CN106934340A - A kind of human face light invariant feature extraction method of utilization logarithmic transformation and Sobel operators - Google Patents
A kind of human face light invariant feature extraction method of utilization logarithmic transformation and Sobel operators Download PDFInfo
- Publication number
- CN106934340A CN106934340A CN201710044647.8A CN201710044647A CN106934340A CN 106934340 A CN106934340 A CN 106934340A CN 201710044647 A CN201710044647 A CN 201710044647A CN 106934340 A CN106934340 A CN 106934340A
- Authority
- CN
- China
- Prior art keywords
- lnr
- sobel operators
- lni
- formula
- human face
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/60—Extraction of image or video features relating to illumination properties, e.g. using a reflectance or lighting model
Abstract
The invention discloses a kind of utilization logarithmic transformation and the human face light invariant feature extraction method of Sobel operators, comprise the following steps:Facial image is transformed into log-domain;Treatment is sharpened to facial image using Sobel operators, the minutia of face horizontal direction is extracted, while eliminating illumination component, acquired results are exactly the human face light invariant features for needing to extract.Illumination of the present invention to different angles has certain inhibitory action, and the face identification rate after treatment after the more existing algorithm process of the discrimination of facial image is high;The present invention does photo-irradiation treatment to facial image using logarithmic transformation and Sobel operators, and method is simple, and recognition speed is fast.
Description
Technical field
The present invention relates to mode identification technology, the face light of specifically a kind of utilization logarithmic transformation and Sobel operators
According to invariant feature extraction method.
Background technology
With the development of information technology, high-tech has incorporated our life in digitized form, to efficiently carrying out
Identity differentiates so that it is guaranteed that the security of various digital informations proposes requirement higher, and identification also there occurs with verification technique
Huge change, by modes such as traditional password authentifications, is more converted to the emerging skill such as digital certificate and living things feature recognition
Art mode.As the face recognition technology of one of biometrics identification technology, in national security, financial security and man-machine interaction etc.
Field has broad application prospects.By the research of many decades, face recognition technology can have been obtained preferably in the ideal case
Recognition performance.But the influence of factor such as it is vulnerable to illumination, attitude in uncontrollable environment, expresses one's feelings, block, makes recognition performance
Drastically decline.Wherein, the problem of uneven illumination is particularly acute, therefore, to improve the performance of the recognition of face in photoenvironment, solution
Certainly the method for lighting issues is broadly divided into three classes in recognition of face:Extract illumination invariant feature, the modeling of illumination variation, illumination bar
Part is standardized.In the method for extracting illumination invariant feature, two kinds are broadly divided into again:The first is in log-domain, by low pass
Filtering eliminates illumination component, for example:In log-domain, treatment is filtered using discrete cosine transform (DCT) or wavelet transformation
Etc. method.Second is construction division arithmetic, the illumination component of slow change is eliminated by division arithmetic, for example:Gradient face
(Gradient-face), the side such as weber face (Weber-face), local binary patterns (Local Binary Pattern, LBP)
Method.The face characteristic of robustness is kept in illumination to strengthen recognition of face application in practice therefore, it is possible to efficiently extract
Become a vital problem.
The content of the invention
It is an object of the invention to provide a kind of utilization logarithmic transformation that can improve face identification rate and Sobel operators
Human face light invariant feature extraction method, to solve the problems, such as to be proposed in above-mentioned background technology.
To achieve the above object, the present invention provides following technical scheme:
A kind of human face light invariant feature extraction method of utilization logarithmic transformation and Sobel operators, comprises the following steps:
(1) original image is transformed into log-domain;
(2) in log-domain, treatment is sharpened to facial image using Sobel operators, extracts face horizontal direction
Minutia, while eliminating illumination component, acquired results are exactly the human face light invariant features for needing to extract.
As further scheme of the invention:According to illumination reflection model, any Gray Face image I (x, y) is reflection
The product of components R (x, y) and illumination component L (x, y), i.e. I (x, y)=R (x, y) L (x, y);In the step (1), to face
Image carries out logarithmic transformation, makes its reflecting component be transformed to be added by being multiplied with illumination component, i.e. ln I (x, y)=ln R (x,
y)+ln L(x,y)。
As further scheme of the invention:With Sobel operatorsMatrix template is sharpened place
Facial image after reason is the gray value I'(x of I' pixels (x, y), y) as shown in formula (1):
I'(x, y)=ln I (x-1, y-1)+2ln I (x, y-1)+ln I (x+1, y-1)-ln I (x-1, y+1) -2ln I
(x,y+1)-ln I(x+1,y+1) (1);
In order to express easily, I (x+m, y+n) is abbreviated as Im,n, similarly, R (x+m, y+n) is abbreviated as Rm,n, L (x+m, y+n)
It is abbreviated as Lm,n, then formula (1) be abbreviated as formula (2), it is as follows:
I'0,0=ln I-1,-1+2ln I0,-1+ln I1,-1-ln I-1,1-2ln I0,1-ln I1,1(2);
Formula (3) is further obtained according to illumination reflection model:
I'0,0=ln R-1,-1L-1,-1+2ln R0,-1L0,-1+ln R1,-1L1,-1-ln R-1,1L-1,1-2ln R0,1L0,1-ln
R1,1L1,1(3);
As further scheme of the invention:Illumination component L (x, y) change is slow, so there is formula (4):
L-1,-1≈L0,-1≈L1,-1≈L-1,1≈L0,1≈L1,1(4);
All variables for formula (4) use LtSubstitute, obtain formula (5):
I'0,0=ln R-1,-1Lt+2ln R0,-1Lt+ln R1,-1Lt-ln R-1,1Lt-2ln R0,1Lt-ln R1,1Lt(5);
=ln R-1,-1+ln Lt+2ln R0,-1+2ln Lt+ln R1,-1+ln Lt-ln R-1,1-ln Lt-2ln
R0,1-2ln Lt-ln R1,1-ln Lt
=ln R-1,-1+2ln R0,-1+ln R1,-1-ln R-1,1-2ln R0,1-ln R1,1
Therefore, illumination component is eliminated.
Compared with prior art, the beneficial effects of the invention are as follows:
Illumination of the present invention to different angles has certain inhibitory action, and the discrimination of facial image is more existing after treatment
Algorithm process after face identification rate it is high;The present invention does photo-irradiation treatment using logarithmic transformation and Sobel operators to facial image,
Method is simple, and recognition speed is fast.
Brief description of the drawings
Fig. 1 is the basic flow sheet that the present invention extracts human face light invariant features.
Fig. 2 is the eight neighborhood gray scale value matrix in the present invention centered on (x, y) and its corresponding simplified expression schematic diagram.
Fig. 3 is to extract the face exemplary plot before and after CMU PIE face database illumination invariant features.
Fig. 4 is to extract the face exemplary plot before and after the illumination invariant feature of extension Yale B front faces storehouse.
Specific embodiment
The technical scheme of this patent is described in more detail with reference to specific embodiment.
A kind of human face light invariant feature extraction method of Fig. 1-4, utilization logarithmic transformation and Sobel operators is referred to, is wrapped
Include following steps:
(1) original image is transformed into log-domain;
(2) in log-domain, treatment is sharpened to facial image using Sobel operators, extracts face horizontal direction
Minutia, while eliminating illumination component, acquired results are exactly the human face light invariant features for needing to extract.
As further scheme of the invention:According to illumination reflection model, any Gray Face image I (x, y) can see
Into the product for being reflecting component R (x, y) and illumination component L (x, y), i.e. I (x, y)=R (x, y) L (x, y);The step (1)
In, logarithmic transformation is carried out to facial image, make its reflecting component that addition, i.e. ln I (x, y) are converted to by being multiplied with illumination component
=ln R (x, y)+ln L (x, y).
With Sobel operatorsIt is I' center pixels that template is sharpened the facial image after treatment
The gray value of point (x, y) is I'(x, y).
Specifically, the human face light invariant feature extraction method of the utilization logarithmic transformation and Sobel operators, specific step
It is rapid as follows:
(1) shown in the 8 neighborhoods such as Fig. 2 (a) centered on a certain pixel (x, y) of original image, stated to simplify, middle imago
Gray value I (x, y) at vegetarian refreshments (x, y) place is expressed as I0,0, such as shown in Fig. 2 (b), gray value I (x+m, y+n) table of its neighborhood point
It is I to statem,n。
According to illumination reflection model, any Gray Face image I (x, y) can regard reflecting component R (x, y) and illumination as
The product of component L (x, y), i.e. I (x, y)=R (x, y) L (x, y), wherein L (x, y) depend on light source and only represent facial illumination
Component, R (x, y) depends on the surface characteristics of object and contains the key message of face, in the present invention I (x+m, y+n)
R can be expressed asm,nLm,n.By facial image, each pixel value transforms to log-domain, and ln I (x+m, y+n) are expressed as in the present invention
ln RM, n+ln LM, n。
(2) Sobel operators can carry out the Edge contrast horizontally and vertically gone up to image, strengthen the thin of image
Section edge and contour feature, and there is certain smoothing effect to image random noise.Wherein extract image level direction thin
Save the template of featureExtract the template of image vertical direction minutia
The present invention is only with DxMatrix template is sharpened treatment to facial image, obtains I'.Specific principle derivation
It is as follows:
If with the D of Sobel operatorsxTemplate is sharpened the gray value of facial image I' pixels (x, y) after treatment
I'(x, y) as shown in formula (1):
I'(x, y)=ln I (x-1, y-1)+2ln I (x, y-1)+ln I (x+1, y-1)-ln I (x-1, y+1) -2ln I
(x,y+1)-ln I(x+1,y+1) (1)
In order to express easily, I (x+m, y+n) is abbreviated as Im,n, similarly, R (x+m, y+n) is abbreviated as Rm,n, L (x+m, y+n)
It is abbreviated as Lm,n, then formula (1) be abbreviated as formula (2), it is as follows:
I'0,0=ln I-1,-1+2ln I0,-1+ln I1,-1-ln I-1,1-2ln I0,1-ln I1,1 (2)
Formula (3) is further obtained according to illumination reflection model:
I'0,0=ln R-1,-1L-1,-1+2ln R0,-1L0,-1+ln R1,-1L1,-1-ln R-1,1L-1,1-2ln R0,1L0,1-ln
R1,1L1,1 (3)
As further scheme of the invention:Illumination component L (x, y) change is slow, so there is formula (4):
L-1,-1≈L0,-1≈L1,-1≈L-1,1≈L0,1≈L1,1 (4)
All variables for formula (4) use LtSubstitute, obtain formula (5):
I'0,0=ln R-1,-1Lt+2ln R0,-1Lt+ln R1,-1Lt-ln R-1,1Lt-2ln R0,1Lt-ln R1,1Lt (5)
=ln R-1,-1+ln Lt+2ln R0,-1+2ln Lt+ln R1,-1+ln Lt-ln R-1,1-ln Lt-2ln
R0,1-2ln Lt-ln R1,1-ln Lt
=ln R-1,-1+2ln R0,-1+ln R1,-1-ln R-1,1-2ln R0,1-ln R1,1
In formula (5), illumination component is eliminated.
The present invention is tested in CMU PIE and extension Yale B face databases:
1st, face database introduction:CMU PIE face databases have 68 the 41368 of people facial images, present invention selection
Be to be tested in front face light group (C27), C27 light groups altogether include 1428 face databases.During experiment, use
Facial size is 32 × 32, and an image is chosen as training sample from the different illumination of everyone 21 kinds, and remaining face is made
It is test sample.
Altogether comprising 38 9 kinds of people different attitudes, every kind of attitude includes 64 kinds of different light to extension Yale B face databases again
According to situation.The present invention is only tested totally in extension Yale B front faces storehouse (2432 faces), and face images are pressed
Illumination incident angle can be divided into 5 subsets:Subset 1 (12 ° of θ <) has 266 samples, subset 2 (13 ° of 25 ° of < θ <) to have 456
Sample, subset 3 (26 ° of 50 ° of < θ <) have 456 samples, subset 4 (51 ° of 77 ° of < θ <) to have 532 samples, (the θ > of subset 5
77 °) there are 722 samples.During experiment, the facial size for using is 96 × 84, and with subset 1 as training sample, its complementary subset is made
It is test sample.
2nd, experimental result:
The human face light invariant features of extraction and original image have been carried out contrast and have found to be processed through the inventive method by experiment 1
Facial image feature afterwards becomes apparent, and effectively overcomes influence of the illumination to face characteristic.In CMU PIE front faces storehouse
With the face example difference before and after extraction illumination invariant feature in extension Yale B front faces storehouse as shown in Figure 3, Figure 4.
Experiment 2, after the inventive method uses L1 norm measurement distances, using nearest neighbor classifier discriminant classification.In CMU
Correct recognition rata on PIE face databases is 96.47%, is 98.85% in the face correct recognition rata of extension Yale B face databases,
There is preferable robustness to face invariant feature extraction.
Illumination of the present invention to different angles has certain inhibitory action, and the discrimination of facial image is more existing after treatment
Algorithm process after face identification rate it is high;The present invention does photo-irradiation treatment using logarithm operation and Sobel operators to facial image,
Method is simple, and recognition speed is fast.
The better embodiment to this patent is explained in detail above, but this patent is not limited to above-mentioned embodiment party
Formula, in the ken that one skilled in the relevant art possesses, can also be on the premise of this patent objective not be departed from
Make a variety of changes.
Claims (4)
1. a kind of human face light invariant feature extraction method of utilization logarithmic transformation and Sobel operators, it is characterised in that including with
Lower step:
(1) original image is transformed into log-domain;
(2) in log-domain, treatment is sharpened to facial image using Sobel operators, extracts the details of face horizontal direction
Feature, while eliminating illumination component, acquired results are exactly the human face light invariant features for needing to extract.
2. the human face light invariant feature extraction method of utilization logarithmic transformation according to claim 1 and Sobel operators, its
It is characterised by, according to illumination reflection model, any Gray Face image I (x, y) can regard reflecting component R (x, y) and light as
According to the product of component L (x, y), i.e. I (x, y)=R (x, y) L (x, y);In the step (1), logarithm change is carried out to facial image
Change, make its reflecting component that addition, i.e. lnI (x, y)=lnR (x, y)+lnL (x, y) are converted to by being multiplied with illumination component.
3. utilization logarithmic transformation and the human face light invariant feature extraction method of Sobel operators according to claim 1-2,
Characterized in that, with Sobel operatorsIt is I' pixels (x, y) that template is sharpened the image after treatment
Gray value I'(x, y) as shown in formula (1):
I'(x, y)=lnI (x-1, y-1)+2lnI (x, y-1)+lnI (x+1, y-1)-lnI (x-1, y+1) -2lnI (x, y+1) -
lnI(x+1,y+1) (1);
In order to express easily, I (x+m, y+n) is abbreviated as Im,n, similarly, R (x+m, y+n) is abbreviated as Rm,n, L (x+m, y+n) brief notes
It is Lm,n, then formula (1) be abbreviated as formula (2), it is as follows:
I'0,0=lnI-1,-1+2lnI0,-1+lnI1,-1-lnI-1,1-2lnI0,1-lnI1,1(2);
Formula (3) is further obtained according to illumination reflection model:
I'0,0=lnR-1,-1L-1,-1+2lnR0,-1L0,-1+lnR1,-1L1,-1-lnR-1,1L-1,1-2lnR0,1L0,1-lnR1,1L1,1 (3)。
4. the human face light invariant feature extraction method of utilization logarithmic transformation according to claim 3 and Sobel operators, its
It is characterised by, illumination component L (x, y) change is slow, so there is formula (4):
L-1,-1≈L0,-1≈L1,-1≈L-1,1≈L0,1≈L1,1(4);
All variables for formula (4) use LtSubstitute, obtain formula (5):
I'0,0=lnR-1,-1Lt+2lnR0,-1Lt+lnR1,-1Lt-lnR-1,1Lt-2lnR0,1Lt-lnR1,1Lt(5);
=lnR-1,-1+lnLt+2lnR0,-1+2lnLt+lnR1,-1+lnLt-lnR-1,1-lnLt-2lnR0,1-2lnLt-lnR1,1-
lnLt
=lnR-1,-1+2lnR0,-1+lnR1,-1-lnR-1,1-2lnR0,1-lnR1,1
Therefore, illumination component is eliminated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710044647.8A CN106934340A (en) | 2017-01-19 | 2017-01-19 | A kind of human face light invariant feature extraction method of utilization logarithmic transformation and Sobel operators |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710044647.8A CN106934340A (en) | 2017-01-19 | 2017-01-19 | A kind of human face light invariant feature extraction method of utilization logarithmic transformation and Sobel operators |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106934340A true CN106934340A (en) | 2017-07-07 |
Family
ID=59423981
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710044647.8A Pending CN106934340A (en) | 2017-01-19 | 2017-01-19 | A kind of human face light invariant feature extraction method of utilization logarithmic transformation and Sobel operators |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106934340A (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103295010A (en) * | 2013-05-30 | 2013-09-11 | 西安理工大学 | Illumination normalization method for processing face images |
CN103530634A (en) * | 2013-10-10 | 2014-01-22 | 中国科学院深圳先进技术研究院 | Face characteristic extraction method |
-
2017
- 2017-01-19 CN CN201710044647.8A patent/CN106934340A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103295010A (en) * | 2013-05-30 | 2013-09-11 | 西安理工大学 | Illumination normalization method for processing face images |
CN103530634A (en) * | 2013-10-10 | 2014-01-22 | 中国科学院深圳先进技术研究院 | Face characteristic extraction method |
Non-Patent Citations (2)
Title |
---|
于秋则: "合成孔径雷达(SAR)图像匹配导航技术研究", 《中国优秀博硕士学位伦文全文数据库(博士)-信息科技辑》 * |
杜波: "人脸识别中光照预处理方法研究", 《中国优秀博硕士学位论文全文数据库(硕士)-信息科技辑》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108229427B (en) | Identity security verification method and system based on identity document and face recognition | |
Ghiani et al. | Fingerprint liveness detection by local phase quantization | |
CN104200239A (en) | Image feature fusion identification based signature authentic identification system and method | |
CN108564040B (en) | Fingerprint activity detection method based on deep convolution characteristics | |
CN108960088A (en) | The detection of facial living body characteristics, the recognition methods of specific environment | |
CN113591747B (en) | Multi-scene iris recognition method based on deep learning | |
CN104809451A (en) | Handwriting authentication system based on stroke curvature detection | |
CN103246877A (en) | Image contour based novel human face recognition method | |
CN103136546A (en) | Multi-dimension authentication method and authentication device of on-line signature | |
Shinde et al. | Gender Classification from Face Images Using LBG Vector Quantization with Data Mining Algorithms | |
CN106845500A (en) | A kind of human face light invariant feature extraction method based on Sobel operators | |
CN105389573B (en) | A kind of face identification method based on three value mode layering manufactures of part | |
CN109815772A (en) | Fingerprint enhancement, recognition methods, device and Fingerprint enhancement identifying system | |
Tian et al. | Research of the Matlab application in the fingerprint identification system | |
CN106971143A (en) | A kind of human face light invariant feature extraction method of utilization logarithmic transformation and smothing filtering | |
Gan et al. | A novel method for palmprint recognition based on wavelet transform | |
CN106897672A (en) | A kind of human face light invariant feature extraction method of utilization logarithmic transformation and Priwitt operators | |
CN106934340A (en) | A kind of human face light invariant feature extraction method of utilization logarithmic transformation and Sobel operators | |
CN111079689B (en) | Fingerprint image enhancement method | |
CN107437061A (en) | It is a kind of to utilize logarithmic transformation and the human face light invariant feature extraction method of Roberts operators | |
CN106934342A (en) | A kind of human face light invariant feature extraction method based on Kirsch operators | |
CN106991369A (en) | A kind of human face light invariant feature extraction method based on smothing filtering | |
DR et al. | Fingerprint verification based on fusion of minutiae and ridges using strength factors | |
Ahmed et al. | A Recognition System for Subjects' Signature Using the Spatial Distribution of Signature Body | |
CN107451591A (en) | A kind of human face light invariant feature extraction method using Wallis operators |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170707 |
|
RJ01 | Rejection of invention patent application after publication |