CN107424141A - A kind of face-image method for evaluating quality based on probability block - Google Patents
A kind of face-image method for evaluating quality based on probability block Download PDFInfo
- Publication number
- CN107424141A CN107424141A CN201710185733.0A CN201710185733A CN107424141A CN 107424141 A CN107424141 A CN 107424141A CN 201710185733 A CN201710185733 A CN 201710185733A CN 107424141 A CN107424141 A CN 107424141A
- Authority
- CN
- China
- Prior art keywords
- face
- image
- probability
- quality
- normalized
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention discloses a kind of face-image method for evaluating quality based on probability block, the described method comprises the following steps：Image is normalized using logarithmic transformation, amplifies hypo-intense pixels and compresses high intensity pixel, reduce the strength difference between the colour of skin；To every piece of normalized in image after conversion, zero-mean and unit variance are made it have, and extracts two-dimensional dct characteristic vector；The probability of individual features vector is calculated using position probabilistic model, oeverall quality fraction is produced by integrating local probability, reflects picture quality.Present invention, avoiding influence of the alignment error to recognition of face caused by being positioned as cast shadow, ambiguity and automatic face, can apply to solve the problems, such as face selection and the recognition of face based on video.
Description
Technical field
The present invention relates to face-image quality evaluation field, more particularly to a kind of face-image quality based on probability block to comment
Estimate method.
Background technology
With the fast development of monitoring system and the reduction of application cost, more and more monitoring systems are applied gives birth in people
Various aspects living, but due to low resolution, blurred picture, full spread position change and presence the problems such as low contrast so that
The identity based on video is inferred challenging under the conditions of monitoring[1]。
In recent years it has been proposed that many methods handle the recognition of face problem in Poor Image[2].A kind of method assumes that
Image is the outlier in sequence, however, when most of images in sequence have the quality of difference, these methods can be by well
The image classification of quality is exceptional value；Another method is explicit subset selection, carries out facial quality automatically to each image and comments
Estimate, select the subset being made up of high quality graphic.Recognition performance is improved, reduces overall calculated load, but is difficult to " facial matter
Amount " finds a definition well.
ISO/IEC 19794-5 and ICAO 9303 is the face-image standard for facial quality evaluation, based on above-mentioned mark
It is accurate, it has been proposed that many methods analyze various faces and image attributes.
Because face recognition performance is influenceed by Multiple factors simultaneously, one or two quality can be detected for robust
The selection of subset is inadequate.Nasrollahi and Moeslund[3]A kind of quality fusion method of weighting is proposed, to combine
Rotation, acutance, brightness and image resolution ratio quality outside face；Rua et al.[4]A similar method for evaluating quality is proposed, is used
Asymmetry analysis and two acutancees measure；Hsu et al.[5]Propose and learn fusion parameters on multiple mass fractions, with realize with
The maximum correlation of fraction is matched between face-image pair；But it is independent measurement due to various properties and has to facial quality
Different influences, the above method are difficult to their single mass fractions of combination output being used for image selection；Luo[6]Propose one kind
Method based on study, wherein quality model are trained to match the mass fraction of hand labeled.However, it is contemplated that the mankind mark
Subjective quality, this method may not produce the best in quality model for face recognition.
The significant challenge that face-image quality evaluation problem faces at present is：Because alignment error, attitudes vibration, image are cloudy
The presence of shadow and the not high various problems of image definition so that the optimal face-image of selection overall performance is by very big system
About；The presence of the problems such as cast shadow, automatic face positioning, very big difficulty is brought to recognition of face in video.
The content of the invention
The invention provides a kind of face-image method for evaluating quality based on probability block, and present invention, avoiding cloudy by projection
Shadow, ambiguity and influence of the alignment error to recognition of face caused by automatic face positioning, improve the degree of accuracy of identification, in detail
See below description：
A kind of face-image method for evaluating quality based on probability block, the described method comprises the following steps：
Image is normalized using logarithmic transformation, amplifies hypo-intense pixels and compresses high intensity pixel, reduce
Strength difference between the colour of skin；
To every piece of normalized in image after conversion, zero-mean and unit variance are made it have, and extract two-dimensional dct
Characteristic vector；
The probability of individual features vector is calculated using position probabilistic model, oeverall quality is produced by integrating local probability
Fraction, reflect picture quality.
Wherein, every piece of normalized after described pair of conversion in image, makes it have zero-mean and unit variance, and carry
The step of taking two-dimensional dct characteristic vector be specially：
To adapt to the contrast change between face-image, every piece in the image after conversion is normalized to equal with zero
Value and unit variance；And from each piece, the dimension DCT characteristic vectors of extraction 2, and exclude the 0th DCT points without normalization information
Amount, retains the preceding d low frequency component containing common face texture.
Wherein, the probability that individual features vector is calculated using position probabilistic model, produced by integrating local probability
Raw oeverall quality fraction, the step of reflecting picture quality be specially：
The model of each position is trained using the frontal face images with frontlighting and natural expression, all to instruct
Experienced face-image is scaled and snaps to fixed size first, each eyes is located at fixed position；
The probability of each block individual features vector is calculated using position probabilistic model, and assumes that the model of each position is
Independent, produce oeverall quality fraction by integrating local probability.
The beneficial effect of technical scheme provided by the invention is：
1st, this method has optimal overall performance, can identify that most positive, align good, illumination and sharp keen image；
2nd, for each given face image set, the method proposed is used to arrange image according to the quality of image
Sequence, the subset of top quality image is only included by selection, significantly improves recognition accuracy.
Brief description of the drawings
Fig. 1 is a kind of face-image method for evaluating quality based on probability block.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, embodiment of the present invention is made below further
It is described in detail on ground.
Embodiment 1
In order to solve problem above, it is desirable to be able to comprehensively, it is automatic, carry out facial selection and face recognition exactly.Research
Show：Block-based partial analysis need not position face by quantifying given face and the similitude of probability mask
Feature and single fraction is exported without recourse to each image in the case of fusion, can be directed to, alignment error is reflected by fraction
Degree, attitudes vibration, shade and image definition.The embodiment of the present invention proposes a kind of face-image matter based on probability block
Appraisal procedure is measured, it is described below referring to Fig. 1：
101：Image is normalized using logarithmic transformation, amplifies hypo-intense pixels and compresses high intensity pixel,
Reduce the strength difference between the colour of skin；
102：To every piece of normalized in image after conversion, zero-mean and unit variance are made it have, and extract two
Tie up DCT (Discrete Cosine Transform, discrete cosine transform) characteristic vector；
103：The probability of individual features vector is calculated using position probabilistic model, is produced totally by integrating local probability
Mass fraction, reflect picture quality.
Wherein, zero-mean and unit side are made it have to every piece of normalized in image after conversion in step 102
Difference, and be specially the step of extract two-dimensional dct characteristic vector：
To adapt to the contrast change between face-image, every piece in the image after conversion is normalized to equal with zero
Value and unit variance；And from each piece, the dimension DCT characteristic vectors of extraction 2, and exclude the 0th DCT points without normalization information
Amount, retains the preceding d low frequency component containing common face texture.
Further, the probability that individual features vector is calculated using position probabilistic model in step 103, passes through integration
Local probability produce oeverall quality fraction, reflect picture quality the step of be specially：
The model of each position is trained using the frontal face images with frontlighting and natural expression, all to instruct
Experienced face-image is scaled and snaps to fixed size first, each eyes is located at fixed position；
The probability of each block individual features vector is calculated using position probabilistic model, and assumes that the model of each position is
Independent, produce oeverall quality fraction by integrating local probability.
In summary, the embodiment of the present invention avoids right as caused by positioning cast shadow, ambiguity and automatic face
Influence of the quasi- error to recognition of face, improves the degree of accuracy of identification.
Embodiment 2
The scheme in embodiment 1 is further introduced with reference to specific calculation formula, Fig. 1, it is as detailed below
Description：
201：Image I is normalized using logarithmic transformation, amplifies hypo-intense pixels and compresses high intensity pixel,
Reduce the strength difference between the colour of skin；
For given image I, non-linear pretreatment (logarithmic transformation) is performed to reduce the dynamic range of data, under use
Formula calculates normalized image Ilog：
Ilog(r, c)=ln [I (r, c)+1]
Wherein, I (r, c) is the image pixel intensities positioned at (r, c) opening position.Logarithm normalization can amplify hypo-intense pixels simultaneously
High intensity pixel is compressed, helps to reduce the strength difference between the colour of skin.
202：To image I after conversionlogIn every piece of normalized, make it have zero-mean and unit variance, and extract
Two-dimensional dct characteristic vector；
Wherein, the image I after conversionlogIt is divided into N number of overlapping block, each block biSize with n × n pixel and
T pixel overlapping with adjacent block, in order to adapt to the change of the contrast between face-image, each piece is normalized to equal with zero
Value and unit variance.
From each piece, the dimension DCT characteristic vectors of extraction 2, the 0th DCT component is excluded, d low frequency component before remaining is low
Frequency component retains common face texture, and largely omits the customizing messages of individual, meanwhile, cast shadow and posture
Change with alignment thereof can change local grain.
203：Individual features vector x is calculated using position probabilistic modeliProbability, by integrate local probability produce it is total
Weight fraction, reflect picture quality.
The model of each position is trained using the frontal face images with frontlighting and natural expression, all to instruct
Experienced face-image is scaled and snaps to fixed size first, each eyes is located at fixed position.
For each piece of position i, characteristic vector x is calculated using position probabilistic modeliProbability：
Wherein, xiIt is characterized vector, μi、∑iIt is the average value and covariance matrix of normal distribution, T represents transposition, and d is represented
The number of low frequency component.
By assuming that the model of each position is independent, calculated using following formula by the N number of piece of image formed I entirety
Probability mass fraction Q (I)：
Resulting mass fraction represents the probability similitude of given face-image and " ideal " face-image, wherein, " reason
Think " face-image represents by one group of training image.
Mass fraction is higher, represents that picture quality is better, conversely, then picture quality is poorer.
In summary, the embodiment of the present invention avoids right as caused by positioning cast shadow, ambiguity and automatic face
Influence of the quasi- error to recognition of face, improves the degree of accuracy of identification.
Embodiment 3
Feasibility is carried out to the scheme in Examples 1 and 2 with reference to specific example, calculation formula, table 1, table 2, table 3
Checking, it is described below：
FERET and PIE databases are used for verifying performance of this method in terms of correct image of the selection with desired characteristic.
Choke Point databases are used for verifying that this method is effective in terms of image subset of the selection based on video is used for face recognition
Property.
FERET (The Facial Recognition Technology Database) database is field of face identification
One of most widely used face database.Wherein, ' fb' subsets have a blurred picture and alignment error image, alignment error by
Horizontal displacement and vertical displacement (0, ± 2, ± 4, ± 6, ± 8 pixels of displacement), face internal rotation (0 °, ± 10 °, ± 20 °, ±
30 °), change of scale (using 0.7,0.8,0.9,1.0,1.1,1.2,1.3 as zoom factor) etc. causes.The subset can be used for
Simulate blurred picture and four alignment error images.
PIE (Pose Illumination Expression) database is created by Carnegie Mellon Univ USA, comprising
The face-image of multi-pose, illumination and expression.It illuminates subset and is used to assess the performance under various projection conditions.In embodiment,
Front view image is divided into by six subsets based on light-source angle, subset 1 has most front light sources, and subset 6 has maximum
Light source angle (54 ° -67 °).The subset can be used for verifying the performance of the method proposed in the presence of cast shadow.
Choke Point databases are sets of video data, designed for carrying out face under the conditions of the monitoring of real world
Identification or the experiment of checking, data set is by 29 in 25 subjects (19 males and 6 women) and entrance 2 in entrance 1
Subject (23 males and 6 women) forms, totally 48 video sequences and 64204 face-images.
This method and following three kinds of methods are contrasted, proposed method for evaluating quality is assessed and quality is provided at the same time
How the face-image of best quality is determined during good and ropy face-image：
Asym_shrp[4]：A kind of fraction fusion method of asymmetric analysis and two acutance analyses based on pixel；
Gabor_asym[7]：A kind of asymmetric analytic approach based on Gabor characteristic；
DFFS[8]：(Distance From Face Space) is a kind of classical apart from space of planes method.
In order to prove validity of this method to various facial-recognition security systems quality evaluations, without loss of generality, make respectively
With multizone histogram (Multi-Region Histograms, referred to as MRH) and local binary patterns (Local Binary
Patterns, referred to as LBP) from each face-image extract feature；Use mutual subspace method (Mutual Subspace
Method, referred to as MSM) face collection is classified.
Table 1
Shown by the result of the displacement shown in table 1, rotation and dimensional variation, this method in most of changes as one man
Realize optimal or close to optimal performance；Performances of the Gabor_asym for image of the detection with the change of various definition
It is bad；Asym_shrp solves this problem by combining asymmetric analysis and two image definition measurements, nevertheless,
Its general performance is still very poor；The performance of DFFS alignment errors fails to detect with optimal acutance less than the method proposed
Image.
Table 2
Show that, even if cast shadow be present, this method can also obtain by the result in 6 PIE illumination subsets in table 2
Good result；Asym_shrp obtains optimum performance (front mark is the frontlighting face of high quality)；Compared to it
Under, Gabor_asym obscures between subset 1 and 4, and DFFS is mistakenly by most of face-images in subset 4 (comprising aobvious
Write shade) it is labeled as that there is first water.
Table 3
Show that number of subsets changes to 16 from 4, no by the recognition of face performance verification result based on video database in table 3
By which kind of facial feature extraction algorithm used, the mass measuring method proposed as one man obtains more more preferable than other three kinds of methods
Face verification performance.The experiment show feasibility and superiority of this method.
Bibliography：
[1] Dong Haibo monitor videos image quality measure [D] Shanghai Communications Universitys, 2013.
[2] research [D] Hebei University Of Engineering of the recognition of face key technology in Yang Yu video images, 2014.
[3]K.Nasrollahi and T.B.Moeslund.Face quality assessment system in
video sequences.In BIOID,Lecture Notes in Computer Science(LNCS),volume 5372,
pages 10–18,2008..
[4]E.A.Rúa,J.L.A.Castro,and C.G.Mateo.Quality-based score
normalization and frame selection for video-based person authentication.In
BIOID,Lecture Notes in Computer Science(LNCS),pages 1–9,2008.
[5]R.-L.V.Hsu,J.Shah,and B.Martin.Quality assessment of facial
images.In Biometrics Symposium,2006.
[6]H.Luo.A training-based no-reference image quality assessment
algorithm.In International Conference on Image Processing(ICIP),pages 2973–
2976,2004.
[7]J.Sang,Z.Lei,and S.Z.Li.Face image quality evaluation for ISO/IEC
standards 19794-5 and 29794-5.In ICB,Lecture Notes in Computer Science(LNCS),
volume 5558,pages 229–238,2009.
[8]H.Bae and S.Kim.Real-time face detection and recognition using
hybrid-information extracted from face space and facial features.Image and
Vision Computing,23(13):1181–1191,2005.
It will be appreciated by those skilled in the art that accompanying drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention
Sequence number is for illustration only, does not represent the quality of embodiment.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc., it should be included in the scope of the protection.
Claims (3)
1. a kind of face-image method for evaluating quality based on probability block, it is characterised in that the described method comprises the following steps：
Image is normalized using logarithmic transformation, amplifies hypo-intense pixels and compresses high intensity pixel, reduce the colour of skin
Between strength difference；
To every piece of normalized in image after conversion, zero-mean and unit variance are made it have, and extracts two-dimensional dct feature
Vector；
The probability of individual features vector is calculated using position probabilistic model, oeverall quality point is produced by integrating local probability
Number, reflect picture quality.
2. a kind of face-image method for evaluating quality based on probability block according to claim 1, it is characterised in that described
To every piece of normalized in image after conversion, zero-mean and unit variance are made it have, and extracts two-dimensional dct characteristic vector
The step of be specially：
To adapt to the contrast change between face-image, every piece in the image after conversion is normalized to zero-mean and
Unit variance；And from each piece, the dimension DCT characteristic vectors of extraction 2, and the 0th DCT component without normalization information is excluded,
Retain the preceding d low frequency component containing common face texture.
3. a kind of face-image method for evaluating quality based on probability block according to claim 1, it is characterised in that described
The probability of individual features vector is calculated using position probabilistic model, produces oeverall quality fraction by integrating local probability, instead
The step of reflecting picture quality be specially：
The model of each position trained using the frontal face images with frontlighting and natural expression, all to be trained
Face-image is scaled and snaps to fixed size first, each eyes is located at fixed position；
The probability of each block individual features vector is calculated using position probabilistic model, and assumes that the model of each position is independent
, produce oeverall quality fraction by integrating local probability.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710185733.0A CN107424141B (en) | 2017-03-26 | 2017-03-26 | Face image quality evaluation method based on probability block |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710185733.0A CN107424141B (en) | 2017-03-26 | 2017-03-26 | Face image quality evaluation method based on probability block |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107424141A true CN107424141A (en) | 2017-12-01 |
| CN107424141B CN107424141B (en) | 2020-07-28 |
Family
ID=60424254
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710185733.0A Active CN107424141B (en) | 2017-03-26 | 2017-03-26 | Face image quality evaluation method based on probability block |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107424141B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108447058A (en) * | 2018-03-30 | 2018-08-24 | 北京理工大学 | A kind of image quality evaluating method and system |
| CN110070010A (en) * | 2019-04-10 | 2019-07-30 | 武汉大学 | A kind of face character correlating method identified again based on pedestrian |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050069214A1 (en) * | 2003-09-29 | 2005-03-31 | Casio Computer Co., Ltd. | Image compression method and image compression device |
| CN101179707A (en) * | 2007-09-21 | 2008-05-14 | 清华大学 | Wireless network video image multi-visual angle cooperation target tracking measurement method |
| CN102855631A (en) * | 2012-08-23 | 2013-01-02 | 浙江大学 | Method for extracting visual energy information for image quality evaluation |
| US20130182952A1 (en) * | 2010-09-17 | 2013-07-18 | I.C.V.T Ltd. | Method of classifying a chroma downsampling error |
| CN103996192A (en) * | 2014-05-12 | 2014-08-20 | 同济大学 | Non-reference image quality evaluation method based on high-quality natural image statistical magnitude model |
| US20140241600A1 (en) * | 2013-02-25 | 2014-08-28 | Siemens Aktiengesellschaft | Combined surface reconstruction and registration for laparoscopic surgery |
| CN105225206A (en) * | 2014-11-22 | 2016-01-06 | 四川大学 | A kind of outdoor optical based on image is according to environment rebuilt method |
| US9256950B1 (en) * | 2014-03-06 | 2016-02-09 | Google Inc. | Detecting and modifying facial features of persons in images |
| CN105976351A (en) * | 2016-03-31 | 2016-09-28 | 天津大学 | Central offset based three-dimensional image quality evaluation method |
| CN106127741A (en) * | 2016-06-17 | 2016-11-16 | 西安电子科技大学 | Non-reference picture quality appraisement method based on improvement natural scene statistical model |
-
2017
- 2017-03-26 CN CN201710185733.0A patent/CN107424141B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050069214A1 (en) * | 2003-09-29 | 2005-03-31 | Casio Computer Co., Ltd. | Image compression method and image compression device |
| CN101179707A (en) * | 2007-09-21 | 2008-05-14 | 清华大学 | Wireless network video image multi-visual angle cooperation target tracking measurement method |
| US20130182952A1 (en) * | 2010-09-17 | 2013-07-18 | I.C.V.T Ltd. | Method of classifying a chroma downsampling error |
| CN102855631A (en) * | 2012-08-23 | 2013-01-02 | 浙江大学 | Method for extracting visual energy information for image quality evaluation |
| US20140241600A1 (en) * | 2013-02-25 | 2014-08-28 | Siemens Aktiengesellschaft | Combined surface reconstruction and registration for laparoscopic surgery |
| US9256950B1 (en) * | 2014-03-06 | 2016-02-09 | Google Inc. | Detecting and modifying facial features of persons in images |
| CN103996192A (en) * | 2014-05-12 | 2014-08-20 | 同济大学 | Non-reference image quality evaluation method based on high-quality natural image statistical magnitude model |
| CN105225206A (en) * | 2014-11-22 | 2016-01-06 | 四川大学 | A kind of outdoor optical based on image is according to environment rebuilt method |
| CN105976351A (en) * | 2016-03-31 | 2016-09-28 | 天津大学 | Central offset based three-dimensional image quality evaluation method |
| CN106127741A (en) * | 2016-06-17 | 2016-11-16 | 西安电子科技大学 | Non-reference picture quality appraisement method based on improvement natural scene statistical model |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108447058A (en) * | 2018-03-30 | 2018-08-24 | 北京理工大学 | A kind of image quality evaluating method and system |
| CN108447058B (en) * | 2018-03-30 | 2020-07-14 | 北京理工大学 | Image quality evaluation method and system |
| CN110070010A (en) * | 2019-04-10 | 2019-07-30 | 武汉大学 | A kind of face character correlating method identified again based on pedestrian |
| CN110070010B (en) * | 2019-04-10 | 2022-06-14 | 武汉大学 | Face attribute association method based on pedestrian re-recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107424141B (en) | 2020-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108009531B (en) | Multi-strategy anti-fraud face recognition method | |
| Wong et al. | Patch-based probabilistic image quality assessment for face selection and improved video-based face recognition | |
| Debiasi et al. | PRNU-based detection of morphed face images | |
| US7848548B1 (en) | Method and system for robust demographic classification using pose independent model from sequence of face images | |
| CN110490158B (en) | Robust face alignment method based on multistage model | |
| CN110378931A (en) | A kind of pedestrian target motion track acquisition methods and system based on multi-cam | |
| Debiasi et al. | PRNU variance analysis for morphed face image detection | |
| WO2016145940A1 (en) | Face authentication method and device | |
| US20120183212A1 (en) | Identifying descriptor for person or object in an image | |
| CN111126240B (en) | Three-channel feature fusion face recognition method | |
| CN102592141A (en) | Method for shielding face in dynamic image | |
| CN107038400A (en) | Face identification device and method and utilize its target person tracks of device and method | |
| CN106407966B (en) | A kind of face identification method applied to attendance | |
| Rukhiran et al. | Effecting of environmental conditions to accuracy rates of face recognition based on IoT solution | |
| CN103605993A (en) | Image-to-video face identification method based on distinguish analysis oriented to scenes | |
| Winarno et al. | Face recognition based on symmetrical half-join method using stereo vision camera | |
| CN110969101A (en) | Face detection and tracking method based on HOG and feature descriptor | |
| CN107424141A (en) | A kind of face-image method for evaluating quality based on probability block | |
| CN109886212A (en) | From the method and apparatus of rolling fingerprint synthesis fingerprint on site | |
| Méndez-Llanes et al. | On the use of local fixations and quality measures for deep face recognition | |
| CN113610071A (en) | Face living body detection method and device, electronic equipment and storage medium | |
| Saini et al. | Improvement in copy-move forgery detection using hybrid approach | |
| Abboud et al. | Quality based approach for adaptive face recognition | |
| Kakadiaris et al. | Face recognition using 3D images | |
| KR100711223B1 (en) | Face recognition method using Zernike/LDA and recording medium storing the method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |