CN101533468A - Method for estimating human age automatically based on digital face images - Google Patents
Method for estimating human age automatically based on digital face images Download PDFInfo
- Publication number
- CN101533468A CN101533468A CN200910031218A CN200910031218A CN101533468A CN 101533468 A CN101533468 A CN 101533468A CN 200910031218 A CN200910031218 A CN 200910031218A CN 200910031218 A CN200910031218 A CN 200910031218A CN 101533468 A CN101533468 A CN 101533468A
- Authority
- CN
- China
- Prior art keywords
- age
- vector
- pattern
- image
- subspace
- 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.)
- Granted
Links
Images
Abstract
The invention relates to a method for estimating a human age automatically based on digital face images, which comprises the following steps: step one, acquiring face images to be estimated; step two, extracting characteristic vectors of the face images to be estimated as growth mode vectors of candidate ages from the images to be estimated; step three, determining whether a subspace of a growth mode is trained well, if the answer is no, training the subspace of the growth mode, and if the answer is yes, going to the next step; step four, according to the well trained subspace of the growth mode, projecting all the growth mode vectors of the candidate ages of the face images to be estimated to the subspace of the growth mode, and then reconstructing integrated growth mode vectors of the candidate ages from the projected vectors; step five, finding out a growth mode with smallest reconstruction errors, namely a growth mode with the best ages by comparing the reconstruction errors between reconstructed images and the original images; and step six, determining that the position r of the face images to be estimated in the growth mode with the best candidate ages is the human age. The method has high precision and can be finished automatically.
Description
Technical field
The present invention relates to utilize computing machine that the human age is carried out automatic estimation approach, particularly a kind of facial image that utilizes carries out estimation approach to the human age.
Background technology
Find at present the technology of utilizing computing machine the human age to be estimated automatically as yet by facial image.But the technology that exists some to utilize digital facial image that people's identity is discerned, another patent of invention " based on the integrated digital facial image recognition method of the many eigen spaces of selectivity " (patent No. ZL 2,004 1 0041173.4) that has as the present patent application people.The identification of numeral facial image has certain getting in touch with the present invention, promptly all is to utilize digital facial image to seek the useful information relevant with personage in the image.But, the two solution be diverse problem, digital facial image recognition objective is identification people's a identity, and the present invention is the age of estimating the people.
Age is people's a important attribute, has determined people's behavior, idea and the rule that should observe.Human automatic estimation technique of age makes intelligence system provide respective service to become possibility according to user's age.In the Chinese society aging population with more and more pay attention under the pupillary overall background of protection, it is particularly important and urgent that this The Application of Technology seems, containing economic and social benefit greatly.Current estimation of Age of in most cases taking or verification mode or depend on people's subjective estimation such as in places refusal minors such as bar, Internet bar go into, or depend on relevant certificate such as passport is open to the custom.Shortcomings such as the intrinsic speed of these modes is slow, cost is high, unfriendly, unreliable, easy forgery can thoroughly be taken on a new look by the application that the automatic age is estimated system.And, the place that originally a lot of estimation of Age and checking are difficult for implementing also can be used this technology and be realized covering, Internet-browser such as selecting to shield some harmful information according to age of user automatically provides the automation services terminal of special service etc. for all ages and classes user.Therefore the Related product based on this technological development will have very vast market prospect.In addition, at aspects such as protection minor and love the elderlys, this The Application of Technology also has good social benefit.It should be noted that human age estimation method involved in the present invention only depends on facial image, the human age estimation method of studying in this and archaeology or the medical jurisprudence has fundamental difference.The latter estimates based on the relevant evidence of bone and tooth after people's death that mainly these evidences can't not invaded the collection of human body, so be difficult to be applied in the daily life.Different therewith, the present invention estimates people's age by people's face digital picture.This mode is as people estimate the mode at other people age in daily life, and is convenient and swift and not rude.Therefore, this automatic estimation of Age technology can directly apply in the intelligence system of friendly interface, and it is had and estimation of Age ability like the mankind.
Summary of the invention
The purpose of this invention is to provide a kind of computing machine that allows to be similar to people's mode, i.e. observer's face, to the automatic mode that the human age is made accurate estimation, the estimated accuracy of this method can reach and level like the mankind.
Technical scheme of the present invention is a kind of automatic method of estimation of human age based on digital facial image, and step is:
The first step is obtained the gray level image facial image promptly to be estimated of user's face by Digital Image Input Device;
Second step, treat that estimated image is placed on the age respectively and becomes in the long pattern on all possible age, extraction is waited to estimate the proper vector of facial image and is generated one group of candidate's age growth pattern vector, in each candidate's age growth pattern vector, only on an age characteristics of image is arranged, remainder all is a missing values;
The 3rd step, judge into the long pattern subspace and whether train, if not, then enter and be trained to the long pattern subspace, if then enter next step;
In the 4th step,, will wait that all candidate's age growth pattern vectors of estimating facial image project in the long pattern subspace, reconstruct complete candidate's age growth pattern vector from projection vector again according to the one-tenth long pattern subspace that trains;
The 5th step, if treating estimated image was placed on the correct age, then reconstruct age of coming out becomes in the long pattern facial image and the original image on the corresponding age very alike, otherwise twisted phenomena will appear in the facial image that reconstruct is come out, by comparing the reconstructed error between reconstructed image and the original image, reconstructed error between age growth pattern vector of reconstruct just and the original age growth pattern vector, find candidate's age of reconstructed error minimum to become long pattern, promptly treat the best age of estimating facial image and become long pattern;
In the 6th step, wait to estimate that the position r of facial image in this optimal candidate age one-tenth long pattern is the estimation to its age, and it is exported as final result.
Related definition: 1. the age becomes long pattern: the image sequence that a kind of facial image of all ages and classes with the someone is arranged from small to large.2. age growth pattern vector: become each width of cloth facial image the long pattern to extract feature from the age, then the proper vector that is spliced from small to large according to the age.3. facial image feature: utilize the principal component analysis (PCA) technology in the higher algebra textbook, the proper vector that from facial image, extracts.4. become the long pattern subspace: becoming long pattern with one group of age is the most representative proper subspace that training sample calculates.5. reconstruct: utilize the one-tenth long pattern subspace train to reduce the process of facial image again.6. candidate's age becomes long pattern: image that will be to be estimated is placed on the one group of age one-tenth long pattern that is generated on the diverse location in the age one-tenth long pattern, and it is only wherein having only an image for the treatment of estimation.7. target age: system the range of age that can estimate, as 0 to 70 years old.
Beneficial effect: 1. this method full automation, without human intervention, and speed is fast, the accuracy height need can be applied to most occasions of estimation of Age.
2. this method face-image of only relying on the people just can make an estimate to its age, people estimate that others' mode at age is similar in this and the daily life, therefore can in daily use, implement very easily and can not allow the people feel trouble or dislike, even can implemented under the ignorant situation, thereby realize the application of a lot of hommizations by the estimator.
3. the age growth pattern vector among the present invention is a kind of special data structure of the people's of being suitable for face estimation of Age problem characteristics, allow to occur missing values in this data structure, the phenomenon of almost unavoidable appearance disappearance image in the age growth pattern in the more realistic application; In this data structure, the face images feature is arranged from small to large according to time sequencing, and this has reflected that the age becomes the sequential of long pattern; The all images feature that comprises in each age growth pattern vector all belongs to same individual, can fully reflect the personalization features that the age grows up like this, and promptly the mode of everyone ageing is different.
Description of drawings
Fig. 1 is based on the automatic estimating system workflow diagram of human age of digital facial image.
Fig. 2 is the generative process example of age growth pattern vector.
Fig. 3 is the process example that becomes disappearance image in the long pattern initialization age.
Fig. 4 is into the training process process flow diagram of long pattern subspace.
Fig. 5 is the process example of individual facial image to be estimated being carried out estimation of Age.
Embodiment
The present invention is described in detail below in conjunction with accompanying drawing and most preferred embodiment.
Based on automatic estimating system workflow diagram of the human age of digital facial image as shown in Figure 1.System at first obtains the gray level image facial image just to be estimated of user's face by Digital Image Input Device, enter computer processing procedure subsequently.This process comprises the proper vector of extracting the input facial image and generates age growth pattern vector as shown in Figures 2 and 3, whether judge into the long pattern subspace then trains, if not, then enter the long pattern subspace process that is trained to, as shown in Figure 4, if, then according to the one-tenth long pattern subspace that trains with wait that the proper vector of estimating facial image calculates the estimation of Age result, as shown in Figure 5, at last estimated result is exported.The visual different practical application of estimation age of system output and trigger corresponding operation.For example, when system applies in based on the man-machine interactive system at age the time, the bigger estimation age may make system with bigger character operation display interface, eyesight with the eldercare, on the contrary, the less estimation age, the system that may make adopted comparatively active and bright-coloured interface, to adapt to youthful hobby.For another example, when system applies during bad network information, belongs to teenage scope if estimate the age in shielding, system will stop the user to browse to be not suitable for pupillary information.
The present invention is based upon on a kind of special data structure basis, i.e. age growth pattern vector, as shown in Figure 2.The so-called age becomes long pattern, the image sequence that promptly a kind of facial image of all ages and classes with the someone is arranged from small to large.The age that Figure 2 shows that someone 0-8 year becomes long pattern.Wherein frame of broken lines is represented last unmanned face image of corresponding age, and this disappearance image becomes almost inevitable in the long pattern at the age, collects its facial image on institute's has age because be difficult to there from the someone.Next, age becomes the every width of cloth image in the long pattern to extract proper vector by the principal component analysis (PCA) technology in the higher algebra textbook, be about to image projection in the subspace that n orthogonal vector base arranged that obtains by principal component analysis (PCA), the n dimensional feature vector that obtains, here n is a predefined integer, for example 100, perhaps select to explain the n value of former data 90% variance, this value generally all is far smaller than number of pixels, but does not have direct relation with number of pixels.The proper vector of Jiang all images is spliced into a big vector according to the age order at last, just is referred to as age growth pattern vector.The corresponding age of missing values becomes the facial image that lacks in the long pattern in this vector, is exactly 0,1,3,4,6,7 in Fig. 2, is changed to m.Just the image projection with 2,5,8 years old the time obtains proper vector b in Fig. 2
2, b
5, b
8Be spliced into a big vector according to the age order, just age growth pattern vector.The description of back of the present invention is all based on age growth pattern vector.Attention is not manual the participation in the generative process of age growth pattern vector, and the facial image of input system can be processed automatically, extracts individual features and generates normalized proper vector, and this system that makes can operation automatically under unmanned situation.Age growth pattern vector is a kind of special data structure of the people's of being suitable for face estimation of Age problem characteristics, this shows: 1. allow to occur missing values in this data structure, this is to consider that age in actual applications becomes the almost inevitable and design carrying out of disappearance image in the long pattern; 2. in this data structure, the face images feature is arranged from small to large according to time sequencing, and this has reflected that the age becomes the sequential of long pattern; 3. all images feature that comprises in each age growth pattern vector all belongs to same individual, can fully reflect the personalization features that the age grows up like this, and promptly the mode of everyone ageing is different.
Often have a large amount of missing values in the age growth pattern vector in order to last method generation, these missing values must just can enter subsequent processes through initialization, and initialization procedure is most important to the validity and the iteration convergence efficient of system's subsequent algorithm.Employed missing values initial method as shown in Figure 3 among the present invention.Each row has shown a people's part age one-tenth long pattern among Fig. 3, and each lists and is identical facial image of age, and frame of broken lines represents to lack image.With the disappearance of the width of cloth wherein image is example, what just solid line was irised out among the figure, direction along this disappearance same age of image, be same row, generally can both become in the long pattern at other people age to find some facial images, two facial images that just dotted line is irised out among the figure, then with these same ages but the proper vector of the facial image of different people be averaged, with average be filled into the disappearance image the position as its initial value.All disappearance images in each age growth pattern vector in the training set are all carried out above processing, finish initialization.Through initialization, institute's has age growth pattern vector all no longer comprises missing values, be used to fill other people characteristic mean of being characterized as of missing values on the same age, can guarantee that so initial filling value and desired value are comparatively approaching, for follow-up one-tenth long pattern subspace training process is had laid a good foundation.
The process flow diagram flow chart that is calculated to be the long pattern subspace by one group of age growth pattern vector as shown in Figure 4.Initial, utilize method shown in Fig. 2 to generate one group of age growth pattern vector then, as training set.Notice that the age growth pattern vector in this set all comprises a large amount of missing values in general.Utilize again method shown in Fig. 3 will train the set in all missing values initialization.Enter step I then, use the principal component analysis (PCA) technology in the higher algebra textbook on the training set after initialization, for age growth pattern vector generate an eigen space that k orthogonal basis vector arranged the long pattern subspace, here k is a predefined subspace dimension, be integer, generally should be far smaller than the dimension of age growth pattern vector, for example 20.Step II is utilized the vector operations in the higher algebra textbook, and each the age growth pattern vector in the training set is projected in the one-tenth long pattern subspace that step I obtains.The subspace projection that Step II I utilizes Step II to obtain conversely reconstructs age growth pattern vectors all in the training set, is about to the transposition that projection vector multiply by projection matrix in the Step II.Step IV compares the age growth pattern vector that Step II I reconstruct is come out with original age growth pattern vector, original herein age growth pattern vector just obtains the former age growth pattern vector in the training set of homolographic projection vector, calculate the quadratic sum of the difference of the element of given value in both corresponding former data, as the reconstructed error of given value.With the average of the given value reconstructed error of institute's has age growth pattern vector in the training set as the reconstructed error of current one-tenth long pattern subspace to given value in the training set.Whether step V judges this reconstructed error less than certain pre-set threshold θ, for example 10
-2If not, then replace missing values in the former data, and then get back to step I, on the training set that upgraded, use principal component analysis (PCA) with the part of missing values in the corresponding former data in the vector that reconstruct is come out among the Step II I.This process constantly circulates, and judges reconstructed error less than θ up to step V, then withdraws from circulation, and the subspace that up-to-date principal component analysis (PCA) calculates is preserved as one-tenth long pattern subspace, finishes.Through this training process, can obtain one and describe the subspace that the age becomes the long pattern basic characteristics.The general character of the proprietary age one-tenth long pattern that comprises in the training set has been reflected in this subspace, and promptly therefore the universal law of growth of human age can be used as next step basis of carrying out estimation of Age and foundation.
A given facial image to be estimated treats that estimated image is placed on the age respectively and becomes in the long pattern on all possible age, extracts and waits to estimate the proper vector of facial image and generate one group of candidate's age growth pattern vector.For example, the hypothetical target age is 0 to 70 years old, and then this step can generate 71 candidate's age growth pattern vectors.In each candidate's age growth pattern vector, only on an age characteristics of image is arranged, remainder all is a missing values.As shown in Figure 5, the hypothetical target age is 0 to p-1 years old, then can be to treat that estimated image generates p candidate's age growth pattern vector, and missing values wherein is changed to m.Then, judge into the long pattern subspace and whether train, if not, then enter and be trained to the long pattern subspace, utilize the training of Fig. 4 describing method to obtain into the long pattern subspace; If then utilize process shown in Figure 5 that it is carried out estimation of Age.At first, according to the one-tenth long pattern subspace that trains, will wait that all candidate's age growth pattern vectors of estimating facial image project in the long pattern subspace, reconstruct complete candidate's age growth pattern vector from projection vector again.In Fig. 5, the p that obtains candidate's age growth pattern vector z
1Z
pObtain vectorial y after projecting to into the long pattern subspace
1Y
p, reconstruct complete candidate's age vector from projection vector again
At this moment, if treating estimated image has been placed on the correct age when generating candidate's age growth pattern vector, then reconstruct age of coming out becomes that the facial image on the corresponding age should be very alike with original image in the long pattern, otherwise twisted phenomena will appear in the facial image that reconstruct is come out.Two reconstructed images that for example amplify among Fig. 5, one of the left side with treat that estimated image is closely similar, and tangible distortion has taken place in of the right.Next step, by the reconstructed error between the reconstructed image proper vector relatively part of original image placement location (in the age growth pattern vector that comes out of reconstruct just corresponding to) and the original image proper vector, find candidate's age of reconstructed error minimum to become long pattern, promptly treat the best age of estimating facial image and become long pattern, and wait to estimate that the position r of facial image in this optimal candidate age one-tenth long pattern is the estimation to its age, and it is exported as final result.Characteristics of image vector that in Fig. 5, just reconstruct is come out and the reconstructed error ε between the original image proper vector
a(1) ... ε
a(P), candidate's age seeking out the reconstructed error minimum become long pattern just best age become long pattern, wait to estimate that facial image becomes the position r in the long pattern as a result of to export at this candidate's age this time.In this course, at first, facial image to be estimated both meets human age growth universal law for finding, promptly reconstructed error minimizes in age one-tenth long pattern subspace, has personalization features again, promptly all candidate's ages become long pattern all to become long pattern by the age that facial image to be estimated derives out, then, wait to estimate that facial image becomes the position in the long pattern just to indicate people's age in the image naturally in this best age.
By above description as can be seen, the inventive method only depend on a facial image can be to wherein personage's age judges.This method is based on a kind of special data structure---age growth pattern vector, rather than single image.Age becomes in the long pattern all images all from same individual, and arranges according to time sequencing, allows missing values to exist in this sequence.Like this, make this method well solve based on three main difficult points in the human estimation of Age of digital facial image, personalization, the imperfect and sequential problem of long pattern.Based on this, the present invention has designed and can training obtain into the method for long pattern subspace in the age growth pattern vector set that a large amount of missing values are arranged, and utilizes into the method that the age of individual input facial image is estimated in the long pattern subspace.After tested, the inventive method can reach and the human similar precision of estimation of Age ability to the stranger.
Claims (4)
1. automatic method of estimation of human age based on digital facial image is characterized in that step is:
The first step is obtained the gray level image facial image promptly to be estimated of user's face by Digital Image Input Device;
Second step, treat that estimated image is placed on the age respectively and becomes in the long pattern on all possible age, extraction is waited to estimate the proper vector of facial image and is generated one group of candidate's age growth pattern vector, in each candidate's age growth pattern vector, only on an age characteristics of image is arranged, remainder all is a missing values;
The 3rd step, judge into the long pattern subspace and whether train, if not, then enter and be trained to the long pattern subspace, if then enter next step;
In the 4th step,, will wait that all candidate's age growth pattern vectors of estimating facial image project in the long pattern subspace, reconstruct complete candidate's age growth pattern vector from projection vector again according to the one-tenth long pattern subspace that trains;
The 5th step, if treating estimated image was placed on the correct age, then reconstruct age of coming out becomes in the long pattern facial image and the original image on the corresponding age very alike, otherwise twisted phenomena will appear in the facial image that reconstruct is come out, by comparing the reconstructed error between reconstructed image and the original image, find candidate's age of reconstructed error minimum to become long pattern, promptly treat the best age of estimating facial image and become long pattern;
In the 6th step, wait to estimate that the position r of facial image in this optimal candidate age one-tenth long pattern is the estimation to its age, and it is exported as final result.
2. the automatic method of estimation of human age based on digital facial image as claimed in claim 1 is characterized in that the step that is trained to the long pattern subspace is:
Step 1: obtaining the one group of facial image collection with all ages and classes section that is used to train is training set;
Step 2: from the training set of step 1, generate one group of age one-tenth long pattern;
Step 3: extract the feature that the age becomes facial image in the long pattern, generate one group of age growth pattern vector;
Step 4: the age growth pattern vector that obtains with step 3 is a training set, is calculated to be the long pattern subspace.
3. the automatic method of estimation of human age based on digital facial image as claimed in claim 2 is characterized in that, step 3 extracts the feature that the age becomes facial image in the long pattern, and the specific practice that generates one group of age growth pattern vector is:
At first obtain the age growth pattern vector that all age in the training set becomes long pattern: become image projection in the long pattern in the subspace that n orthogonal vector base arranged that obtains by principal component analysis (PCA) an age in the training set, obtain a n dimensional feature vector, n is predefined integer, use the same method and obtain the proper vector that this age becomes the remaining image in the long pattern, should become the proper vector of all images in the long pattern to be spliced into a big vector then at the age according to the age order, just obtain this age growth pattern vector, the corresponding age of missing values becomes the facial image that lacks in the long pattern in this vector, obtains the age growth pattern vector of the institute's has age growth pattern in the whole training set then with same method;
Then, missing values is carried out initialization: the age that has shown a people on each row becomes long pattern, each lists and is identical facial image of age, wherein each row lists the facial image that contains several disappearances with each, with the disappearance of the width of cloth wherein image is the center, direction along this same age of disappearance image is same row, become in the long pattern to find some facial images at other people age, the proper vector of the facial image of the same age of then these being found but different people is averaged, average is filled into the position of disappearance image as initial value, all disappearance images in each age growth pattern vector in the training set are all carried out same processing, finish the initialization of training set.
4. the automatic method of estimation of human age based on digital facial image as claimed in claim 2 is characterized in that, step 4, and the age growth pattern vector that obtains with step 3 is a training set, the specific practice that is calculated to be the long pattern subspace is:
Step I uses the principal component analysis (PCA) technology on the training set after the initialization, for age growth pattern vector generate an eigen space that k orthogonal basis vector arranged the long pattern subspace, k is a sub spaces dimension, is predefined integer;
Step II is utilized vector operations, and each the age growth pattern vector in the training set is projected in the one-tenth long pattern subspace that step I obtains;
Step II I, the one-tenth long pattern subspace projection that utilizes Step II to obtain reconstructs age growth pattern vectors all in the training set, is about to the transposition that projection vector multiply by projection matrix in the Step II;
Step IV, the age growth pattern vector that Step II I reconstruct is come out is compared with original age growth pattern vector, calculate the quadratic sum of the difference of the element of given value in both corresponding former data, as the reconstructed error of given value, with the average of the given value reconstructed error of institute's has age growth pattern vector in the training set as the reconstructed error of current one-tenth long pattern subspace to given value in the training set;
Step V, whether judge this reconstructed error less than pre-set threshold θ, if not, then replace missing values in the former data with the part of missing values in the corresponding former data in the vector that reconstruct is come out among the Step II I, and then get back to step I, on the training set that upgraded, use principal component analysis (PCA);
If then the subspace that up-to-date principal component analysis (PCA) is calculated is as becoming the long pattern subspace to preserve; Finish.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100312182A CN101533468B (en) | 2009-04-27 | 2009-04-27 | Method for estimating human age automatically based on digital face images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100312182A CN101533468B (en) | 2009-04-27 | 2009-04-27 | Method for estimating human age automatically based on digital face images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101533468A true CN101533468A (en) | 2009-09-16 |
| CN101533468B CN101533468B (en) | 2012-05-23 |
Family
ID=41104053
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009100312182A Expired - Fee Related CN101533468B (en) | 2009-04-27 | 2009-04-27 | Method for estimating human age automatically based on digital face images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101533468B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101877054A (en) * | 2009-11-23 | 2010-11-03 | 北京中星微电子有限公司 | Method and device for determining age of face image |
| CN102306281A (en) * | 2011-07-13 | 2012-01-04 | 东南大学 | Multi-mode automatic estimating method for human age |
| CN102393907A (en) * | 2010-06-30 | 2012-03-28 | 卡西欧计算机株式会社 | Image processing apparatus, method, and program that classifies data of images |
| CN104486587A (en) * | 2014-12-20 | 2015-04-01 | 江阴市电工合金有限公司 | Juvenile type recognition method for electric alloy workshop |
| CN104992151A (en) * | 2015-06-29 | 2015-10-21 | 华侨大学 | Age estimation method based on TFIDF face image |
| CN107194868A (en) * | 2017-05-19 | 2017-09-22 | 成都通甲优博科技有限责任公司 | A kind of Face image synthesis method and device |
-
2009
- 2009-04-27 CN CN2009100312182A patent/CN101533468B/en not_active Expired - Fee Related
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101877054A (en) * | 2009-11-23 | 2010-11-03 | 北京中星微电子有限公司 | Method and device for determining age of face image |
| CN102393907A (en) * | 2010-06-30 | 2012-03-28 | 卡西欧计算机株式会社 | Image processing apparatus, method, and program that classifies data of images |
| CN102393907B (en) * | 2010-06-30 | 2014-05-07 | 卡西欧计算机株式会社 | Image processing apparatus and method |
| CN102306281A (en) * | 2011-07-13 | 2012-01-04 | 东南大学 | Multi-mode automatic estimating method for human age |
| CN102306281B (en) * | 2011-07-13 | 2013-11-27 | 东南大学 | Multi-mode automatic estimating method for humage |
| CN104486587A (en) * | 2014-12-20 | 2015-04-01 | 江阴市电工合金有限公司 | Juvenile type recognition method for electric alloy workshop |
| CN104486587B (en) * | 2014-12-20 | 2017-12-12 | 叶丽琴 | A kind of workshop minor identification system |
| CN104992151A (en) * | 2015-06-29 | 2015-10-21 | 华侨大学 | Age estimation method based on TFIDF face image |
| CN107194868A (en) * | 2017-05-19 | 2017-09-22 | 成都通甲优博科技有限责任公司 | A kind of Face image synthesis method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101533468B (en) | 2012-05-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101533468B (en) | Method for estimating human age automatically based on digital face images | |
| CN103649987B (en) | Face impression analysis method, beauty information providing method and face image generation method | |
| Lanitis et al. | Modeling the process of ageing in face images | |
| CN112002009B (en) | Unsupervised three-dimensional face reconstruction method based on generation of confrontation network | |
| CN106897671A (en) | A kind of micro- expression recognition method encoded based on light stream and FisherVector | |
| Punyani et al. | Human age-estimation system based on double-level feature fusion of face and gait images | |
| Lanitis et al. | Towards automatic face identification robust to ageing variation | |
| CN116188912A (en) | Training method, device, medium and equipment for image synthesis model of theme image | |
| CN113361646A (en) | Generalized zero sample image identification method and model based on semantic information retention | |
| Li et al. | Speckle noise removal based on structural convolutional neural networks with feature fusion for medical image | |
| Oh et al. | Deep visual discomfort predictor for stereoscopic 3d images | |
| Atzori et al. | Demographic bias in low-resolution deep face recognition in the wild | |
| CN116030077B (en) | Video salient region detection method based on multi-dataset collaborative learning | |
| Fujiwara et al. | Estimating image bases for visual image reconstruction from human brain activity | |
| CN113298895A (en) | Convergence guarantee-oriented unsupervised bidirectional generation automatic coding method and system | |
| Farazdaghi et al. | Backward face ageing model (B‐FAM) for digital face image rejuvenation | |
| Liu et al. | A3GAN: An attribute-aware attentive generative adversarial network for face aging | |
| Zhang et al. | ST-GAN: Unsupervised Facial Image Semantic Transformation Using Generative Adversarial Networks. | |
| CN110378979A (en) | The method automatically generated based on the generation confrontation customized high-resolution human face picture of network implementations | |
| Kumar | Nonlocal means image denoising using orthogonal moments | |
| CN110287761A (en) | A kind of face age estimation method analyzed based on convolutional neural networks and hidden variable | |
| Akamatsu et al. | Perceived image decoding from brain activity using shared information of multi-subject fmri data | |
| Ding et al. | InjectionGAN: unified generative adversarial networks for arbitrary image attribute editing | |
| CN114694074A (en) | Method, device and storage medium for generating video by using image | |
| CN112329736B (en) | Face recognition method and financial system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120523 Termination date: 20150427 |
|
| EXPY | Termination of patent right or utility model |