CN101739546A - Image cross reconstruction-based single-sample registered image face recognition method - Google Patents
Image cross reconstruction-based single-sample registered image face recognition method Download PDFInfo
- Publication number
- CN101739546A CN101739546A CN200810228567A CN200810228567A CN101739546A CN 101739546 A CN101739546 A CN 101739546A CN 200810228567 A CN200810228567 A CN 200810228567A CN 200810228567 A CN200810228567 A CN 200810228567A CN 101739546 A CN101739546 A CN 101739546A
- Authority
- CN
- China
- Prior art keywords
- image
- face
- facial image
- identified
- people
- 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 image cross reconstruction-based single-sample registered image face recognition method. The method adopting the image cross reconstruction realizes the personal identity authentication by computer analysis and processing, and particularly comprises the following steps: registering low-frequency information of standard face images in a face database; acquiring a front image of a face to be recognized; extracting high-frequency information of the image of the face to be recognized; performing the image cross reconstruction on the high-frequency information of the image of the face to be recognized and the low-frequency information of the registered face images; and finally, performing matching judgment on the fused images and the image of the face to be recognized to realize the identity recognition. The image cross reconstruction-based single-sample registered image face recognition method has the advantages of small space of the database registering the face information, simple algorithm, short recognition time, high recognition rate and the like.
Description
Technical field
The invention belongs to the identification field, relate in particular to a kind of image cross reconstruction, single-sample registered image face identification method that carries out identification of passing through based on the image cross reconstruction.
Background technology
Recognition of face utilizes the Computer Analysis facial image exactly, therefrom extracts the information of facial image, carries out a kind of technology of identification.With respect to the other biological feature identification, recognition of face has the unique technique advantage aspect availability, and this is mainly reflected in:
1, can hidden operation, be particularly useful for security monitoring;
2, contactless collection, the property invaded is not accepted easily;
3, has convenient, fast, powerful trace ability afterwards;
4, the image capture device cost is low;
5, more meet human identification custom, interaction is strong.
For these reasons, recognition of face has become an important component part of biometrics identification technology, and is subjected to increasing attention.
Single sample face recognition technology just utilizes everyone single width facial image of Computer Analysis, therefrom extracts effective recognition information, is used for a special kind of skill of " identification " identity.
Single sample recognition of face is compared with traditional recognition of face based on multiple image, not only has above-mentioned advantage, very important advantage of tool also simultaneously, and face database exactly is easy to get.At be easy to get usually individual's single width photo of national government security department or some small-sized departments, as I.D., employee's card, student's identity card, passport, diploma and admission card for entrance examination etc.Train and traditional recognition of face based on multiple image is everyone multiple image of requirement, discern on this basis, this just requires the user to cooperate, and to gather multiple different photo, this is very difficult in practical operation.Because it is single sample recognition of face has these advantages, therefore very practical in some large-scale video monitorings or criminal's tracking.
Present face recognition technology mainly is based on several training images, need to gather many middle change situation facial images (as angle, illumination, expression etc.), comprehensively extract feature as training image, but in practice, the face database that obtains everyone various variations is very difficult, breadboard personnel can cooperate energetically and finish the various requirement of setting up specific face database, do not gear to actual circumstances and require the ordinary person also to cooperate so in actual applications.In many practical matter, we have only everyone photo, for example may be photos above I.D., employee's card, student's identity card, passport, diploma and the admission card for entrance examination etc.Therefore it is very significant carrying out recognition of face with single sample.
From paper and the statistical data of delivering at present, recognition of face problem for individual registered images, the most frequently used method is according to the faceform, generate many training images by single sample image, and identifying schemes mainly still adopts existing face recognition technology, improve discrimination by optimizing feature extraction and sorting technique, make it be suitable for single sample identification problem as far as possible.Although above-mentioned face identification method sets out from different perspectives, the recognition of face performance is increased, for the identification problem of individual registered images, still there is not special effective solution.
Existing face identification method is of all kinds.Generally speaking be divided into for three steps:
1, data input
Data input aspect is exactly to gather different people's face pictures, carries out data fusion.For example: the three-dimensional face identification of various visual angles, the recognition of face of colourful attitude, the recognition of face of coloured image, dynamic human face image recognition etc.
2, feature extraction
The method of recognition of face recent years mainly is to concentrate on the feature extraction aspect.Feature extracting methods is a lot, as the face identification method based on the Gabor small echo, based on the face identification method of PCA, based on the face identification method of svd, based on face identification method of small echo etc.
3, decision-making classification
Decision-making classification go up be exactly seek a good sorter differentiate between class and class in.Existing recognition methods mainly contains neural network method, the method for support vector machine, the method for Hidden Markov Model (HMM) etc.
But be the independent a certain class methods of use, the effect of identification is not fine, so most of researchist just uses the method for above two classes or three classes comprehensively to discern people's face, to improve discrimination.This just exists the complexity of algorithm to improve, and the long drawback of recognition time.
Summary of the invention
The present invention is intended to overcome the deficiencies in the prior art part and provides a kind of algorithm simple, and recognition time is short, the single-sample registered image face identification method based on the image cross reconstruction of strong robustness.
For achieving the above object, the present invention proposes to utilize camera system to obtain and comprises people's face direct picture of eyes, and a kind of single-sample registered image face identification method based on the image cross reconstruction proposed, adopt the method that merges two facial image each several part information, by computing machine facial image is analyzed, it mainly may further comprise the steps:
(1) extracts the low-frequency information of standard faces image as log-on data;
(2) obtain the people's face direct picture that comprises eyes of people's face to be identified;
(3) high-frequency information of extraction facial image to be identified;
(4) high-frequency information of facial image to be identified and the log-on data of standard faces image are carried out image reconstruction;
(5) image and facial image to be identified after rebuilding are mated differentiation.
At first, as a kind of preferred version, step of the present invention is carried out wavelet transformation with the standard faces image in (1), and low frequency coefficient is as log-on message after the extraction conversion.
As a kind of preferred version, obtain the direct picture of people's face to be identified in the step of the present invention (2), this direct picture is meant and comprises the eyes facial image.
As another kind of preferred version, extract the high-frequency information of facial image to be identified in the step of the present invention (3) and operate as follows:
(A) image to be identified is positioned, obtain the people face part;
(B) facial image to be identified is carried out photo-irradiation treatment;
(C) facial image to be identified is carried out angularity correction;
(D) facial image to be identified is carried out scale;
(E) facial image to be identified is carried out wavelet transformation, extract the high-frequency information of image.
In addition, in the step of the present invention (4) log-on data of the high-frequency information of facial image to be identified and the width of cloth standard faces in the database is constituted the wavelet conversion coefficient that will newly form again behind one group of new wavelet conversion coefficient and carry out wavelet inverse transformation, obtain reconstructed image.
Secondly, the present invention carries out distance calculation to image and the facial image to be identified after rebuilding, and judges whether image to be identified and standard picture belong to a people; If do not belong to same people, then repeating step (4) is rebuild and computed range with next width of cloth log-on data in the database, up to the match is successful.
Once more, the registration facial image described in step of the present invention (1), step (3), step (4) and the step (5) only need be registered the partial information of facial image.
The present invention compares with other people face recognition method, has following several characteristics:
(a) Image Acquisition is simple.Registered images of the present invention only needs a standard faces image, can be I.D., employee's card, student's identity card, passport, diploma and admission card for entrance examination etc., can directly obtain in the process of taking pictures, and sets up database.
(b) algorithm is simple.The present invention only need extract the partial information of people's face, and does not need whole people's face is carried out feature extraction.
(c) required face database storage space is little.Face database of the present invention only need be stored the partial information of people's face, and does not need to store whole facial image.
Description of drawings
The invention will be further described below in conjunction with the drawings and specific embodiments.Protection scope of the present invention will not only be confined to the statement of following content.
Fig. 1 is a face identification system block diagram of the present invention;
Fig. 2-1 is a facial image of the present invention;
Fig. 2-2 decomposes facial image for small echo one-level of the present invention;
Fig. 2-3 decomposes the facial image synoptic diagram for small echo one-level of the present invention;
Fig. 2-4 decomposes facial image for small echo secondary of the present invention;
Fig. 2-5 decomposes the facial image synoptic diagram for small echo secondary of the present invention;
Fig. 3-1 is a synoptic diagram before inventor's face image irradiation is handled;
Fig. 3-2 is that inventor's face image irradiation is handled the back synoptic diagram;
Fig. 4-1 is inventor's face eye location synoptic diagram;
Fig. 4-2 is a synoptic diagram after inventor's face eye location;
Fig. 5-1 is the present invention's facial image to be identified;
Fig. 5-2 is inventor's face angularity correction image;
Fig. 5-3 is inventor's face scale figure;
Fig. 6-1 registers facial image for the present invention;
Fig. 6-2 registers the low frequency restructuring graph of facial image for the present invention;
Fig. 7-1 is the present invention's facial image to be identified;
Fig. 7-2 is the high-frequency information reconstruct synoptic diagram of the present invention's people's face to be identified;
Image after the high-frequency information that Fig. 8 registers the low-frequency information of facial image and facial image to be identified for the present invention is rebuild.
Embodiment
As shown in Figure 1, concrete implementation step of the present invention is as follows:
Step 1: the low-frequency information of extracting the standard faces image
The standard faces image is carried out wavelet transformation, and low frequency coefficient is as log-on message after the extraction conversion;
Step 2: the people's face direct picture that comprises eyes that obtains people's face to be identified
Adopt a stationkeeping, but the camera that self can rotate up and down forms certain contour relation with the people, collects the facial image that comprises eyes; Perhaps adopt a camera that can all move up and down, collect the facial image that comprises eyes.
Step 3: the high-frequency information that extracts facial image to be identified
(A) image to be identified is positioned, obtain the people face part;
(B) facial image to be identified is carried out photo-irradiation treatment;
(C) facial image to be identified is carried out angularity correction;
(D) facial image to be identified is carried out scale;
(E) facial image to be identified is carried out wavelet transformation, extract the high-frequency information of image.
Step 4: the low-frequency information of registration facial image and the high-frequency information of facial image to be identified are carried out image reconstruction
(1) high-frequency information of facial image to be identified and the log-on data of the width of cloth standard faces in the database are constituted one group of new wavelet conversion coefficient;
(2) wavelet conversion coefficient that will newly form carries out wavelet inverse transformation, obtains reconstructed image.
Step 5: image and facial image to be identified after rebuilding are carried out distance calculation, judge whether image to be identified and standard picture belong to a people; If do not belong to same people, then repeating step three, rebuild and computed range with next width of cloth log-on data in the database, up to the match is successful.
Wherein the embodiment of step 1 is:
(1) sets up the facial image database with certificate photographs such as I.D., employee's card, student's identity card, passport, diploma and admission cards for entrance examination;
(2) select wavelet basis according to the face database characteristics of image, the size of the registration facial image that the present invention uses is 92*112, selects the db3 small echo for use;
(3) determine to decompose the number of plies according to the face database image, the present invention is defined as three grades of decomposition;
(4) face images in the database is carried out three grades of decomposition of small echo, decomposition can obtain 10 subgraphs, extracts low frequency subgraph as log-on message.
Wherein the embodiment of step 2 is:
The camera of camera system at first is set, and camera and people's spatial relation can be realized according to fixed form and removable mode.No matter be fixed form or removable mode, all require camera can photograph the facial image that contains eyes.For convenience of treatment of picture, it is constant that camera and people's shooting distance should keep, and makes in the image of each shooting, and the size of people's face is constant relatively.
For fixed form, but the mounting points height reference man's of camera average height.If people's average height is a m rice, then the liftoff setting height(from bottom) of camera is slightly less than m rice (considering a bit of distance on the eyes and the crown), makes that camera can be over against people's eyes.
For removable mode, the initial mounting points of camera can be finished according to the mounting points of fixed form, this mounting points locate up and down slidably track of camera is installed again, make camera to move up and down along track.
Camera requires in the captured scope of camera a people to be arranged in the process of taking facial image, can occur following three kinds of situations like this:
(1) can not find people's eyes.Because people's height differences, somebody's height can depart from people's average height, makes camera seek the eyes less than the people;
(2) can only find people's eyes.People's face still because the people is not over against camera, the deflection phenomenon of people's face then can occur in the coverage of camera, make camera can only photograph half of face;
(3) can find people's eyes.
For first kind of situation, the measure that fixed form is taked is: control the rotation up and down of camera self by seeking the human eye algorithm, make camera can find people's eyes.For example, when someone's height greater than (less than) during people's average height, camera is sought the eyes less than this people, then camera self can rotate the eyes that seek the people as last (descend).The measure that removable mode is taked is: move up and down camera automatically by seeking the human eye algorithm.For example, when someone's height greater than (less than) during people's average height, then camera self can be mobile as last (descend), till the eyes that find the people.Because camera and people have certain shooting distance, so camera only needs to rotate up and down (moving) sub-fraction distance, just can change a big angular field of view, and not need the rotation up and down (moving) of wide-angle.
For second kind of situation, the measure that fixed form is taked is: control the left rotation and right rotation of camera self by seeking the human eye algorithm, make camera can find people's eyes.The measure that removable mode is taked is: control the move left and right of camera self by seeking the human eye algorithm, find up to camera till people's the eyes.
For the third situation, fixed form and removable mode can be taken people's face normally.
Camera be fixed form or removable mode to obtaining the influence of facial image:
For fixed form, after camera is through rotation up and down, the visual angle of camera and people's face not over against, there is the deflection on the locus at this moment captured facial image, for example pitching rotation etc.
For removable mode, when camera through after moving up and down, the visual angle of camera and people's face still over against, therefore there is not the deflection on the locus in captured facial image, still this mode in actual mechanical process than fixed form complexity.
No matter camera is fixed form or removable mode is taken facial image, before taking, all requires at camera under all angles, take a background image earlier, just object to be identified not situation under, take earlier image, an image as a setting.
Wherein the embodiment of step 3 is:
(1) image to be identified is positioned, obtain the people face part
Obtain facial image with image pick-up card, wherein obtaining of frame of video adopted direct show technology, follows every two field picture is carried out people's face location.
Native system adopts the method for detecting human face based on skin color segmentation.The colour of skin is the important information of people's face, has relative stability, and distinguishes mutually with the color of most of background objects.Generally the colorized face images that we saw all is based on the RGB color space, but in the RGB color space, chrominance information and monochrome information mix, because the change of surrounding environment illumination, brightness may make the detection of people's face become more complicated, make that the skin color segmentation result is unreliable, so the present invention uses the color space based on YCrCb.Y indicates diopter, i.e. brightness (in fact representing gray-scale value), and Cr and Cb then are meant colourity, promptly describe the saturation degree of color.Experimental study shows: although the face complexion of different nationalities, all ages and classes, different sexes looks different, but this difference mainly concentrates in the brightness, in some color space of removing brightness, it is consistent that the colour of skin of different people face distributes, and concentrate in the less zone, promptly have the cluster characteristic.
Concrete implementation step is:
1. obtain the colorized face images of each frame;
2. the rgb value of colorized face images is converted into the YCrCb value;
3. by being experimentized, a large amount of colorized face images find the interval of colour of skin Cr value;
4. the area of skin color of people's face is judged in the interval of rule of thumb resulting colour of skin Cr value; The facial image that collects is carried out the pointwise picture element scan,, then think the class skin pixel, and this pixel is made as white, otherwise think it is the non-colour of skin, be made as black, generate binary image thus if certain puts the Cr value of pixel in this interval;
5. the image after the binaryzation is carried out opening and closing operation, the non-face zone of fraction of class colour of skin characteristic is arranged with elimination.Orient people's face thus;
6. the colorized face images of orienting is converted into the gray scale facial image.
(2) facial image to be identified is carried out photo-irradiation treatment
The illumination problem generally shows as intensity of illumination and differs (strong or weak), and uneven illumination is even etc.The present invention here uses a kind of preconditioning technique based on regional regularization.Suppose to have virtual " standard " illumination condition, under these standard conditions, the pixel value distribution range of facial image is 0～1.To obtain the target that the facial image under this " standard " illumination condition is handled as the image regularization.Correspondingly, can think that existing intensity profile is that facial image between 0～255 all is to obtain under the illumination condition that is better than " standard " illumination.If the facial image of thinking to obtain under the different illumination conditions " normalizing ", then should be considered the power influence of illumination to identical fixed-illumination condition.
Therefore, in order to draw the image that distributes between 0～1 under " standard " illumination, should be with original pixel value divided by a numerical value greater than itself.Can suppose that the image under " standard " illumination is
Wherein: I
XyFor image at (x, gray-scale value y);
Pixel average for entire image; R is a positive coefficient.Here pixel average and the illumination power of supposing facial image are linear.In addition, the I in (1) formula denominator
XyThe influence that suppresses light source direction there is certain effect.Under the bigger situation of the deflection of pointolite direction, to compare with the imaging under the positive dirction light source condition, a facial image possibility part is bright partially, and another part is dark partially.Handle if with (1) formula image is carried out canonical, a kind of so bigger light and shade difference can obtain weakening.Concrete implementation step is:
1. about facial image being divided into, about identical four I of size
1, I
2, I
3And I
4
2. each piece image of facial image is asked the pixel average respectively
With
3. the image pixel value of each piece and average thereof are brought into formula (2), obtain the pixel value I ' of the corresponding point of each piece
i(i=1,2,3,4);
4. four little image I after the linear transformation '
i(i=1,2,3,4) are reassembled into a new facial image, are the facial image after the photo-irradiation treatment.
(3) facial image to be identified is carried out angularity correction
People's face of being oriented by said method has just been oriented the approximate region of people's face, near the background area that facial contour, can have fraction, this will impact the location of human eye, therefore go out human eye for accurate in locating, just need to reduce unnecessary interference region, need further find the approximate region of human eye.
Experiment shows with the background area to be compared, and human face region often has higher brightness.At people's face left and right sides boundary, the summation of brightness value reduces rapidly on the vertical direction, thereby forms a tangible protruding peak.Therefore, only need carry out vertical integral projection, determine the border, the left and right sides at main protruding peak in the vertical gray integration drop shadow curve, can obtain the border, the left and right sides of people's face facial image.In like manner facial image is carried out horizontal integral projection.Experiment is found, the corresponding people's of first minimum point of horizontal integral projection curve the crown, because the low gray scale of hair has produced the low ebb of horizontal gray integration drop shadow curve, the inferior maximum of points of curve and maximum of points be corresponding people's forehead position and people's nose middle part then, therefore, thereby orient the up-and-down boundary of facial image as long as find the inferior maximum of points and the maximum of points of horizontal gray integration drop shadow curve.So just the approximate region of human eye delimited, shown in Fig. 4-1.
With edge detection operator the approximate region of the human eye of delimitation is carried out rim detection, can obtain human eye pupil boundary zone, because the human eye pupil is similar round, therefore can find the pupil center of circle in this zone with the Hough conversion, promptly orient the position at human eye place.Because the human face region in the video image is very little, even therefore have deviation with the detected next human eye of the method, deviation also can be very little, can ignore, shown in Fig. 4-2.
By to two location, find the straight line at eyes place and horizontal direction straight line between angle theta, according to this angle θ, adopt formula (2) that image is rotated.
Wherein a (x, y) and b (x, y) expression former coordinate x and the determined new coordinate of y.Rotation back facial image synoptic diagram is shown in Fig. 5-2.
(4) facial image to be identified is carried out scale
Postrotational facial image is carried out convergent-divergent:, therefore it is considered herein that and can the distance between everyone two be considered as equating because facial image is very little.Thus can by calculate distance between two and standard according in the ratio γ of distance between two of facial image, be the proportional zoom coefficient with γ, postrotational facial image is carried out convergent-divergent.
Facial image behind the convergent-divergent is carried out cutting: the people in the facial image behind the convergent-divergent is bold little consistent with the standard photograph, but expanding may appear in its background, cause the size of whole facial image and the size of standard photograph to differ, therefore need be cropped to facial image consistent according to size with standard, with convenient follow-up feature extraction, the synoptic diagram of facial image is shown in Fig. 5-3 behind the convergent-divergent.
(5) facial image to be identified is carried out wavelet transformation, extract the high-frequency information of image
The to be identified facial image intact to pre-service carries out 3 grades of decomposition of small echo.The input facial image can be from facial image in the video monitoring, or the fixing facial image of gathering in the work attendance, but the size of input facial image must and registration facial image big or small identical, make the registration facial image and import facial image and can under a yardstick, mate.Here, the present invention extracts the frequency domain information of input facial image, and the method for use is for extracting with small echo.The same db3 small echo that uses carries out three grades of decomposition of small echo to the input facial image, and decomposition can obtain 10 subgraphs, wherein has only one to be low frequency subgraph, and other is the high frequency subgraph.Remove its low frequency subgraph, and stay 9 high frequency subgraphs, as the high-frequency information of input facial image.9 high frequency subgraphs with the db3 wavelet reconstructions after shown in Fig. 7-2.
Wherein the embodiment of step 4 is:
1) obtained the low-frequency information of registration facial image in the extraction step one
2) obtained the high-frequency information of facial image to be identified in the extraction step two, three.
3) above low frequency and high-frequency information are formed new set of wavelet coefficients, carry out three grades of reconstruct of db3 small echo, the image after obtaining rebuilding, as shown in Figure 8, the image size after the reconstruction is big or small identical with the input facial image.
Wherein the embodiment of step 5 is:
If the arbitrfary point pixel value of input facial image is f
1(i, j), the arbitrfary point pixel value of the image after the reconstruction is f
2(i, j).The row-coordinate of i presentation video wherein, the ordinate of j presentation video.Image after rebuilding is carried out Euclidean distance with the input facial image subtract each other, image after obtaining rebuilding and the distance A of importing facial image, that is:
The low-frequency information of each standards registration facial image in the low-frequency information of input facial image and the database is rebuild, several images after rebuilding are all carried out Euclidean distance with the input facial image to be subtracted each other, find and import the image of facial image after apart from the reconstruction of minimum, can judge that this input facial image is same individual with registrant's face of the low-frequency information of the reconstructed image that provides.
Claims (7)
1. single-sample registered image face identification method based on the image cross reconstruction, adopt camera system to obtain the facial image that comprises eyes of people's face to be identified, with the method that merges two facial image each several part information, by computing machine facial image is analyzed, be it is characterized in that may further comprise the steps:
(1) extracts the low-frequency information of individual standard faces image as log-on data;
(2) obtain the facial image that comprises eyes of people's face to be identified;
(3) high-frequency information of extraction facial image to be identified;
(4) high-frequency information of facial image to be identified and the log-on data of standard faces image are carried out image reconstruction;
(5) image and facial image to be identified after rebuilding are mated differentiation.
2. the single-sample registered image face identification method based on the image cross reconstruction according to claim 1, it is characterized in that: described step is carried out wavelet transformation with the standard faces image in (1), and low frequency coefficient is as log-on message after the extraction conversion.
3. the single-sample registered image face identification method based on the image cross reconstruction according to claim 1, it is characterized in that: be to adopt a stationkeeping in the described step (2), but the camera that camera lens can rotate up and down, form certain contour relation with the people, collect the people's face direct picture that comprises eyes; Perhaps adopt a camera that can all move up and down, collect the people's face direct picture that comprises eyes.
4. the single-sample registered image face identification method based on the image cross reconstruction according to claim 1 and 2 is characterized in that: the high-frequency information that extracts facial image to be identified in the described step (3) is operated as follows:
(A) image to be identified is positioned, obtain the people face part;
(B) facial image to be identified is carried out photo-irradiation treatment;
(C) facial image to be identified is carried out angularity correction;
(D) facial image to be identified is carried out scale;
(E) facial image to be identified is carried out wavelet transformation, extract the high-frequency information of image.
5. the single-sample registered image face identification method based on the image cross reconstruction according to claim 4, it is characterized in that: in the described step (4) log-on data of the high-frequency information of facial image to be identified and the width of cloth standard faces in the database is constituted the wavelet conversion coefficient that will newly form again behind one group of new wavelet conversion coefficient and carry out wavelet inverse transformation, obtain reconstructed image.
6. the single-sample registered image face identification method based on the image cross reconstruction according to claim 5, it is characterized in that: image and facial image to be identified after rebuilding are carried out distance calculation, judge whether image to be identified and standard picture belong to a people; If do not belong to same people, then repeating step (4) is rebuild and computed range with next width of cloth log-on data in the database, up to the match is successful.
7. the single-sample registered image face identification method based on the image cross reconstruction according to claim 6 is characterized in that: the registration facial image described in described step (1), step (3), step (4) and the step (5) only need be registered the partial information of facial image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810228567A CN101739546A (en) | 2008-11-05 | 2008-11-05 | Image cross reconstruction-based single-sample registered image face recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810228567A CN101739546A (en) | 2008-11-05 | 2008-11-05 | Image cross reconstruction-based single-sample registered image face recognition method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101739546A true CN101739546A (en) | 2010-06-16 |
Family
ID=42463016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810228567A Pending CN101739546A (en) | 2008-11-05 | 2008-11-05 | Image cross reconstruction-based single-sample registered image face recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101739546A (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102142154A (en) * | 2011-05-10 | 2011-08-03 | 中国科学院半导体研究所 | Method and device for generating virtual face image |
| CN102456137A (en) * | 2010-10-20 | 2012-05-16 | 上海青研信息技术有限公司 | Sight line tracking preprocessing method based on near-infrared reflection point characteristic |
| CN103413119A (en) * | 2013-07-24 | 2013-11-27 | 中山大学 | Single sample face recognition method based on face sparse descriptors |
| CN103543339A (en) * | 2013-10-31 | 2014-01-29 | 国家电网公司 | Method and device for analyzing alternating current-and-direct current space charge test data of dielectric sample |
| CN104484854A (en) * | 2014-12-24 | 2015-04-01 | 北京奇虎科技有限公司 | Cropping method and device for people pictures |
| CN104680161A (en) * | 2015-01-09 | 2015-06-03 | 安徽清新互联信息科技有限公司 | Digit recognition method for identification cards |
| WO2015154205A1 (en) * | 2014-04-11 | 2015-10-15 | Xiaoou Tang | Methods and systems for verifying face images based on canonical images |
| CN105049732A (en) * | 2015-08-27 | 2015-11-11 | 广东欧珀移动通信有限公司 | Camera adjusting method and user terminal |
| CN105606864A (en) * | 2016-03-25 | 2016-05-25 | 宋健 | Using method of automatic electricity stealing preventing watt-hour meter |
| CN105628996A (en) * | 2016-03-25 | 2016-06-01 | 胡荣 | Image processing-based electric energy meter |
| CN105675935A (en) * | 2016-03-25 | 2016-06-15 | 李英 | Intelligent electric energy meter |
| CN105675945A (en) * | 2016-03-25 | 2016-06-15 | 李娜 | Image detection-based intelligent electric meter using method |
| CN105699933A (en) * | 2016-02-04 | 2016-06-22 | 任红霞 | Direct current electric energy meter detecting system for electric automobile battery charging pile |
| CN105842501A (en) * | 2016-03-25 | 2016-08-10 | 宋健 | Electric meter capable of automatically preventing electricity stealing |
| CN105842532A (en) * | 2016-03-25 | 2016-08-10 | 高秀丽 | Method for using image identification-based automatic electric meter |
| CN105866536A (en) * | 2016-03-25 | 2016-08-17 | 高秀丽 | Automatic electric meter based on image recognition |
| CN105866528A (en) * | 2016-03-25 | 2016-08-17 | 胡荣 | Electric energy meter application method based on image processing |
| CN103714326B (en) * | 2013-12-26 | 2017-04-12 | 江南大学 | One-sample face identification method |
| CN109284685A (en) * | 2018-08-22 | 2019-01-29 | 石数字技术成都有限公司 | A kind of face identification method and system based on integration method |
| CN110210414A (en) * | 2019-06-05 | 2019-09-06 | 北京京投信安科技发展有限公司 | The quick intersection identification technology of magnanimity face database |
| CN112884030A (en) * | 2021-02-04 | 2021-06-01 | 重庆邮电大学 | Cross reconstruction based multi-view classification system and method |
| CN113033514A (en) * | 2021-05-24 | 2021-06-25 | 南京伯索网络科技有限公司 | Classroom student aggressiveness evaluation method based on network |
| CN116052222A (en) * | 2023-03-06 | 2023-05-02 | 吉林大学 | Cattle face recognition method for naturally collecting cattle face image |
| CN111401456B (en) * | 2020-03-20 | 2023-08-22 | 杭州涂鸦信息技术有限公司 | Training method, system and device for face gesture recognition model |
-
2008
- 2008-11-05 CN CN200810228567A patent/CN101739546A/en active Pending
Non-Patent Citations (3)
| Title |
|---|
| 何家忠 等: "基于小波的训练样本增强的单样本人脸识别", 《计算机工程与应用》 * |
| 何家忠 等: "结合小波低频子带的主成分分析方法", 《华南理工大学学报（自然科学版）》 * |
| 杨绍华: "一种基于小波和DCT的人脸识别算法", 《宁夏师范学院学报（自然科学）》 * |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102456137A (en) * | 2010-10-20 | 2012-05-16 | 上海青研信息技术有限公司 | Sight line tracking preprocessing method based on near-infrared reflection point characteristic |
| CN102456137B (en) * | 2010-10-20 | 2013-11-13 | 上海青研信息技术有限公司 | Sight line tracking preprocessing method based on near-infrared reflection point characteristic |
| CN102142154A (en) * | 2011-05-10 | 2011-08-03 | 中国科学院半导体研究所 | Method and device for generating virtual face image |
| CN103413119A (en) * | 2013-07-24 | 2013-11-27 | 中山大学 | Single sample face recognition method based on face sparse descriptors |
| CN103543339A (en) * | 2013-10-31 | 2014-01-29 | 国家电网公司 | Method and device for analyzing alternating current-and-direct current space charge test data of dielectric sample |
| CN103714326B (en) * | 2013-12-26 | 2017-04-12 | 江南大学 | One-sample face identification method |
| WO2015154205A1 (en) * | 2014-04-11 | 2015-10-15 | Xiaoou Tang | Methods and systems for verifying face images based on canonical images |
| CN106462724B (en) * | 2014-04-11 | 2019-08-02 | 北京市商汤科技开发有限公司 | Method and system based on normalized images verification face-image |
| CN106462724A (en) * | 2014-04-11 | 2017-02-22 | 北京市商汤科技开发有限公司 | Methods and systems for verifying face images based on canonical images |
| CN104484854A (en) * | 2014-12-24 | 2015-04-01 | 北京奇虎科技有限公司 | Cropping method and device for people pictures |
| CN104680161A (en) * | 2015-01-09 | 2015-06-03 | 安徽清新互联信息科技有限公司 | Digit recognition method for identification cards |
| CN105049732A (en) * | 2015-08-27 | 2015-11-11 | 广东欧珀移动通信有限公司 | Camera adjusting method and user terminal |
| CN105049732B (en) * | 2015-08-27 | 2018-05-29 | 广东欧珀移动通信有限公司 | A kind of camera shooting head adjusting method and user terminal |
| CN105699933A (en) * | 2016-02-04 | 2016-06-22 | 任红霞 | Direct current electric energy meter detecting system for electric automobile battery charging pile |
| CN105866528A (en) * | 2016-03-25 | 2016-08-17 | 胡荣 | Electric energy meter application method based on image processing |
| CN105606864A (en) * | 2016-03-25 | 2016-05-25 | 宋健 | Using method of automatic electricity stealing preventing watt-hour meter |
| CN105842532A (en) * | 2016-03-25 | 2016-08-10 | 高秀丽 | Method for using image identification-based automatic electric meter |
| CN105866536A (en) * | 2016-03-25 | 2016-08-17 | 高秀丽 | Automatic electric meter based on image recognition |
| CN105866497A (en) * | 2016-03-25 | 2016-08-17 | 张超 | Automatic electricity-stealing-preventing electric meter |
| CN105842501A (en) * | 2016-03-25 | 2016-08-10 | 宋健 | Electric meter capable of automatically preventing electricity stealing |
| CN105675945A (en) * | 2016-03-25 | 2016-06-15 | 李娜 | Image detection-based intelligent electric meter using method |
| CN105675935A (en) * | 2016-03-25 | 2016-06-15 | 李英 | Intelligent electric energy meter |
| CN105628996A (en) * | 2016-03-25 | 2016-06-01 | 胡荣 | Image processing-based electric energy meter |
| CN105842495A (en) * | 2016-03-25 | 2016-08-10 | 李军 | Image processing-based electric energy meter |
| CN109284685A (en) * | 2018-08-22 | 2019-01-29 | 石数字技术成都有限公司 | A kind of face identification method and system based on integration method |
| CN109284685B (en) * | 2018-08-22 | 2021-08-31 | 一石数字技术成都有限公司 | Face recognition method and system based on integral method |
| CN110210414A (en) * | 2019-06-05 | 2019-09-06 | 北京京投信安科技发展有限公司 | The quick intersection identification technology of magnanimity face database |
| CN111401456B (en) * | 2020-03-20 | 2023-08-22 | 杭州涂鸦信息技术有限公司 | Training method, system and device for face gesture recognition model |
| CN112884030A (en) * | 2021-02-04 | 2021-06-01 | 重庆邮电大学 | Cross reconstruction based multi-view classification system and method |
| CN113033514A (en) * | 2021-05-24 | 2021-06-25 | 南京伯索网络科技有限公司 | Classroom student aggressiveness evaluation method based on network |
| CN116052222A (en) * | 2023-03-06 | 2023-05-02 | 吉林大学 | Cattle face recognition method for naturally collecting cattle face image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101739546A (en) | Image cross reconstruction-based single-sample registered image face recognition method | |
| Kuruvilla et al. | A review on image processing and image segmentation | |
| CN102902959B (en) | Face recognition method and system for storing identification photo based on second-generation identity card | |
| Li et al. | Illumination invariant face recognition using near-infrared images | |
| US8345936B2 (en) | Multispectral iris fusion for enhancement and interoperability | |
| Gu et al. | Feature points extraction from faces | |
| Gao et al. | Standardization of face image sample quality | |
| CN106446872A (en) | Detection and recognition method of human face in video under low-light conditions | |
| CN103632132A (en) | Face detection and recognition method based on skin color segmentation and template matching | |
| CN105469076B (en) | Face alignment verification method based on multi-instance learning | |
| CN103914676A (en) | Method and apparatus for use in face recognition | |
| CN102663413A (en) | Multi-gesture and cross-age oriented face image authentication method | |
| CN109858439A (en) | A kind of biopsy method and device based on face | |
| CN108898125A (en) | One kind being based on embedded human face identification and management system | |
| Gupta et al. | Face detection using modified Viola jones algorithm | |
| CN106650606A (en) | Matching and processing method for face image and face image model construction system | |
| CN109359577A (en) | A kind of Complex Background number detection system based on machine learning | |
| Shrivastava et al. | Conceptual model for proficient automated attendance system based on face recognition and gender classification using Haar-Cascade, LBPH algorithm along with LDA model | |
| Wati et al. | Security of facial biometric authentication for attendance system | |
| Storer et al. | Occlusion detection for ICAO compliant facial photographs | |
| Sharma et al. | 3D face recognition techniques-a review | |
| Monwar et al. | Eigenimage based pain expression recognition | |
| Osadchy et al. | Surface dependent representations for illumination insensitive image comparison | |
| Bissoon et al. | Gender classification using face recognition | |
| Mekami et al. | Towards a new approach for real time face detection and normalization |
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: 20100616 |