CN108446686A - A kind of face recognition features' extraction algorithm based on image local graph structure - Google Patents
A kind of face recognition features' extraction algorithm based on image local graph structure Download PDFInfo
- Publication number
- CN108446686A CN108446686A CN201810520509.7A CN201810520509A CN108446686A CN 108446686 A CN108446686 A CN 108446686A CN 201810520509 A CN201810520509 A CN 201810520509A CN 108446686 A CN108446686 A CN 108446686A
- Authority
- CN
- China
- Prior art keywords
- pixel
- image
- value
- face recognition
- graph structure
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of face recognition features' extraction algorithms based on image local graph structure, include the following steps：In face gray level image, the neighborhood of 5 × 3 sizes is chosen from left to right and successively from top to bottom；Build local digraph structure in certain sequence using central pixel point as starting point；It is each directed edge assignment according to the size in the direction of the digraph successively gray value of more each directed edge both ends pixel, and based on given principle；The weights on all sides are multiplied by corresponding weight and obtain 16 binary system string value, finally centered on the actual numerical value of the binary string pixel characteristic value；It repeats the above steps to obtain the characteristic value of whole facial image.Reasonable design of the present invention can make full use of the relationship between each pixel in neighborhood, accurately describe the local feature of facial image, and then effectively improves discrimination；Meanwhile relative to LGS and SLGS, the utilization rate of texture information can be effectively improved, there is better robustness to image recognition.
Description
Technology neighborhood
The invention belongs to technical field of biometric identification, it is related to the algorithm for gray level image feature extraction, it is especially a kind of
Face recognition features' extraction algorithm based on image local graph structure.
Background technology
In image texture characteristic extraction algorithm, LBP is obtained as a kind of simple and very effective local grain operator
Extensive use, each pixel is compared by it with the pixel near it, and result is saved as binary string.Since LBP is calculated
Method discrimination is powerful and calculates simply, and therefore, local binary patterns texture operator is applied under different scenes.
The most important attribute of LBP algorithms is the robustness to grey scale change caused by illumination variation etc..Another of LBP algorithms
Key property is that its calculating is simple, this allows it to analyze image in real time.But it is carried out using LBP operators special
Sign extraction can lose portion image space information.
To reduce the loss of image space information, 2011, Eimad Abusham et al. were in document Face
recognition using local graph structure(LGS).Lecture Notes in Computer
Science.2011：A kind of face recognition algorithms based on local graph structure (LGS) are proposed in 169-175.The base of the algorithm
This thought is：In 3 × 4 neighborhood centered on some pixel, digraph (as shown in Figure 1) is built counterclockwise, so
The gray value of the pixel at every directed edge both ends in figure is compared afterwards, if the gray value of starting point is more than terminating point
The side is then assigned a value of 1, is otherwise assigned a value of 0 by gray value.
2014, for the deficiency of LGS algorithms, Mohd Fikri Azli bin Abdullah et al. were in document Face
recognition with Symmetric Local Graph Structure(SLGS).Expert Systems with
Applications.2014,41 (14):A kind of recognition of face being based on symmetrical graph structure (SLGS) is proposed in 6131-6137
Algorithm.The basic thought of the algorithm is：The gray level image for converting the image into 128 × 128 first, the pixel centered on certain pixel
Point, the left side of its 5 × 3 neighborhood by counterclockwise construction digraph, on the right of it by clockwise construction digraph, then according to
The principle that LGS calculates the value of directed edge calculates the value of each directed edge and obtains the corresponding grey scale value (as shown in Figure 2) of the pixel, this
Sample can extract 15624 characteristic values from every pictures.
However, either LGS or SLGS, what is mainly extracted is the texture information of central pixel point or so consecutive points,
The neighbouring texture information of the pixel has been abandoned, has led to the missing of texure information in this way, to reduce recognition of face
Efficiency.
By retrieval, not yet find and the relevant patent publication us of present patent application.
Invention content
It is an object of the invention to overcome the deficiencies in the prior art, propose that a kind of face based on image local graph structure is known
Other feature extraction algorithm solves the space of each pixel in neighborhood cannot be made full use of to believe when feature extraction in biometric identification process
Breath carries out the problem of feature extraction.
The present invention solves its technical problem and following technical scheme is taken to realize：
A kind of face recognition features' extraction algorithm based on image local graph structure, includes the following steps：
Step 1, in face gray level image, choose the neighborhood of 5 × 3 sizes successively from left to right and from top to bottom；
Step 2 builds local digraph structure using central pixel point as starting point in certain sequence；
Step 3, according to the size in the direction of the digraph successively gray value of more each directed edge both ends pixel, and be based on
Given principle is each directed edge assignment；
Step 4：The weights on all sides are multiplied by corresponding weight and obtain 16 binary system string value, finally with this two into
The characteristic value of pixel centered on the actual numerical value of system string；
Step 5 repeats step 1 to step 4, obtains the characteristic value of whole facial image.
Further, the neighborhood of 5 × 3 sizes of step 1 selection is：With X0Centered on 5 × 3 neighborhoods in each point from a left side
To right and pixel is respectively from top to bottom：X1、X2、X3、X4、X5、X6、X7、X0、X8、X9、X10、X11、X12、X13、X14、X15。
Further, the step 2 build local digraph structure be in the following order structure and it is indicated with arrows：X0、
X4、X2、X0、X7、X1、X10、X7、X0、X12、X14、X0、X8、X15、X5、X8、X0。
Further, the given principle of the step 3 is：If the gray value of the arc tail pixel of directed edge is more than arc head portrait element
Gray value, then the weights on the side are set as 0, the weights on the side are otherwise set as 1.
Further, the specific method of the step 4 is：By the weights on all sides according to X0→X4→X2→X0→X7→X1
→X10→X7→X0→X12→X14→X0→X8→X15→X5→X8→X0Direction arrangement, then turn to non-negative decimal number,
Finally to the decimal number evolution and downward rounding, central pixel point X is obtained0Characteristic value.
The advantages and positive effects of the present invention are：
1, the present invention builds digraph using 5 × 3 traditional neighborhoods, and central pixel point can preferably utilize its neighborhood in figure
The information of point indicates, to more can accurately describe face characteristic, and then effectively improves the discrimination of face, solves biology and knows
The problem of spatial information of each pixel in neighborhood cannot be made full use of to carry out feature extraction during not when feature extraction.
2, the present invention carries out feature extraction using local digraph, and makes full use of the gray scale in neighborhood between each pixel
Relationship, since the algorithm is other than the neighborhood point information on 4 angles using some pixel, neighborhood point on four angles
Texture relationship is also considered into, therefore relative to LGS and SLGS, can effectively improve the utilization rate of texture information, to figure
As identification has better robustness.
Description of the drawings
Fig. 1 is that LGS neighborhoods in the prior art and its digraph build schematic diagram；
Fig. 2 is that SLGS neighborhoods in the prior art and its digraph build schematic diagram；
Fig. 3 is that neighborhood of the present invention and its digraph build schematic diagram；
Fig. 4 is the present invention and the discrimination comparison diagram of LBP, LGS, SLGS on YALE face databases.
Specific implementation mode
The embodiment of the present invention is further described below in conjunction with attached drawing.
A kind of face recognition features' extraction algorithm based on image local graph structure is acquired with ordinary video collecting device
The color video of human face region be original signal, the extraction foundation characterized by the gray level image of original signal, and use is based on
The feature extraction algorithm of image local graph structure carries out feature extraction, and concrete methods of realizing includes the following steps：
Step 1, in face gray level image, from left to right, from top to bottom, successively choose 5 × 3 sizes neighborhood.
In this step, the ROI region for including face from the extraction segmentation of face gray level image, and gray processing are needed, then
Choose the neighborhood of 5 × 3 sizes.As shown in figure 3, selected neighborhood territory pixel (16 pixels from left to right and from top to bottom：X1、
X2、X3、X4、X5、X6、X7、X0、X8、X9、X10、X11、X12、X13、X14、X15) corresponding grey scale value be respectively 54,62,92,84,59,
7、82、55、15、84、5、17、62、42、55。
Step 2, in selected neighborhood according to X0→X4→X2→X0→X7→X1→X10→X7→X0→X12→X14→
X0→X8→X15→X5→X8→X0Direction build digraph, as shown in Figure 3.
Step 3, according to the size of graph structure successively more every directed edge arc head and the gray value of arc tail pixel, and press
According to " if the gray value of the arc tail pixel of directed edge is more than the gray value of arc head portrait element, the weights on the side are set as 0, it is no
It is then set as 1 " principle, it is respectively 1,0,0,1,0,0,1,0,0,1,1,0,1,1,0,1 to obtain binary system string value.
Step 4 arranges the sequence of the binary system string value big-endian obtained in step 3, obtains one 16
Position binary string 1001001001101101, it is 37485 to be then converted into decimal number, is then carried out to the decimal number
Evolution and downward rounding, obtain 193, finally by 193 characteristic value as central pixel point in neighborhood.
This step is the binary string that the weights for 16 directed edges that step 3 obtains sequentially are formed to one 16, wherein
Weights are combined by the sequence of step 3 from a high position to low level.Since the practical greatest measure of obtained binary string is 215+
214+…+21+20, that is, 65535.Since numerical value is excessive, be unfavorable for calculating, thus by until characteristic value evolution and downward
Rounding, the maximum eigenvalue for finally obtaining central pixel point are 255.
Step 5 repeats step 1 to step 4, obtains the characteristic value of whole facial image.
By experimental result further verification can be done to the effect of the present invention.As shown in figure 4, on YALE databases,
When training sample data take 4,5,6,7,8, the discrimination of (ILGS) of the invention is above LBP, LGS, SLGS algorithm.Later,
Gradually increase with the quantity of training sample, the discrimination of face is also constantly being increased, is being reached when the quantity of training sample takes
Total number of samples amount 80% when, the discrimination of (ILGS) algorithm of the invention then reaches 99.02%, LBP, LGS and SLGS of the same period
The discrimination of algorithm is respectively 93.32%, 96.99%, 97.56%.
It is not filled in conclusion the present invention for LBP, LGS and SLGS algorithm, overcomes their feature information extractions
The problem of dividing, make the feature of extraction that more fully face be described.Therefore the discrimination of facial image is superior to
State algorithm.
It is emphasized that embodiment of the present invention is illustrative, without being restrictive, therefore packet of the present invention
Include the embodiment being not limited to described in specific implementation mode, it is every by this neighborhood technique personnel according to the technique and scheme of the present invention
The other embodiment obtained, also belongs to the scope of protection of the invention.
Claims (5)
1. a kind of face recognition features' extraction algorithm based on image local graph structure, it is characterised in that include the following steps：
Step 1, in face gray level image, choose the neighborhood of 5 × 3 sizes successively from left to right and from top to bottom；
Step 2 builds local digraph structure using central pixel point as starting point in certain sequence；
Step 3, according to the size in the direction of the digraph successively gray value of more each directed edge both ends pixel, and based on given
Principle is each directed edge assignment；
Step 4：The weights on all sides are multiplied by corresponding weight and obtain 16 binary system string value, finally with the binary string
Actual numerical value centered on pixel characteristic value；
Step 5 repeats step 1 to step 4, obtains the characteristic value of whole facial image.
2. a kind of face recognition features' extraction algorithm based on image local graph structure according to claim 1, feature
It is：The neighborhood that the step 1 chooses 5 × 3 sizes is：With X0Centered on 5 × 3 neighborhoods in each point from left to right and from upper
It is respectively to lower pixel：X1、X2、X3、X4、X5、X6、X7、X0、X8、X9、X10、X11、X12、X13、X14、X15。
3. a kind of face recognition features' extraction algorithm based on image local graph structure according to claim 1, feature
It is：The step 2 build local digraph structure be in the following order structure and it is indicated with arrows：X0、X4、X2、X0、X7、
X1、X10、X7、X0、X12、X14、X0、X8、X15、X5、X8、X0。
4. a kind of face recognition features' extraction algorithm based on image local graph structure according to claim 1, feature
It is：The given principle of the step 3 is：If the gray value of the arc tail pixel of directed edge is more than the gray value of arc head portrait element,
The weights on the side are then set as 0, the weights on the side are otherwise set as 1.
5. a kind of face recognition features' extraction algorithm based on image local graph structure according to claim 1, feature
It is：The specific method of the step 4 is：By the weights on all sides according to X0→X4→X2→X0→X7→X1→X10→X7→
X0→X12→X14→X0→X8→X15→X5→X8→X0Direction arrangement, then turn to non-negative decimal number, finally to this ten
System number evolution and downward rounding, obtain central pixel point X0Characteristic value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810520509.7A CN108446686A (en) | 2018-05-28 | 2018-05-28 | A kind of face recognition features' extraction algorithm based on image local graph structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810520509.7A CN108446686A (en) | 2018-05-28 | 2018-05-28 | A kind of face recognition features' extraction algorithm based on image local graph structure |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108446686A true CN108446686A (en) | 2018-08-24 |
Family
ID=63204645
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810520509.7A Pending CN108446686A (en) | 2018-05-28 | 2018-05-28 | A kind of face recognition features' extraction algorithm based on image local graph structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108446686A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111931590A (en) * | 2020-07-15 | 2020-11-13 | 杭州电子科技大学 | Balanced face feature extraction method of self-adaptive four-corner star local graph structure |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056059A (en) * | 2016-05-20 | 2016-10-26 | 合肥工业大学 | Multidirectional SLGS characteristic description and performance cloud weight fusion face recognition method |
| CN107358147A (en) * | 2017-05-22 | 2017-11-17 | 天津科技大学 | Face recognition features' extraction algorithm based on local circulation graph structure |
| CN107862267A (en) * | 2017-10-31 | 2018-03-30 | 天津科技大学 | Face recognition features' extraction algorithm based on full symmetric local weber description |
-
2018
- 2018-05-28 CN CN201810520509.7A patent/CN108446686A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056059A (en) * | 2016-05-20 | 2016-10-26 | 合肥工业大学 | Multidirectional SLGS characteristic description and performance cloud weight fusion face recognition method |
| CN107358147A (en) * | 2017-05-22 | 2017-11-17 | 天津科技大学 | Face recognition features' extraction algorithm based on local circulation graph structure |
| CN107862267A (en) * | 2017-10-31 | 2018-03-30 | 天津科技大学 | Face recognition features' extraction algorithm based on full symmetric local weber description |
Non-Patent Citations (2)
| Title |
|---|
| MOHD FIKRI AZLI ABDULLAH 等: "Face recognition with Symmetric Local Graph Structure", 《EXPERT SYSTEMS WITH APPLICATIONS》 * |
| SONG DONG 等: "Finger Vein Recognition Based on Multi-Orientation Weighted Symmetric Local Graph Structure", 《KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111931590A (en) * | 2020-07-15 | 2020-11-13 | 杭州电子科技大学 | Balanced face feature extraction method of self-adaptive four-corner star local graph structure |
| CN111931590B (en) * | 2020-07-15 | 2023-09-29 | 杭州电子科技大学 | Balanced face feature extraction method of self-adaptive four-corner star-shaped partial graph structure |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100461204C (en) | Method for recognizing facial expression based on 2D partial least square method | |
| CN104063706B (en) | Video fingerprint extraction method based on SURF algorithm | |
| CN107316031A (en) | The image characteristic extracting method recognized again for pedestrian | |
| CN103295197B (en) | Based on the image super-resolution rebuilding method of dictionary learning and bilateral canonical | |
| CN104361096B (en) | The image search method of a kind of feature based rich region set | |
| WO2015024383A1 (en) | Similarity acquisition method for colour distribution and texture distribution image retrieval | |
| CN104021523B (en) | A kind of method of the image super-resolution amplification based on marginal classification | |
| CN104598933A (en) | Multi-feature fusion based image copying detection method | |
| CN103984963B (en) | Method for classifying high-resolution remote sensing image scenes | |
| CN107358147A (en) | Face recognition features' extraction algorithm based on local circulation graph structure | |
| CN109242796A (en) | Character image processing method, device, electronic equipment and computer storage medium | |
| Kumar et al. | Near lossless image compression using parallel fractal texture identification | |
| CN103839074B (en) | Image classification method based on matching of sketch line segment information and space pyramid | |
| CN103778430B (en) | Rapid face detection method based on combination between skin color segmentation and AdaBoost | |
| CN110399760A (en) | A kind of batch two dimensional code localization method, device, electronic equipment and storage medium | |
| CN111652240A (en) | Image local feature detection and description method based on CNN | |
| CN110287847A (en) | Vehicle grading search method based on Alexnet-CLbpSurf multiple features fusion | |
| CN104463091B (en) | A kind of facial image recognition method based on image LGBP feature subvectors | |
| CN106897724A (en) | A kind of plant leaf identification method based on contour line shape facility | |
| CN104881668A (en) | Method and system for extracting image fingerprints based on representative local mode | |
| CN107292821A (en) | A kind of super-resolution image reconstruction method and system | |
| Kim et al. | Ancient coin recognition based on spatial coding | |
| CN108446686A (en) | A kind of face recognition features' extraction algorithm based on image local graph structure | |
| CN113688821A (en) | OCR character recognition method based on deep learning | |
| CN104063855A (en) | Super-resolution image reconstruction method and device based on classified dictionary database |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180824 |