CN104021372A - Face recognition method and device thereof - Google Patents

Face recognition method and device thereof Download PDF

Info

Publication number
CN104021372A
CN104021372A CN201410213460.2A CN201410213460A CN104021372A CN 104021372 A CN104021372 A CN 104021372A CN 201410213460 A CN201410213460 A CN 201410213460A CN 104021372 A CN104021372 A CN 104021372A
Authority
CN
China
Prior art keywords
cbp
image
ldp
operator
histogram
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
CN201410213460.2A
Other languages
Chinese (zh)
Inventor
叶剑英
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.)
Beijing Ingenic Semiconductor Co Ltd
Original Assignee
Beijing Ingenic Semiconductor 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 Beijing Ingenic Semiconductor Co Ltd filed Critical Beijing Ingenic Semiconductor Co Ltd
Priority to CN201410213460.2A priority Critical patent/CN104021372A/en
Publication of CN104021372A publication Critical patent/CN104021372A/en
Pending legal-status Critical Current

Links

Abstract

The invention discloses a face recognition method and a device thereof so as to raise real-time performance and accuracy of face recognition. The face recognition method provided by the embodiment of the invention comprises the following steps: CBP histograms of an image to be recognized are determined based on a CBP operator, and the CBP operator determines a CBP code value of central pixel points according to gray value of the central pixel points within the pixel 3*3 neighborhood and gray value of diagonal element; LDP histograms of the image to be recognized are determined based on a LDP operator; the CBP histograms and the LDP histograms are connected in order to form a combined histogram which is used as an extracted feature; and face recognition is carried out on the image to be recognized based on the extracted feature.

Description

A kind of face identification method and device
Technical field
The present invention relates to biometrics identification technology field, relate in particular to a kind of face identification method and device.
Background technology
At society, identity validation has very important application.In recent years, the mankind's biological characteristic is more and more widely used in individual identity validation, and the intrinsic biological characteristic of the mankind mainly comprises: DNA (DNA (deoxyribonucleic acid)), fingerprint, iris, voice, gait, palmmprint, people's face etc.Recognition of face is a kind of biometrics identification technology that face feature information based on people is carried out identity validation.The image that contains people's face with video camera or camera collection or video, and detection and tracking people face in image automatically, and then the people's face detecting is carried out to a series of correlation techniques of face.Compare other biometrics identification technology, recognition of face has following advantage:
(1) without user, too much participate in, contactless collection, without the property invaded;
(2) to user without any obvious stimulation, be convenient to hide;
(3) equipment cost is cheap, is mainly to adopt video camera or camera to gather people's face.
Therefore, face recognition technology has obtained more and more people's concern and research.But, the committed step in face recognition technology---feature extraction, but, because be subject to the impact of the factors such as illumination, attitude, expression, dress ornament, age, its robustness need to improve.
LBP (Local Binary Patterns, local binary patterns) is a kind of operator that is used for Description Image Local textural feature; Its effect is to carry out feature extraction, and the feature of extracting is the Local textural feature of image.At present, LBP operator has been successfully applied to recognition of face field.
Initial LBP operator is that in image, each pixel has defined pixel 3 * 3 neighborhoods centered by this pixel, the gray-scale value of central pixel point in pixel 3 * 3 neighborhoods of take is threshold value, the gray-scale value of 8 adjacent surrounding pixel points is compared with it, if the gray-scale value of surrounding pixel point is greater than the gray-scale value of central pixel point, this surrounding pixel point is labeled as to 1, otherwise is labeled as 0.Like this, 8 surrounding pixel points in pixel 3 * 3 neighborhoods are through relatively producing 8 bits, conventionally 8 bits are converted to decimal number, obtain the LBP code value of central pixel point in this neighborhood, and with this LBP code value, reflect the texture information in this region.The computing method of LBP operator are as shown in formula [1]:
LBP ( x c , y c ) = Σ p = 0 7 2 p s ( g p - g c ) - - - [ 1 ]
Wherein, (x c, y c) represent the central pixel point in pixel 3 * 3 neighborhoods, g cthe gray-scale value that represents central pixel point, g pthe gray-scale value that represents surrounding pixel point, s (x) represents sign function, definition is as shown in formula [2]:
Describe for example, Fig. 1 has described concrete LBP operator, in pixel 3 * 3 neighborhoods, the gray-scale value " 5 " of central pixel point is threshold value, the gray-scale value of 8 adjacent surrounding pixel points is compared with it successively, produce 8 bits " 00010011 ", it is " 19 " that 8 bits are converted to the LBP code value that decimal number obtain central pixel point in these pixel 3 * 3 neighborhoods.
LBP operator is to extract Local textural feature as distinguishing rule, and its significant advantage is to illumination-insensitive, and still, the calculation cost of LBP operator is very large, is difficult to real-time characteristic and extracts, and causes the real-time of the recognition of face based on LBP operator poor; And LBP operator is poor to noise robustness, be subject to especially the impact of white Gaussian noise, do not consider the effect of central pixel point simultaneously and use the factors such as half-tone information is relatively unstable, cause the accuracy of the recognition of face based on LBP operator low.
Summary of the invention
The embodiment of the present invention provides a kind of face identification method and device, in order to promote real-time and the accuracy of recognition of face.
The face identification method that the embodiment of the present invention provides, comprising:
Based on centralization local binary patterns CBP operator, determine the CBP histogram of image to be identified, described CBP operator is determined the CBP code value of central pixel point according to the gray-scale value of central pixel point in pixel 3 * 3 neighborhoods and the gray-scale value of diagonal element; And
Based on local orientation's pattern LDP operator, determine the LDP histogram of described image to be identified;
Described CBP histogram and LDP histogram are linked in sequence to formation joint histogram as the feature of extracting; And
Feature based on extracting is carried out recognition of face to described image to be identified.
The face identification method that the embodiment of the present invention provides and device, comprising:
CBP processing module, for determine the CBP histogram of image to be identified based on centralization local binary patterns CBP operator, described CBP operator is determined the CBP code value of central pixel point according to the gray-scale value of central pixel point in pixel 3 * 3 neighborhoods and the gray-scale value of diagonal element;
LDP processing module, for determining the LDP histogram of described image to be identified based on local orientation's pattern LDP operator;
Union feature module, for being linked in sequence formation joint histogram as the feature of extracting using described CBP histogram and LDP histogram;
Face recognition module, carries out recognition of face for the feature based on extracting to described image to be identified.
The face identification method that the embodiment of the present invention provides and device, new CBP operator has been proposed on the basis of existing LBP operator, simultaneously in conjunction with LDP operator, taken into full account the effect of central pixel point, and use the half-tone information of diagonal element relatively stable, noise is had to good robustness, thereby effectively promoted the accuracy of recognition of face; CBP operator is compared LBP operator with LDP operator, dimension has had obvious reduction, thereby has significantly reduced calculated amount and computation complexity, and the while is the textural characteristics of Description Image effectively, be applicable to real-time characteristic extraction, thereby effectively promoted the real-time of recognition of face.
The application's further feature and advantage will be set forth in the following description, and, partly from instructions, become apparent, or understand by implementing the application.The application's object and other advantages can be realized and be obtained by specifically noted structure in the instructions write, claims and accompanying drawing.
Accompanying drawing explanation
Accompanying drawing is used to provide a further understanding of the present invention, and forms a part for instructions, is used from explanation the present invention with the embodiment of the present invention one, is not construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the computing method schematic diagram of LBP operator in prior art;
Fig. 2 is face identification method process flow diagram in the embodiment of the present invention;
Fig. 3 is the face identification method process flow diagram based on CBP operator and LDP operator in the embodiment of the present invention;
Fig. 4 is pixel 3 * 3 neighborhood subimage schematic diagram of Kirsch operator in the embodiment of the present invention;
Fig. 5 is the computing method schematic diagram of Kirsch operator and LDP operator in the embodiment of the present invention;
Fig. 6 is a kind of possibility structured flowchart of face identification device in the embodiment of the present invention;
Fig. 7 is a kind of possibility structured flowchart of CBP processing module in the embodiment of the present invention;
Fig. 8 is that a kind of of LDP processing module in the embodiment of the present invention may structured flowchart.
Embodiment
The embodiment of the present invention provides a kind of face identification method and device, in order to promote real-time and the accuracy of recognition of face.Below in conjunction with Figure of description, the preferred embodiments of the present invention are described, should be appreciated that preferred embodiment described herein, only for description and interpretation the present invention, is not intended to limit the present invention.And in the situation that not conflicting, embodiment and the feature in embodiment in the application can combine mutually.
First operator face identification method in the embodiment of the present invention being adopted describes.
1, CBP operator (centralization LBP operator): because LBP operator does not fully take into account the effect of central pixel point, for this situation embodiment of the present invention, centralization LBP operator has been proposed, in the embodiment of the present invention by centralization LBP operator called after CBP operator.CBP is a kind of operator that is used for Description Image Local textural feature, and CBP operator adopts " diagonal angle principle ", calculates the gray-scale value of diagonal element, and taken into full account the effect of central pixel point, draw CBP code value, and then obtain CBP histogram by CBP code value, as the feature of extracting.The dimension of CBP operator has had obvious reduction, and the speed of therefore calculating is improved.
2, LDP (Local Directional Pattern, local orientation's pattern) operator: LDP is a kind of operator that is used for Description Image Local textural feature, by calculating the skirt response value of 8 directions, calculate LDP code value, and then obtain LDP histogram by LDP code value, as the feature of extracting.
3, Kirsch operator: Kirsch operator is by 8 directional operators that template forms that represent 8 directions, for 3 * 3 neighborhoods of pixel in image, process, make each pixel in image adopt 8 templates to carry out convolutional calculation, wherein maximal value can be used as the edge output of image.
Introduction based on operator, the face identification method that the embodiment of the present invention provides, as shown in Figure 2, comprises the steps:
S201, based on CBP operator, determine the CBP histogram of image to be identified, wherein, CBP operator is determined the CBP code value of central pixel point according to the gray-scale value of central pixel point in pixel 3 * 3 neighborhoods and the gray-scale value of diagonal element;
In the concrete enforcement of S201, in order to reduce computation complexity, a kind of preferably implementation method, can first carry out piecemeal processing by this image to be identified, obtains non-overlapping some image subblocks; Based on CBP operator, determine respectively the CBP histogram of each image subblock again; The CBP histogram of this image to be identified of formation is finally linked in sequence the CBP histogram of each image subblock.
S202, based on LDP operator, determine the LDP histogram of the image that this is to be identified.
S203, CBP histogram and LDP histogram are linked in sequence to formation joint histogram as the feature of extracting.
S204, the feature based on extracting are carried out recognition of face to image to be identified;
In the concrete enforcement of S204, carrying out the sorting algorithm that recognition of face adopts generally can comprise: KNN (k-Nearest Neighbor algorithm, arest neighbors sorting algorithm), algorithm of support vector machine, PCN (airport composite runway pavement algorithm) etc., the sorting algorithm adopting belongs to the category of prior art, specifically repeats no more.
It should be noted that, be only to consider based on describing easily in the embodiment of the present invention, and each step is provided to serial number, one of ordinary skill in the art will appreciate that, S201 and S202 there is no strict sequential context.
Below, the face identification method based on CBP operator and LDP operator providing in connection with 3 pairs of embodiment of the present invention of Figure of description is elaborated.
S301, image to be identified is carried out to piecemeal processing, obtain non-overlapping some image subblocks;
In concrete enforcement, the size of piece can be set based on experience value flexibly.
S302, based on CBP operator, determine the CBP histogram of each image subblock;
In concrete enforcement, for each image subblock, be handled as follows respectively:
Step 321, based on CBP operator, determine the CBP code value of each pixel in present image sub-block, concrete, for each pixel in present image sub-block, in pixel 3 * 3 neighborhoods centered by current pixel point, determine the CBP code value of current pixel point, the method for determining is as shown in formula [3]:
CBP ( x c , y c ) = Σ p = 0 3 s ( g p - g p + 4 ) 2 p + s ( g c - g m ) 2 4 - - - [ 3 ]
Wherein, (x c, y c) represent the central pixel point in pixel 3 * 3 neighborhoods, g cthe gray-scale value that represents central pixel point, g pthe gray-scale value that represents surrounding pixel point, s (x) represents sign function, defines as shown in formula [2] g mthe gray-scale value average that represents each pixel in pixel 3 * 3 neighborhoods, definition is as shown in formula [4]:
g m = ( Σ i = 0 7 g i + g c ) / 9 - - - [ 4 ]
Step 322, according to the CBP code value of each pixel in present image sub-block, obtain the CBP histogram of present image sub-block;
Obviously, the histogrammic dimension of CBP is 32, with respect to the histogrammic dimension of LBP, is 256, and histogram dimension is greatly reduced.The histogrammic computing method of CBP of present image sub-block are as shown in formula [5]:
H CBP i = Σ x , y f ( CBP ( x , y ) , c i ) - - - [ 5 ]
Wherein, c icBP code value within the scope of the CBP code value that represents to set.
S303, by the be linked in sequence CBP histogram of formation image to be identified of the CBP histogram of each image subblock;
Step S304, based on LDP operator, determine the LDP histogram of image to be identified;
In concrete enforcement, based on LDP operator, determine that the LDP histogram of image to be identified comprises following treatment step:
Step 341, for each pixel in image to be identified, in pixel 3 * 3 neighborhoods centered by current pixel point, use Kirsch operator to carry out convolutional calculation and obtain skirt response value;
Conventionally Kirsch operator is comprised of 8 templates, and each pixel is used after these 8 template computings, selects wherein maximal value as the edge strength of this pixel.Pixel 3 * 3 neighborhood subimages centered by current pixel point, as shown in Figure 4, Kirsch operator is investigated the grey scale change of its eight surrounding pixel points to current pixel point, deduct the gray-scale value weighted sum of remaining five surrounding pixel points with the gray-scale value weighted sum of three surrounding pixel points wherein; Make three surrounding pixel points around continuous displacement, get the maximal value of its difference as Kirsch operator value, i.e. the edge strength of current pixel point.Computing method are as shown in formula [6]:
m(i，j)＝max{1，max{|5s k-3t k|：k＝0，1，…，7}} [6]
Wherein: s k=a k+ a k+1+ a k+2, t k=a k+3+ a k+4+ ... + a k+7, a kthe gray-scale value that represents surrounding pixel point, if subscript employing over 7 is determined divided by the mode of 8 remainder numbers.
Step 342, according to the skirt response value of using Kirsch operator to calculate, based on LDP operator, determine the LDP code value of current pixel point, the method for determining is as shown in formula [7]:
LDP k = Σ i = 0 7 s i ( m i - m k ) 2 i - - - [ 7 ]
Wherein, m krepresent k the skirt response value of arranging according to size order in each skirt response value.
Describe for example, as shown in Figure 5, in pixel 3 * 3 neighborhoods centered by pixel a (i, j)=50,8 surrounding pixel points of this pixel are respectively: a 0=10, a 1=26, a 2=32, a 3=85, a 4=53, a 5=60, a 6=38, a 7=45, by Kirsch operator edge calculation response, be respectively: 503,97,313,537,161,97,161,393; Wherein make k=3, k skirt response value is 393, thereby the LDP code value of determining current pixel point is that 19 (binary number is: 00010011).
Step 343, according to the LDP code value of each pixel in image to be identified, obtain the LDP histogram of image to be identified, when k=3, the histogrammic dimension of LDP is 56;
The histogrammic computing method of LDP of image to be identified are as shown in formula [8]:
H LDP i = Σ x , y f ( LD P k ( x , y ) , c i ) - - - [ 8 ]
Wherein, c ilDP code value within the scope of the LDP code value that represents to set.
S305, CBP histogram and LDP histogram are linked in sequence to formation joint histogram as the feature of extracting;
S306, the feature based on extracting, adopt KNN to carry out recognition of face to image to be identified;
In concrete enforcement, when the feature based on extracting is carried out recognition of face, can adopt various existing sorting algorithms, preferably, adopt the simplest KNN can reach requirement.KNN is carried out to brief description below:
KNN is one of the simplest machine learning algorithm, adopts vector space model to classify, and due to the example of identical category, similarity is each other high, therefore can, by calculating the similarity with known class example, assess the possible classification of unknown classification example.KNN classifies according to the similarity between some sample instance and other examples.The example of feature similarity is adjacent to each other, the dissimilar example of feature mutually away from.Thereby distance that can be using between two examples is as a kind of module of their " dissimilar degree ".KNN can support two kinds of method calculated examples spacings, respectively: Euclidean Distance (Euclidean distance method) and City-block Distance (city Furthest Neighbor).
Based on same technical conceive, the embodiment of the present invention provides a kind of face identification device, and because the principle that this device is dealt with problems is consistent with face identification method, so the enforcement of this device can, referring to the enforcement of method, repeat part and not repeat.
As shown in Figure 6, the face identification device that the embodiment of the present invention provides, comprising:
CBP processing module 601, for determine the CBP histogram of image to be identified based on CBP operator, wherein, CBP operator is determined the CBP code value of central pixel point according to the gray-scale value of central pixel point in pixel 3 * 3 neighborhoods and the gray-scale value of diagonal element;
LDP processing module 602, for determining the LDP histogram of the image that this is to be identified based on LDP operator;
Union feature module 603, for being linked in sequence formation joint histogram as the feature of extracting using CBP histogram and LDP histogram;
Face recognition module 604, carries out recognition of face for the feature based on extracting to image to be identified.
In concrete enforcement, a kind of possibility structure of CBP processing module 601, as shown in Figure 7, specifically can comprise:
Piecemeal processing unit 701, for image to be identified is carried out to piecemeal processing, obtains non-overlapping some image subblocks;
CBP processing unit 702, for determining respectively the CBP histogram of each image subblock based on CBP operator;
Merge cells 703, for the CBP histogram of this image to be identified of formation that the CBP histogram of each image subblock is linked in sequence.
Wherein, CBP processing unit 702, specifically can comprise:
CBP operator is processed subelement, for for each image subblock, determines the CBP code value of each pixel in present image sub-block based on CBP operator, specifically passes through formula CBP ( x c , y c ) = Σ p = 0 3 s ( g p - g p + 4 ) 2 p + s ( g c - g m ) 2 4 Realize, wherein: (x c, y c) represent the central pixel point in pixel 3 * 3 neighborhoods, g cthe gray-scale value that represents central pixel point, g pthe gray-scale value that represents surrounding pixel point, s (x) represents sign function, specifically passes through formula realize g mrepresent the gray-scale value average of each pixel in pixel 3 * 3 neighborhoods, specifically pass through formula realize;
CBP histogram generates subelement, for according to the CBP code value of each pixel of present image sub-block, obtains the CBP histogram of present image sub-block, specifically passes through formula realize, wherein: c icBP code value within the scope of the CBP code value that represents to set.
In concrete enforcement, a kind of possibility structure of LDP processing module 602, as shown in Figure 8, specifically can comprise:
Kirsch operator processing unit 801, for each pixel of the image for to be identified, in pixel 3 * 3 neighborhoods centered by current pixel point, use Kirsch operator to carry out convolutional calculation and obtain skirt response value, concrete by formula m (i, j)=max{1, max{|5s k-3t k|: k=0,1 ..., 7}} realizes, wherein: s k=a k+ a k+1+ a k+2, t k=a k+3+ a k+4+ ... + a k+7, a kthe gray-scale value that represents surrounding pixel point;
LDP operator processing unit 802, for according to the skirt response value of using Kirsch operator to obtain, determines the LDP code value of current pixel point based on LDP operator, specifically pass through formula realize, wherein: s (x) represents sign function, specifically passes through formula m krepresent k the skirt response value of arranging according to size order in each skirt response value;
LDP histogram generation unit 803, for according to the LDP code value of each pixel of image to be identified, obtains the LDP histogram of image to be identified, specifically passes through formula realize, wherein: c ilDP code value within the scope of the LDP code value that represents to set.
In concrete enforcement, face recognition module 604, specifically for the feature based on extracting, adopts KNN to carry out recognition of face to image to be identified.
The face identification device that the application's embodiment provides can be realized by computer program.Those skilled in the art should be understood that; above-mentioned Module Division mode is only a kind of in numerous Module Division modes; if be divided into other modules or do not divide module, as long as the through server of search has above-mentioned functions, all should be within the application's protection domain.
The face identification method that the embodiment of the present invention provides and device, new CBP operator has been proposed on the basis of existing LBP operator, simultaneously in conjunction with LDP operator, taken into full account the effect of central pixel point, and use the half-tone information of diagonal element relatively stable, noise is had to good robustness, thereby effectively promoted the accuracy of recognition of face; CBP operator is compared LBP operator with LDP operator, dimension has had obvious reduction, thereby has significantly reduced calculated amount and computation complexity, and the while is the textural characteristics of Description Image effectively, be applicable to real-time characteristic extraction, thereby effectively promoted the real-time of recognition of face.
It will be understood by those skilled in the art that embodiments of the invention can be provided as method, system, equipment or computer program.Therefore, the present invention can adopt complete hardware implementation example, implement software example or in conjunction with the form of the embodiment of software and hardware aspect completely.And the present invention can adopt the form that wherein includes the upper computer program of implementing of computer-usable storage medium (including but not limited to magnetic disk memory, CD-ROM, optical memory etc.) of computer usable program code one or more.
The present invention is with reference to describing according to process flow diagram and/or the block scheme of the method for the embodiment of the present invention, equipment (system) and computer program.Should understand can be in computer program instructions realization flow figure and/or block scheme each flow process and/or the flow process in square frame and process flow diagram and/or block scheme and/or the combination of square frame.Can provide these computer program instructions to the processor of multi-purpose computer, special purpose computer, Embedded Processor or other programmable data processing device to produce a machine, the instruction of carrying out by the processor of computing machine or other programmable data processing device is produced for realizing the device in the function of flow process of process flow diagram or a plurality of flow process and/or square frame of block scheme or a plurality of square frame appointments.
These computer program instructions also can be stored in energy vectoring computer or the computer-readable memory of other programmable data processing device with ad hoc fashion work, the instruction that makes to be stored in this computer-readable memory produces the manufacture that comprises command device, and this command device is realized the function of appointment in flow process of process flow diagram or a plurality of flow process and/or square frame of block scheme or a plurality of square frame.
These computer program instructions also can be loaded in computing machine or other programmable data processing device, make to carry out sequence of operations step to produce computer implemented processing on computing machine or other programmable devices, thereby the instruction of carrying out is provided for realizing the step of the function of appointment in flow process of process flow diagram or a plurality of flow process and/or square frame of block scheme or a plurality of square frame on computing machine or other programmable devices.
Although described the preferred embodiments of the present invention, once those skilled in the art obtain the basic creative concept of cicada, can make other change and modification to these embodiment.So claims are intended to all changes and the modification that are interpreted as comprising preferred embodiment and fall into the scope of the invention.
Obviously, those skilled in the art can carry out various changes and modification and not depart from the spirit and scope of the present invention the present invention.Like this, if within of the present invention these are revised and modification belongs to the scope of the claims in the present invention and equivalent technologies thereof, the present invention is also intended to comprise these changes and modification interior.

Claims (10)

1. a face identification method, is characterized in that, comprising:
Based on centralization local binary patterns CBP operator, determine the CBP histogram of image to be identified, described CBP operator is determined the CBP code value of central pixel point according to the gray-scale value of central pixel point in pixel 3 * 3 neighborhoods and the gray-scale value of diagonal element; And
Based on local orientation's pattern LDP operator, determine the LDP histogram of described image to be identified;
Described CBP histogram and LDP histogram are linked in sequence to formation joint histogram as the feature of extracting; And
Feature based on extracting is carried out recognition of face to described image to be identified.
2. the method for claim 1, is characterized in that, determines the CBP histogram of image to be identified based on centralization local binary patterns CBP operator, specifically comprises:
Described image to be identified is carried out to piecemeal processing, obtain non-overlapping some image subblocks;
Based on CBP operator, determine respectively the CBP histogram of each image subblock; And
By the be linked in sequence CBP histogram of the described image to be identified of formation of the CBP histogram of each image subblock.
3. method as claimed in claim 2, is characterized in that, the described CBP histogram of determining respectively each image subblock based on CBP operator, specifically comprises:
For each image subblock, based on CBP operator, determine the CBP code value of each pixel in present image sub-block, specifically pass through formula CBP ( x c , y c ) = Σ p = 0 3 s ( g p - g p + 4 ) 2 p + s ( g c - g m ) 2 4 Realize, wherein:
(x c, y c) represent the central pixel point in pixel 3 * 3 neighborhoods, g cthe gray-scale value that represents central pixel point, g pthe gray-scale value that represents surrounding pixel point, s (x) represents sign function, specifically passes through formula realize g mrepresent the gray-scale value average of each pixel in pixel 3 * 3 neighborhoods, specifically pass through formula g m = ( Σ i = 0 7 g i + g c ) / 9 Realize;
According to the CBP code value of each pixel in present image sub-block, obtain the CBP histogram of present image sub-block, specifically pass through formula realize, wherein:
c icBP code value within the scope of the CBP code value that represents to set.
4. the method for claim 1, is characterized in that, the described LDP histogram of determining described image to be identified based on local orientation's pattern LDP operator, specifically comprises:
For each pixel in described image to be identified, in pixel 3 * 3 neighborhoods centered by current pixel point, use Kirsch operator to carry out convolutional calculation and obtain skirt response value, specifically by formula m (i, j)=max{1, max{|5s k-3t k|: k=0,1 ..., 7}} realizes, wherein:
S k=a k+ a k+1+ a k+2, t k=a k+3+ a k+4+ ... + a k+7, a kthe gray-scale value that represents surrounding pixel point;
According to the skirt response value of using Kirsch operator to calculate, based on LDP operator, determine the LDP code value of current pixel point, specifically pass through formula realize, wherein:
S (x) represents sign function, specifically passes through formula realize m krepresent k the skirt response value of arranging according to size order in each skirt response value;
According to the LDP code value of each pixel in image to be identified, obtain the LDP histogram of image to be identified, specifically pass through formula realize, wherein:
c ilDP code value within the scope of the LDP code value that represents to set.
5. the method as described in as arbitrary in claim 1 to 4, is characterized in that, the described feature based on extracting is carried out recognition of face to described image to be identified, specifically comprises:
Feature based on extracting, adopts arest neighbors sorting algorithm KNN to carry out recognition of face to described image to be identified.
6. a face identification device, is characterized in that, comprising:
CBP processing module, for determine the CBP histogram of image to be identified based on centralization local binary patterns CBP operator, described CBP operator is determined the CBP code value of central pixel point according to the gray-scale value of central pixel point in pixel 3 * 3 neighborhoods and the gray-scale value of diagonal element;
LDP processing module, for determining the LDP histogram of described image to be identified based on local orientation's pattern LDP operator;
Union feature module, for being linked in sequence formation joint histogram as the feature of extracting using described CBP histogram and LDP histogram;
Face recognition module, carries out recognition of face for the feature based on extracting to described image to be identified.
7. device as claimed in claim 6, is characterized in that, described CBP processing module, specifically comprises:
Piecemeal processing unit, for described image to be identified is carried out to piecemeal processing, obtains non-overlapping some image subblocks;
CBP processing unit, for determining respectively the CBP histogram of each image subblock based on CBP operator;
Merge cells, for the CBP histogram of the described image to be identified of formation that the CBP histogram of each image subblock is linked in sequence.
8. device as claimed in claim 7, is characterized in that, described CBP processing unit, specifically comprises:
CBP operator is processed subelement, for for each image subblock, determines the CBP code value of each pixel in present image sub-block based on CBP operator, specifically passes through formula CBP ( x c , y c ) = Σ p = 0 3 s ( g p - g p + 4 ) 2 p + s ( g c - g m ) 2 4 Realize, wherein: (x c, y c) represent the central pixel point in pixel 3 * 3 neighborhoods, g cthe gray-scale value that represents central pixel point, g pthe gray-scale value that represents surrounding pixel point, s (x) represents sign function, specifically passes through formula realize g mrepresent the gray-scale value average of each pixel in pixel 3 * 3 neighborhoods, specifically pass through formula realize;
CBP histogram generates subelement, for according to the CBP code value of each pixel of present image sub-block, obtains the CBP histogram of present image sub-block, specifically passes through formula realize, wherein: c icBP code value within the scope of the CBP code value that represents to set.
9. device as claimed in claim 6, is characterized in that, LDP processing module, specifically comprises:
Kirsch operator processing unit, for for described each pixel of image to be identified, in pixel 3 * 3 neighborhoods centered by current pixel point, use Kirsch operator to carry out convolutional calculation and obtain skirt response value, concrete by formula m (i, j)=max{1, max{|5s k-3t k|: k=0,1 ..., 7}} realizes, wherein: s k=a k+ a k+1+ a k+2, t k=a k+3+ a k+4+ ... + a k+7, a kthe gray-scale value that represents surrounding pixel point;
LDP operator processing unit, for according to the skirt response value of using Kirsch operator to obtain, determines the LDP code value of current pixel point based on LDP operator, specifically pass through formula realize, wherein: s (x) represents sign function, specifically passes through formula realize m krepresent k the skirt response value of arranging according to size order in each skirt response value;
LDP histogram generation unit, for according to the LDP code value of each pixel of image to be identified, obtains the LDP histogram of image to be identified, specifically passes through formula realize, wherein: c ilDP code value within the scope of the LDP code value that represents to set.
10. the device as described in as arbitrary in claim 6 to 9, is characterized in that, described face recognition module, specifically for the feature based on extracting, adopts arest neighbors sorting algorithm KNN to carry out recognition of face to described image to be identified.
CN201410213460.2A 2014-05-20 2014-05-20 Face recognition method and device thereof Pending CN104021372A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410213460.2A CN104021372A (en) 2014-05-20 2014-05-20 Face recognition method and device thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410213460.2A CN104021372A (en) 2014-05-20 2014-05-20 Face recognition method and device thereof

Publications (1)

Publication Number Publication Date
CN104021372A true CN104021372A (en) 2014-09-03

Family

ID=51438117

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410213460.2A Pending CN104021372A (en) 2014-05-20 2014-05-20 Face recognition method and device thereof

Country Status (1)

Country Link
CN (1) CN104021372A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105117707A (en) * 2015-08-29 2015-12-02 电子科技大学 Regional image-based facial expression recognition method
CN105825192A (en) * 2016-03-24 2016-08-03 深圳大学 Facial expression identification method and system
CN106127251A (en) * 2016-06-23 2016-11-16 合肥工业大学 A kind of computer vision methods for describing face characteristic change
CN106503718A (en) * 2016-09-20 2017-03-15 南京邮电大学 A kind of local binary patterns Image Description Methods based on wave filter group
CN107194351A (en) * 2017-05-22 2017-09-22 天津科技大学 Face recognition features' extraction algorithm based on weber Local Symmetric graph structure
CN112491844A (en) * 2020-11-18 2021-03-12 西北大学 Voiceprint and face recognition verification system and method based on trusted execution environment

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120070041A1 (en) * 2010-09-16 2012-03-22 Jie Wang System And Method For Face Verification Using Video Sequence
CN103761515A (en) * 2014-01-27 2014-04-30 中国科学院深圳先进技术研究院 Human face feature extracting method and device based on LBP
CN103778434A (en) * 2014-01-16 2014-05-07 重庆邮电大学 Face recognition method based on multi-resolution multi-threshold local binary pattern

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120070041A1 (en) * 2010-09-16 2012-03-22 Jie Wang System And Method For Face Verification Using Video Sequence
CN103778434A (en) * 2014-01-16 2014-05-07 重庆邮电大学 Face recognition method based on multi-resolution multi-threshold local binary pattern
CN103761515A (en) * 2014-01-27 2014-04-30 中国科学院深圳先进技术研究院 Human face feature extracting method and device based on LBP

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
龚劬等: "结合改进的LBP和LDP的人脸表情识别", 《计算机工程与应用》 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105117707A (en) * 2015-08-29 2015-12-02 电子科技大学 Regional image-based facial expression recognition method
CN105825192A (en) * 2016-03-24 2016-08-03 深圳大学 Facial expression identification method and system
CN105825192B (en) * 2016-03-24 2019-06-25 深圳大学 A kind of facial expression recognizing method and system
CN106127251A (en) * 2016-06-23 2016-11-16 合肥工业大学 A kind of computer vision methods for describing face characteristic change
CN106503718A (en) * 2016-09-20 2017-03-15 南京邮电大学 A kind of local binary patterns Image Description Methods based on wave filter group
CN106503718B (en) * 2016-09-20 2019-11-22 南京邮电大学 A kind of local binary patterns Image Description Methods based on wave filter group
CN107194351A (en) * 2017-05-22 2017-09-22 天津科技大学 Face recognition features' extraction algorithm based on weber Local Symmetric graph structure
CN107194351B (en) * 2017-05-22 2020-06-23 天津科技大学 Face recognition feature extraction method based on Weber local symmetric graph structure
CN112491844A (en) * 2020-11-18 2021-03-12 西北大学 Voiceprint and face recognition verification system and method based on trusted execution environment

Similar Documents

Publication Publication Date Title
Zhang et al. Joint face detection and alignment using multitask cascaded convolutional networks
EP3096263B1 (en) Human body orientation recognition method and system based on two-lens camera
CN104021372A (en) Face recognition method and device thereof
US9020186B2 (en) Method and apparatus for detecting object using volumetric feature vector and 3D haar-like filters
CN105956560A (en) Vehicle model identification method based on pooling multi-scale depth convolution characteristics
CN111626371B (en) Image classification method, device, equipment and readable storage medium
CN102855461B (en) In image, detect the method and apparatus of finger
CN103971102A (en) Static gesture recognition method based on finger contour and decision-making trees
Kadam et al. Detection and localization of multiple image splicing using MobileNet V1
Arif et al. Automated body parts estimation and detection using salient maps and Gaussian matrix model
Li et al. Pushing the “Speed Limit”: high-accuracy US traffic sign recognition with convolutional neural networks
CN109145742A (en) A kind of pedestrian recognition method and system
JP2016062610A (en) Feature model creation method and feature model creation device
CN105718889A (en) Human face identity recognition method based on GB(2D)2PCANet depth convolution model
CN103413119A (en) Single sample face recognition method based on face sparse descriptors
Hasan et al. Features fitting using multivariate gaussian distribution for hand gesture recognition
JP5906071B2 (en) Information processing method, information processing apparatus, and storage medium
CN105893941B (en) A kind of facial expression recognizing method based on area image
CN105160305A (en) Finger multi-mode characteristic fusion method
CN110263726B (en) Finger vein identification method and device based on deep correlation feature learning
WO2022247403A1 (en) Keypoint detection method, electronic device, program, and storage medium
Li et al. Cascade Classifier Using Combination of Histograms of Oriented Gradients for Rapid Pedestrian Detection.
Mehryar et al. Automatic landmark detection for 3d face image processing
Chen et al. Fresh tea sprouts detection via image enhancement and fusion SSD
CN102819754A (en) Fingerprint score fusion system and method based on Sigmoid expansion

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20140903

WD01 Invention patent application deemed withdrawn after publication