、
Figure 201310003415X100002DEST_PATH_IMAGE084
are all constants.
Wherein,
Figure 843217DEST_PATH_IMAGE078
=0.625、=0.391、
Figure 953573DEST_PATH_IMAGE082
=1.563、
Figure 263332DEST_PATH_IMAGE084
=0.281。
in the eyebrow identification method based on SIFT features, in the step e), the matrix of the corresponding sub-region
Figure 201310003415X100002DEST_PATH_IMAGE086
And
Figure 201310003415X100002DEST_PATH_IMAGE088
the method for calculating the similarity between any two lines is Euclidean distance, Mahalanobis distance or vector inner product algorithm. Euclidean distance, Mahalanobis distance or vector inner product are all the existing methods for calculating similarity.
According to the eyebrow identification method based on SIFT features, probability distribution of statistical similarity in step f) is a discrete model or a continuous model; under the condition of a discrete model, the similarity is obtained by counting the probability value of each similarity value falling in each numerical value interval; in the case of a continuous model, a probability density function is formed by fitting using Gaussian mixture probability modeling.
The eyebrow identification method based on SIFT features, provided by the invention, comprises the step a) of capacity identification
Figure 201310003415X100002DEST_PATH_IMAGE090
=2000，
Figure 583323DEST_PATH_IMAGE004
=200，
Figure 636730DEST_PATH_IMAGE006
= 10; the different shooting conditions refer to different postures and different illumination.
The invention has the beneficial effects that: the eyebrow recognition method based on SIFT features comprises the steps of firstly establishing a face photo library of a plurality of people under different shooting conditions, dividing an eyebrow area into sub-areas, calculating the similarity of the sub-areas of every two photos by adopting an SIFT feature matrix, and then establishing probability distribution of eyebrows of the same person and eyebrows of people which are not the same by taking the similarity as a sample; and finally, obtaining the probability that the two compared images are the eyebrows of the same person according to a Bayesian formula, and judging whether the two images are from the same person or not according to the probability.
The scale invariant feature transform SIFT algorithm is suitable for feature description and feature matching of rigid objects, and has the characteristic of keeping the rotation and scale variation of images invariant. The SIFT feature is used for eyebrow recognition, the influence of illumination and posture change can be effectively reduced, and manual participation is not needed.
Drawings
FIG. 1 shows the similarity between the subregions of the eyebrows
Figure 915658DEST_PATH_IMAGE038
Probability distribution when the eyebrows are the same person;
FIG. 2 shows the similarity between the sub-regions of the eyebrows
Figure 334001DEST_PATH_IMAGE038
Probability distribution when not the same person is eyebrow.
Detailed Description
The invention is further described with reference to the following figures and examples.
The eyebrow identification method based on SIFT features comprises the following steps:
a) collect the face photo, compile the volume as
Figure DEST_PATH_IMAGE091
The face photo library of (1), wherein
Figure 875710DEST_PATH_IMAGE004
Is the number of people in the face library,
Figure 670490DEST_PATH_IMAGE006
is shown as
Figure 364777DEST_PATH_IMAGE008
Face picture of person under different shooting conditions, wherein 0 <
Figure 954021DEST_PATH_IMAGE008
≤
Figure 364450DEST_PATH_IMAGE004
；
Capacity in step a)
Figure 388687DEST_PATH_IMAGE090
It may be selected to be 2000 a,
Figure 388347DEST_PATH_IMAGE004
=200，
Figure 476389DEST_PATH_IMAGE006
= 10; the different shooting conditions refer to different postures and different illumination;
b) selecting an eyebrow area
Figure 681105DEST_PATH_IMAGE010
Selecting a face photo
Figure 129273DEST_PATH_IMAGE012
Of the eyebrow areaUsing the face image as a region for calculating the similarity of the face image;
the eyebrow area selected in step b)
Figure 791515DEST_PATH_IMAGE010
Can be a left eyebrow area, a right eyebrow area or all eyebrow areas and is provided with a selected eyebrow area
Figure 483528DEST_PATH_IMAGE010
Respectively has a width and a height of
Figure 535054DEST_PATH_IMAGE054
、
Figure 996122DEST_PATH_IMAGE056
Which comprises the following steps:
b-1) positioning the pupil position, and positioning the face photo by using the pupil positioning method
Figure 425967DEST_PATH_IMAGE012
The pupil position of (1) is the connecting line between two pupils
Figure 588964DEST_PATH_IMAGE058
The axes establish a plane rectangular coordinate system, and the coordinates of the left pupil and the right pupil are respectively set as
Figure DEST_PATH_IMAGE092
、
Figure DEST_PATH_IMAGE062A
；
b-2) calculating the interpupillary distance according to a formula
Figure DEST_PATH_IMAGE064A
Calculating the distance between two pupils;
b-3) if region
Figure 194913DEST_PATH_IMAGE010
When the area is the left eyebrow area, the selected eyebrow area
Figure 776067DEST_PATH_IMAGE010
Width of (2)
Figure 111233DEST_PATH_IMAGE066
Height of，
Figure 346399DEST_PATH_IMAGE010
The coordinates of the center point of the region are
Figure 657425DEST_PATH_IMAGE070
(ii) a If regionWhen the right eyebrow area is selected, the selected eyebrow areaWidth of (2)
Figure 698434DEST_PATH_IMAGE066
Height of
Figure 178482DEST_PATH_IMAGE068
，
Figure 121030DEST_PATH_IMAGE010
The coordinates of the center point of the region are(ii) a If region
Figure 917265DEST_PATH_IMAGE010
When the eyebrow area is all, the selected eyebrow areaWidth of (2)Height of
Figure 851351DEST_PATH_IMAGE068
，
Figure 75659DEST_PATH_IMAGE010
The coordinates of the center point of the region are
Figure 278101DEST_PATH_IMAGE076
；
Wherein,
Figure 562452DEST_PATH_IMAGE078
、
Figure 974979DEST_PATH_IMAGE080
、
Figure 675081DEST_PATH_IMAGE082
、
Figure 292882DEST_PATH_IMAGE084
are all constants; in particular=0.625、
Figure 585640DEST_PATH_IMAGE080
=0.391、
Figure 89434DEST_PATH_IMAGE082
=1.563、
Figure 63206DEST_PATH_IMAGE084
=0.281。
c) Dividing the area of eyebrow into sub-areas
Figure 128508DEST_PATH_IMAGE010
Is divided into
Figure 515627DEST_PATH_IMAGE014
Sub-regions, for sub-regions
Figure 557532DEST_PATH_IMAGE016
Represents; the intersection between different subregions may or may not be empty;
Figure 651390DEST_PATH_IMAGE016
signFirst of the region
Figure 994964DEST_PATH_IMAGE018
Sub-region, 0 <
Figure 604674DEST_PATH_IMAGE018
≤
Figure 490722DEST_PATH_IMAGE014
；
d) Acquiring an SIFT feature matrix, and acquiring an eyebrow region by using an SIFT algorithmIs/are as follows
Figure 554810DEST_PATH_IMAGE014
SIFT feature matrix of individual sub-regions
Figure DEST_PATH_IMAGE093
，The first representing the area A
Figure 457835DEST_PATH_IMAGE018
The number of feature points extracted by the sub-regions,the dimension of SIFT feature matrix is represented;indicating the number of lines
Figure 305202DEST_PATH_IMAGE022
The number of rows is
Figure 726694DEST_PATH_IMAGE024
A matrix of (a);
e) calculating the similarity of the two face photos, and aiming at the two face photos
Figure 36452DEST_PATH_IMAGE026
And
Figure 44860DEST_PATH_IMAGE028
respectively obtaining SIFT feature matrixes according to the steps b), c) and d)
Figure DEST_PATH_IMAGE094
And
Figure DEST_PATH_IMAGE095
(ii) a Calculating a matrix of corresponding sub-regions
Figure 973633DEST_PATH_IMAGE094
And
Figure 813413DEST_PATH_IMAGE095
the similarity between any two lines and all the similarity values form a matrix
Figure DEST_PATH_IMAGE096
And define
Figure DEST_PATH_IMAGE097
，
Figure 608587DEST_PATH_IMAGE038
Presentation photograph
Figure 274929DEST_PATH_IMAGE026
And
Figure 7393DEST_PATH_IMAGE028
similarity corresponding to jth sub-region, and its size is taken as matrixMaximum value of all elements in;
in this step e), a matrix of corresponding sub-regions
Figure 290924DEST_PATH_IMAGE086
And
Figure 336240DEST_PATH_IMAGE088
the method for calculating the similarity between any two lines is an Euclidean distance algorithm, a Mahalanobis distance algorithm or a vector inner product algorithm;
f) statistical similarityProbability distribution of degree by two pictures of eyebrow areaThe similarity of the sub-regions is a sample, and the probability distribution of the similarity of each sub-region of the eyebrows when the similarity is the eyebrows of the same person and the similarity is not the eyebrows of the same person is counted;
in the step f), the probability distribution of the statistical similarity is a discrete model or a continuous model; under the condition of a discrete model, the similarity is obtained by counting the probability value of each similarity value falling in each numerical value interval; in the case of a continuous model, modeling by using Gaussian mixture probability, and fitting to form a probability density function; as shown in FIG. 1 and FIG. 2, the similarity is given
Figure 594757DEST_PATH_IMAGE038
Fitting probability distribution when the eyebrows of the same person are the eyebrows of the same person and when the eyebrows of the different person are not the eyebrows of the same person to form a probability density function image;
g) calculating the similarity probability of the eyebrow regions of the two compared images
Figure DEST_PATH_IMAGE098
Calculating two images to be compared according to steps b), c), d) and e)
Figure 620481DEST_PATH_IMAGE014
Similarity of individual sub-regions
Figure 323733DEST_PATH_IMAGE042
、
Figure 257054DEST_PATH_IMAGE042
、…、Obtaining by probability distribution in step f
Figure DEST_PATH_IMAGE100
And
Figure DEST_PATH_IMAGE102
，
Figure 437073DEST_PATH_IMAGE100
indicates to correspond to
Figure 801189DEST_PATH_IMAGE048
When the sub-regions are eyebrows of the same person, the similarity is
Figure 741463DEST_PATH_IMAGE038
A probability value of (d);
Figure 701067DEST_PATH_IMAGE102
indicates to correspond toWhen the sub-regions are not eyebrows of the same person, the similarity is
Figure 920007DEST_PATH_IMAGE038
0 < (r) >, a probability value of
Figure 460710DEST_PATH_IMAGE018
≤
Figure 141397DEST_PATH_IMAGE014
(ii) a Obtaining the probability that the two comparison images are the eyebrows of the same person under each similarity according to a Bayes formula:
Figure DEST_PATH_IMAGE103
Figure DEST_PATH_IMAGE105
；
and judging whether the two contrast images come from the same person or not according to the probability value of the obtained two contrast images being the eyebrows of the same person.

Claims (6)

1. An eyebrow identification method based on SIFT features is characterized by comprising the following steps:
a) collect the face photo, compile the volume as
Figure 201310003415X100001DEST_PATH_IMAGE002
The face photo library of (1), wherein
Figure 201310003415X100001DEST_PATH_IMAGE004
Is the number of people in the face library,
Figure 201310003415X100001DEST_PATH_IMAGE006
is shown asFace picture of person under different shooting conditions, wherein 0 <
Figure 407197DEST_PATH_IMAGE008
≤
Figure 808222DEST_PATH_IMAGE004
；
b) Selecting an eyebrow area
Figure 201310003415X100001DEST_PATH_IMAGE010
Selecting a face photo
Figure 201310003415X100001DEST_PATH_IMAGE012
Of the eyebrow area
Figure 467130DEST_PATH_IMAGE010
Using the face image as a region for calculating the similarity of the face image;
c) dividing the area of eyebrow into sub-areas
Figure 965107DEST_PATH_IMAGE010
Is divided into
Figure 201310003415X100001DEST_PATH_IMAGE014
Sub-regions, for sub-regions
Figure 201310003415X100001DEST_PATH_IMAGE016
Represents; the intersection between different subregions may or may not be empty;
Figure 782760DEST_PATH_IMAGE016
sign
Figure 936660DEST_PATH_IMAGE010
First of the region
Figure 201310003415X100001DEST_PATH_IMAGE018
Sub-region, 0 <≤；
d) Acquiring an SIFT feature matrix, and acquiring an eyebrow region by using an SIFT algorithm
Figure 723985DEST_PATH_IMAGE010
Is/are as follows
Figure 161919DEST_PATH_IMAGE014
SIFT feature matrix of individual sub-regions
Figure 201310003415X100001DEST_PATH_IMAGE020
，
Figure 201310003415X100001DEST_PATH_IMAGE022
The first representing the area AThe number of feature points extracted by the sub-regions,
Figure 201310003415X100001DEST_PATH_IMAGE024
the dimension of SIFT feature matrix is represented;
Figure 307303DEST_PATH_IMAGE020
indicating the number of lines
Figure 732337DEST_PATH_IMAGE022
The number of rows is
Figure 657568DEST_PATH_IMAGE024
A matrix of (a);
e) calculating the similarity of the two face photos, and aiming at the two face photos
Figure 201310003415X100001DEST_PATH_IMAGE026
And
Figure 201310003415X100001DEST_PATH_IMAGE028
respectively obtaining SIFT feature matrixes according to the steps b), c) and d)
Figure 201310003415X100001DEST_PATH_IMAGE030
And
Figure 201310003415X100001DEST_PATH_IMAGE032
(ii) a Calculating a matrix of corresponding sub-regions
Figure 196390DEST_PATH_IMAGE030
And
Figure 452360DEST_PATH_IMAGE032
the similarity between any two lines and all the similarity values form a matrixAnd define，
Figure 201310003415X100001DEST_PATH_IMAGE038
Presentation photographAnd
Figure 901237DEST_PATH_IMAGE028
similarity corresponding to jth sub-region, and its size is taken as matrix
Figure 929236DEST_PATH_IMAGE034
Maximum value of all elements in;
f) counting the probability distribution of similarity to make two pictures of eyebrow region
Figure 48501DEST_PATH_IMAGE010
The similarity of the sub-regions is a sample, and the similarity of each sub-region of the eyebrow is countedProbability distribution when the eyebrows of the same person are the same and when the eyebrows of the different persons are not the same;
g) calculating the similarity probability of the eyebrow regions of the two compared images
Figure 201310003415X100001DEST_PATH_IMAGE040
Calculating two images to be compared according to steps b), c), d) and e)
Figure 777478DEST_PATH_IMAGE014
Similarity of individual sub-regions
Figure 201310003415X100001DEST_PATH_IMAGE042
、、…、
Figure 631875DEST_PATH_IMAGE038
Obtaining by probability distribution in step fAnd
Figure 201310003415X100001DEST_PATH_IMAGE046
，
Figure 195711DEST_PATH_IMAGE044
indicates to correspond to
Figure 201310003415X100001DEST_PATH_IMAGE048
When the sub-regions are eyebrows of the same person, the similarity isA probability value of (d);
Figure 60954DEST_PATH_IMAGE046
indicates to correspond to
Figure 217129DEST_PATH_IMAGE048
When the sub-regions are not eyebrows of the same person, the similarity is
Figure 951867DEST_PATH_IMAGE038
0 < (r) >, a probability value of
Figure 560702DEST_PATH_IMAGE018
≤
Figure 406299DEST_PATH_IMAGE014
(ii) a Obtaining the probability that the two comparison images are the eyebrows of the same person under each similarity according to a Bayes formula:
Figure 201310003415X100001DEST_PATH_IMAGE050
Figure 201310003415X100001DEST_PATH_IMAGE052
；
and judging whether the two contrast images come from the same person or not according to the probability value of the obtained two contrast images being the eyebrows of the same person.
2. The method as claimed in claim 1, wherein the selected eyebrow area in step b) is selected based on SIFT feature
Figure 731494DEST_PATH_IMAGE010
Setting selected eyebrow regions for the left eyebrow region, the right eyebrow region or the whole eyebrow region
Figure 699450DEST_PATH_IMAGE010
Respectively has a width and a height of
Figure 201310003415X100001DEST_PATH_IMAGE054
、
Figure 201310003415X100001DEST_PATH_IMAGE056
Which comprises the following steps:
b-1) positioning the pupil position, and positioning the face photo by using the pupil positioning method
Figure 670948DEST_PATH_IMAGE012
The pupil position of (1) is the connecting line between two pupils
Figure 201310003415X100001DEST_PATH_IMAGE058
The axes establish a plane rectangular coordinate system, and the coordinates of the left pupil and the right pupil are respectively set as
Figure DEST_PATH_IMAGE060
、
Figure DEST_PATH_IMAGE062
；
b-2) solvingTaking the interpupillary distance, according to the formula
Figure DEST_PATH_IMAGE064
Calculating the distance between two pupils;
b-3) if region
Figure 631820DEST_PATH_IMAGE010
When the area is the left eyebrow area, the selected eyebrow areaWidth of (2)
Figure DEST_PATH_IMAGE066
Height of
Figure DEST_PATH_IMAGE068
，
Figure 17289DEST_PATH_IMAGE010
The coordinates of the center point of the region are
Figure DEST_PATH_IMAGE070
(ii) a If region
Figure 525018DEST_PATH_IMAGE010
When the right eyebrow area is selected, the selected eyebrow area
Figure 977996DEST_PATH_IMAGE010
Width of (2)
Figure 963270DEST_PATH_IMAGE066
Height of
Figure 709247DEST_PATH_IMAGE068
，
Figure 766589DEST_PATH_IMAGE010
The coordinates of the center point of the region are
Figure DEST_PATH_IMAGE072
(ii) a If region
Figure 819996DEST_PATH_IMAGE010
When the eyebrow area is all, the selected eyebrow area
Figure 597459DEST_PATH_IMAGE010
Width of (2)
Figure DEST_PATH_IMAGE074
Height of
Figure 466668DEST_PATH_IMAGE068
，
Figure 460907DEST_PATH_IMAGE010
The coordinates of the center point of the region are
Figure DEST_PATH_IMAGE076
；
Wherein,
Figure DEST_PATH_IMAGE078
、
Figure DEST_PATH_IMAGE080
、
Figure DEST_PATH_IMAGE082
、
Figure DEST_PATH_IMAGE084
are all constants.
3. The SIFT-feature-based eyebrow recognition method according to claim 2, wherein: the above-mentioned
Figure 865475DEST_PATH_IMAGE078
=0.625、
Figure 264488DEST_PATH_IMAGE080
=0.391、=1.563、
Figure 571153DEST_PATH_IMAGE084
=0.281。
4. The SIFT-feature-based eyebrow recognition method according to claim 1 or 2, wherein: the matrix of the corresponding sub-region in the step e)
Figure DEST_PATH_IMAGE086
And
Figure DEST_PATH_IMAGE088
the method for calculating the similarity between any two lines is Euclidean distance, Mahalanobis distance or vector inner product algorithm.
5. The SIFT-feature-based eyebrow recognition method according to claim 1 or 2, wherein: the probability distribution of the statistical similarity in the step f) is a discrete model or a continuous model; under the condition of a discrete model, the similarity is obtained by counting the probability value of each similarity value falling in each numerical value interval; in the case of a continuous model, a probability density function is formed by fitting using Gaussian mixture probability modeling.
6. The SIFT-feature-based eyebrow recognition method according to claim 1 or 2, wherein: capacity in said step a)
Figure DEST_PATH_IMAGE090
=2000，
Figure 153313DEST_PATH_IMAGE004
=200，
Figure 282199DEST_PATH_IMAGE006
= 10; the different shooting conditions refer to different postures and different illumination.
CN201310003415XA 2013-01-06 2013-01-06 Scale-invariant feature transform (SIFT) feature based method for identifying eyebrows Pending CN102999751A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310003415XA CN102999751A (en) 2013-01-06 2013-01-06 Scale-invariant feature transform (SIFT) feature based method for identifying eyebrows

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310003415XA CN102999751A (en) 2013-01-06 2013-01-06 Scale-invariant feature transform (SIFT) feature based method for identifying eyebrows

Publications (1)

Publication Number Publication Date
CN102999751A true CN102999751A (en) 2013-03-27

Family

ID=47928300

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310003415XA Pending CN102999751A (en) 2013-01-06 2013-01-06 Scale-invariant feature transform (SIFT) feature based method for identifying eyebrows

Country Status (1)

Country Link
CN (1) CN102999751A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104898704A (en) * 2015-03-12 2015-09-09 哈尔滨理工大学 Intelligent eyebrow penciling machine device based on DSP image processing
CN108985153A (en) * 2018-06-05 2018-12-11 成都通甲优博科技有限责任公司 A kind of face recognition method and device

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1801180A (en) * 2005-02-24 2006-07-12 北京工业大学 Identity recognition method based on eyebrow recognition
CN1926575A (en) * 2004-03-03 2007-03-07 日本电气株式会社 Image similarity calculation system, image search system, image similarity calculation method, and image similarity calculation program
CN101510257A (en) * 2009-03-31 2009-08-19 华为技术有限公司 Human face similarity degree matching method and device
CN101901353A (en) * 2010-07-23 2010-12-01 北京工业大学 Eyebrow Image Recognition Method Based on Subregion Matching

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1926575A (en) * 2004-03-03 2007-03-07 日本电气株式会社 Image similarity calculation system, image search system, image similarity calculation method, and image similarity calculation program
CN1801180A (en) * 2005-02-24 2006-07-12 北京工业大学 Identity recognition method based on eyebrow recognition
CN101510257A (en) * 2009-03-31 2009-08-19 华为技术有限公司 Human face similarity degree matching method and device
CN101901353A (en) * 2010-07-23 2010-12-01 北京工业大学 Eyebrow Image Recognition Method Based on Subregion Matching

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
余文峰: "《基于特征的肖像画自动生成系统》", 《中国优秀硕士学位论文全文数据库》, no. 9, 15 September 2006 (2006-09-15), pages 24 - 25 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104898704A (en) * 2015-03-12 2015-09-09 哈尔滨理工大学 Intelligent eyebrow penciling machine device based on DSP image processing
CN108985153A (en) * 2018-06-05 2018-12-11 成都通甲优博科技有限责任公司 A kind of face recognition method and device

Similar Documents

Publication Publication Date Title
Jian et al. Simultaneous hallucination and recognition of low-resolution faces based on singular value decomposition
US10083366B2 (en) Edge-based recognition, systems and methods
CN105389554B (en) Living body distinguishing method and device based on face recognition
US10438059B2 (en) Image recognition method, image recognition apparatus, and recording medium
Zhao et al. Learning mid-level filters for person re-identification
CN106372629B (en) Living body detection method and device
JP5899179B2 (en) Person recognition device, person recognition method, person recognition program, and recording medium thereof
CN107346414B (en) Pedestrian attribute recognition method and device
CN103729649B (en) A kind of image rotation angle detection method and device
CN104598871B (en) A kind of facial age computational methods based on correlation regression
JP2008538998A (en) Use of time when recognizing a person in an image
EP1886255A1 (en) Using photographer identity to classify images
CN111178195A (en) Facial expression recognition method and device and computer readable storage medium
CN109725721B (en) Human eye positioning method and system for naked eye 3D display system
Parde et al. Face and image representation in deep CNN features
US11783192B2 (en) Hybrid deep learning method for recognizing facial expressions
KR101326691B1 (en) Robust face recognition method through statistical learning of local features
JP6410450B2 (en) Object identification device, object identification method, and program
Burgos-Artizzu et al. Distance estimation of an unknown person from a portrait
CN102999751A (en) Scale-invariant feature transform (SIFT) feature based method for identifying eyebrows
US20160292529A1 (en) Image collation system, image collation method, and program
CN107480628A (en) A kind of face identification method and device
JP2018022416A (en) Face direction estimation apparatus and program thereof
CN100418106C (en) Method and device for detecting human faces
CN114627327B (en) A method for image classification based on feature enhancement

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C12 Rejection of a patent application after its publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20130327