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 PDFInfo
- 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
- zone
- similarity
- sift
- same person
- 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
Links
Images
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
Technical field
The present invention relates to a kind of eyebrow recognition method based on the SIFT feature, relate to the Digital Image Processing application.
Background technology
Eyebrow is feature important in recognition of face, than people's face further feature, has better stability and otherness, but usually is subject to the impact of the factors such as illumination, attitude.
(publication number: 1645406) the RGB color component of each pixel of use eyebrow is poor is basis of characterization to patent " authentication identifying method based on eyebrow recognition ".The method is subject to the impact of illumination and attitude larger, and particularly, under the sidelight condition, recognition effect is poor.
Patent " eyebrow image identifying method based on the subregion coupling " (publication number: 101901353A) the manual pure eyebrow image of choosing each user, template as each user, use eyebrow image to be identified to carry out convolution algorithm with the eyebrow template of having preserved successively, obtain being identified after similarity.Although the method can reduce the some effects of illumination and attitude, the process of choosing by hand pure eyebrow image is very complicated loaded down with trivial details.
Summary of the invention
The present invention, in order to overcome the shortcoming of above-mentioned technical matters, provides a kind of eyebrow recognition method based on the SIFT feature that can effectively reduce illumination, attitude variable effect.
Eyebrow recognition method based on the SIFT feature of the present invention, its special feature is, comprises the following steps: a). gather human face photo, the establishment capacity is
The human face photo storehouse, wherein
The number of people in face database,
Mean the
The human face photo of individual under different shooting conditions, wherein 0<
≤
B). choose the eyebrow zone
, choose human face photo
The eyebrow zone
, the zone that utilizes it to calculate as the human face photo similarity; C). divide subregion, by the eyebrow zone
Be divided into
Sub regions, subregion is used
Mean; Common factor between different subregions can be sky, can not be also empty;
Indicate
The of zone
Sub regions, 0<
≤
D). obtain the SIFT eigenmatrix, utilize the SIFT algorithm to obtain the eyebrow zone
The SIFT eigenmatrix of sub regions
,
Mean of regional A
The characteristic point number that sub regions is extracted,
The dimension that means the SIFT eigenmatrix;
Mean that line number is
, columns is
Matrix; E). calculate the similarity of two human face photos, for two human face photos
With
, all according to step b), c) and d) obtain respectively the SIFT eigenmatrix
With
Calculate the matrix of corresponding subregion
With
Similarity between any two row, form matrix by all similarity values
, and definition
,
Mean photo
With
The similarity of corresponding j sub regions, its size is got matrix
The maximum of middle all elements; F). the probability distribution of statistics similarity, with photo eyebrow zone in twos
The similarity of subregion be sample, the similarity of each sub regions of statistics eyebrow
Be same person eyebrow and the probability distribution while not being the same person eyebrow; G). calculate the similar probability in two contrast images eyebrow zones
, according to step b), c), d) and e) calculate two and treat contrast images
The similarity of sub regions
,
...,
, by the probability distribution in step f, obtain
With
,
Mean corresponding the
When sub regions is the eyebrow of same person, similarity is
Probable value;
Mean corresponding the
When sub regions is not the eyebrow of same person, similarity is
Probable value, 0<
≤
Obtain under each similarity the probability that two contrast images are the same person eyebrow according to Bayesian formula:
The probable value that is the same person eyebrow according to two contrast images of obtaining, can judge whether two contrast images come from same person.
Step e) in, corresponding subregion should be the eyebrow zone of two images according to identical subregion division methods, the corresponding part of obtaining, only have corresponding part just to have meaning and necessity of calculating similarity.In this step, due to
, calculate
With
The matrix that similarity between any two row is obtained
In, no matter how element is arranged, and can not affect similarity
The size of numerical value.Step f), in, the similarity of take is obtained as sample
Be same person eyebrow and the probability distribution while not being the same person eyebrow; Like this, in step g) in, by step f) in the probability distribution function that obtains can calculate
With
Numerical value.Step g) can set threshold values in, when two contrast images of the obtaining similar probable value that is the same person eyebrow is greater than while setting threshold values, think that two images come from same people; If be less than while setting threshold values, think that two images are not same people's images.
Eyebrow recognition method based on the SIFT feature of the present invention, step b) the eyebrow zone of choosing in
For left eyebrow zone, right eyebrow zone or whole eyebrows zone, establish the eyebrow zone of choosing
Width, highly be respectively
,
, it comprises the following steps: b-1). and the location pupil position adopts pupil positioning method location human face photo
Pupil position, the line of take between two pupils as
Axle is set up plane right-angle coordinate, and the coordinate of establishing left and right pupil is respectively
,
B-2). ask for interocular distance, according to formula
, ask for the distance between two pupils; B-3) if. the zone
During for left eyebrow zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
If regional
During for right eyebrow zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
If regional
During for whole eyebrows zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
Wherein,
,
,
,
Be constant.
Eyebrow recognition method based on the SIFT feature of the present invention, described step e) matrix of corresponding subregion in
With
Between any two row, the computing method of similarity are Euclidean distance, mahalanobis distance or inner product of vectors algorithm.Euclidean distance, mahalanobis distance or inner product of vectors are the existing method of calculating similarity.
Eyebrow recognition method based on the SIFT feature of the present invention, step f) in, the probability distribution of statistics similarity is discrete model or continuous model; Be under the situation of discrete model, obtaining by adding up the probable value that each similarity value drops on each numerical value interval; Be under the situation of continuous model, using the mixed Gaussian probabilistic Modeling, matching formation probability density function.
Eyebrow recognition method based on the SIFT feature of the present invention, described step is middle capacity a)
=2000,
=200,
=10; Described different shooting condition refers to different attitudes, different light.
The invention has the beneficial effects as follows: the eyebrow recognition method based on the SIFT feature of the present invention, human face photo storehouse under the different shooting conditions of the some people of model, the eyebrow zone is divided into to subregion and adopts the SIFT eigenmatrix to calculate the regional similarity of taking an X-ray in twos, then take similarity as Sample Establishing be same person eyebrow and the probability distribution while not being the same person eyebrow; Finally, according to Bayesian formula, can obtain the probability that two contrast images are the same person eyebrow, by the probability size, can judge whether two images come from same people.
Yardstick invariant features conversion SIFT algorithm, be applicable to, in the feature description and characteristic matching of rigid objects, to have the rotation to image, the characteristic that dimensional variation remains unchanged.Use the SIFT feature to carry out eyebrow recognition, can effectively reduce the impact of illumination, attitude variation, and not need artificial participation.
The accompanying drawing explanation
The similarity that Fig. 1 is the eyebrow all subregion
Probability distribution when being the same person eyebrow;
Embodiment
Below in conjunction with accompanying drawing and embodiment, the invention will be further described.
Eyebrow recognition method based on the SIFT feature of the present invention comprises the following steps:
A). gather human face photo, the establishment capacity is
The human face photo storehouse, wherein
The number of people in face database,
Mean the
The human face photo of individual under different shooting conditions, wherein 0<
≤
This step is middle capacity a)
Can be chosen for 2000,
=200,
=10; Described different shooting condition refers to different attitudes, different light;
B). choose the eyebrow zone
, choose human face photo
The eyebrow zone
, the zone that utilizes it to calculate as the human face photo similarity;
This step b) the eyebrow zone of choosing in
Can, for left eyebrow zone, right eyebrow zone or whole eyebrows zone, establish the eyebrow zone of choosing
Width, highly be respectively
,
, it comprises the following steps:
B-1). the location pupil position adopts pupil positioning method location human face photo
Pupil position, the line of take between two pupils as
Axle is set up plane right-angle coordinate, and the coordinate of establishing left and right pupil is respectively
,
B-3) if. the zone
During for left eyebrow zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
If regional
During for right eyebrow zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
If regional
During for whole eyebrows zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
C). divide subregion, by the eyebrow zone
Be divided into
Sub regions, subregion is used
Mean; Common factor between different subregions can be sky, can not be also empty;
Indicate
The of zone
Sub regions, 0<
≤
D). obtain the SIFT eigenmatrix, utilize the SIFT algorithm to obtain the eyebrow zone
The SIFT eigenmatrix of sub regions
,
Mean of regional A
The unique point number that sub regions is extracted,
The dimension that means the SIFT eigenmatrix;
Mean that line number is
, columns is
Matrix;
E). calculate the similarity of two human face photos, for two human face photos
With
, all according to step b), c) and d) obtain respectively the SIFT eigenmatrix
With
Calculate the matrix of corresponding subregion
With
Similarity between any two row, form matrix by all similarity values
, and definition
,
Mean photo
With
The similarity of corresponding j sub regions, its size is got matrix
The maximal value of middle all elements;
This step e) in, the matrix of corresponding subregion
With
Between any two row, the computing method of similarity are Euclidean distance, mahalanobis distance or inner product of vectors algorithm;
F). the probability distribution of statistics similarity, with photo eyebrow zone in twos
The similarity of subregion be sample, the similarity of each sub regions of statistics eyebrow is being same person eyebrow and the probability distribution while not being the same person eyebrow;
This step f), in, the probability distribution of statistics similarity is discrete model or continuous model; Be under the situation of discrete model, obtaining by adding up the probable value that each similarity value drops on each numerical value interval; Be under the situation of continuous model, using the mixed Gaussian probabilistic Modeling, matching formation probability density function; As depicted in figs. 1 and 2, provided respectively similarity
Be same person eyebrow and the Probability Distribution Fitting formation probability density function image while not being the same person eyebrow;
G). calculate the similar probability in two contrast images eyebrow zones
, according to step b), c), d) and e) calculate two and treat contrast images
The similarity of sub regions
,
...,
, by the probability distribution in step f, obtain
With
,
Mean corresponding the
When sub regions is the eyebrow of same person, similarity is
Probable value;
Mean corresponding the
When sub regions is not the eyebrow of same person, similarity is
Probable value, 0<
≤
Obtain under each similarity the probability that two contrast images are the same person eyebrow according to Bayesian formula:
The probable value that is the same person eyebrow according to two contrast images of obtaining, can judge whether two contrast images come from same person.
Claims (6)
1. the eyebrow recognition method based on the SIFT feature, is characterized in that, comprises the following steps:
A). gather human face photo, the establishment capacity is
The human face photo storehouse, wherein
The number of people in face database,
Mean the
The human face photo of individual under different shooting conditions, wherein 0<
≤
B). choose the eyebrow zone
, choose human face photo
The eyebrow zone
, the zone that utilizes it to calculate as the human face photo similarity;
C). divide subregion, by the eyebrow zone
Be divided into
Sub regions, subregion is used
Mean; Common factor between different subregions can be sky, can not be also empty;
Indicate
The of zone
Sub regions, 0<
≤
D). obtain the SIFT eigenmatrix, utilize the SIFT algorithm to obtain the eyebrow zone
The SIFT eigenmatrix of sub regions
,
Mean of regional A
The unique point number that sub regions is extracted,
The dimension that means the SIFT eigenmatrix;
Mean that line number is
, columns is
Matrix;
E). calculate the similarity of two human face photos, for two human face photos
With
, all according to step b), c) and d) obtain respectively the SIFT eigenmatrix
With
Calculate the matrix of corresponding subregion
With
Similarity between any two row, form matrix by all similarity values
, and definition
,
Mean photo
With
The similarity of corresponding j sub regions, its size is got matrix
The maximal value of middle all elements;
F). the probability distribution of statistics similarity, with photo eyebrow zone in twos
The similarity of subregion be sample, the similarity of each sub regions of statistics eyebrow
Be same person eyebrow and the probability distribution while not being the same person eyebrow;
G). calculate the similar probability in two contrast images eyebrow zones
, according to step b), c), d) and e) calculate two and treat contrast images
The similarity of sub regions
,
...,
, by the probability distribution in step f, obtain
With
,
Mean corresponding the
When sub regions is the eyebrow of same person, similarity is
Probable value;
Mean corresponding the
When sub regions is not the eyebrow of same person, similarity is
Probable value, 0<
≤
Obtain under each similarity the probability that two contrast images are the same person eyebrow according to Bayesian formula:
The probable value that is the same person eyebrow according to two contrast images of obtaining, can judge whether two contrast images come from same person.
2. the eyebrow recognition method based on the SIFT feature according to claim 1, is characterized in that step b) in the eyebrow zone chosen
For left eyebrow zone, right eyebrow zone or whole eyebrows zone, establish the eyebrow zone of choosing
Width, highly be respectively
,
, it comprises the following steps:
B-1). the location pupil position adopts pupil positioning method location human face photo
Pupil position, the line of take between two pupils as
Axle is set up plane right-angle coordinate, and the coordinate of establishing left and right pupil is respectively
,
B-3) if. the zone
During for left eyebrow zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
If regional
During for right eyebrow zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
If regional
During for whole eyebrows zone, the eyebrow zone of choosing
Width
, highly
,
The regional center point coordinate is
4. the eyebrow recognition method based on the SIFT feature according to claim 1 and 2, is characterized in that: the matrix of corresponding subregion described step e)
With
Between any two row, the computing method of similarity are Euclidean distance, mahalanobis distance or inner product of vectors algorithm.
5. the eyebrow recognition method based on the SIFT feature according to claim 1 and 2 is characterized in that: step f), the probability distribution of statistics similarity is discrete model or continuous model; Be under the situation of discrete model, obtaining by adding up the probable value that each similarity value drops on each numerical value interval; Be under the situation of continuous model, using the mixed Gaussian probabilistic Modeling, matching formation probability density function.
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)
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)
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 | Subregion-based matched eyebrow image identifying method |
-
2013
- 2013-01-06 CN CN201310003415XA patent/CN102999751A/en active Pending
Patent Citations (4)
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 | Subregion-based matched eyebrow image identifying method |
Non-Patent Citations (1)
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)
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 |
---|---|---|
Saavedra et al. | Sketch based Image Retrieval using Learned KeyShapes (LKS). | |
EP3084682B1 (en) | System and method for identifying faces in unconstrained media | |
CN109583298B (en) | Cross-view gait recognition method based on set | |
US20170032222A1 (en) | Cross-trained convolutional neural networks using multimodal images | |
US20150261803A1 (en) | Edge-based recognition, systems and methods | |
Xiaofeng et al. | Discriminatively trained sparse code gradients for contour detection | |
CN101593205B (en) | Video-based method for searching three-dimension model | |
CN107085731B (en) | Image classification method based on RGB-D fusion features and sparse coding | |
CN109509142A (en) | A kind of face ageing image processing method, system, readable storage medium storing program for executing and equipment | |
JP2010108494A (en) | Method and system for determining characteristic of face within image | |
CN105809182B (en) | Image classification method and device | |
CN103902992A (en) | Human face recognition method | |
Su et al. | Bayesian depth estimation from monocular natural images | |
CN102999751A (en) | Scale-invariant feature transform (SIFT) feature based method for identifying eyebrows | |
CN112329662A (en) | Multi-view saliency estimation method based on unsupervised learning | |
CN109711231A (en) | A kind of face alignment method based on average face-Pu Shi transformation | |
Häfner et al. | Colonic polyp classification in high-definition video using complex wavelet-packets | |
Arandjelovic et al. | A new look at filtering techniques for illumination invariance in automatic face recognition | |
Duan et al. | Local feature learning for face recognition under varying poses | |
Li et al. | Iris recognition using ordinal encoding of log-euclidean covariance matrices | |
Mohammad et al. | Short-term user authentication using eyebrows biometric for smartphone devices | |
Quraishi et al. | A novel approach for face detection using artificial neural network | |
Ciobanu et al. | A novel iris clustering approach using LAB color features | |
CN104573682A (en) | Face anti-counterfeiting method based on face similarity | |
Geng et al. | A person re-identification algorithm by using region-based feature selection and feature fusion |
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 |