CN103188427B - Image capture unit and the control method thereof of image feature value group can be simplified - Google Patents
Image capture unit and the control method thereof of image feature value group can be simplified Download PDFInfo
- Publication number
- CN103188427B CN103188427B CN201110454290.3A CN201110454290A CN103188427B CN 103188427 B CN103188427 B CN 103188427B CN 201110454290 A CN201110454290 A CN 201110454290A CN 103188427 B CN103188427 B CN 103188427B
- Authority
- CN
- China
- Prior art keywords
- eigenvalue
- group
- image
- cluster
- eigenvalue cluster
- 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.)
- Active
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention discloses a kind of image capture unit simplifying image feature value group and control method thereof.This image capture unit comprises Feature Conversion module, feature reduction module, template storage module and identification module.The video conversion that this image capture unit is captured is characterized image by Feature Conversion module, and this characteristic image comprises an eigenvalue group.Feature reduction module performs simplification program according to a look-up table, to produce the eigenvalue group after simplifying.Finally, the eigenvalue group after identification module will simplify is compared with several determinand templates (Template) being stored in template storage module, with the determinand in identification image.The image capture unit of the tool object identification function that the present invention can solve known technology need to expend substantial amounts of memory body and calculation resources, the problem causing its high cost.
Description
Technical field
The present invention relates to a kind of image capture unit and control method thereof, the video conversion of acquisition can be characterized image in particular to a kind of, and simplify the eigenvalue cluster in characteristic image, to reduce the memory body demand of image capture unit and to save image capture unit and the control method thereof of its calculation resources.
Background technology
In recent years, due to keen competition on market, electronic product more and more par, function is the most powerful.Wherein, object detecting function has begun to be added in image capture unit miscellaneous.Such as, digital camera (Digital
Still Camera), digital code camera (Digital
Camera), monitor (Monitor) even camera cell phone (Camera
Phone) etc., and there is multiple application, the such as identification of human face recognition, pedestrian, car plate, vehicle recognition or other object.But generally requiring the substantial amounts of memory body of cost to store the data such as skin detection (Template) owing to possessing the image capture unit of object detecting function, with greater need for consuming huge calculation resources, therefore the cost of image capture unit also can significantly increase.
Above-mentioned situation is obvious especially for small hand-held camera head, and small hand-held camera head is due to considering on cost, and its resource is the most fairly limited, is most particularly obvious in the use of memory body, and the memory body cost of higher-order is the highest.Therefore, if being intended to by object detecting Function Integration Mechanism on small hand-held camera head, the demand of its memory body is also inevitable significantly to be increased, and direct reaction is in its manufacturing cost, causes the competitiveness of product to decline.
Therefore, how to develop a kind of image capture unit, possess object detecting function, and the hardware cost that under the premise not affecting object detecting accuracy, can be effectively saved image capture unit is the problem that the present invention to be solved.
Summary of the invention
Because the problem of above-mentioned known technology, the purpose of the present invention is exactly to provide a kind of image capture unit simplifying image feature value group and control method thereof, substantial amounts of memory body and calculation resources need to be expended, the problem causing its high cost with the image capture unit of the tool object identification function of solution known technology.
For achieving the above object, the present invention takes techniques below scheme:
The present invention proposes a kind of image capture unit simplifying image feature value group, comprises: Feature Conversion module, and video conversion to be measured is characterized image, and characteristic image comprises eigenvalue group；Feature reduction module, according to a threshold value, perform simplification program, from eigenvalue group, the eigenvalue cluster that would not exist in look-up table is deleted, each eigenvalue cluster being deleted is then pre-conditioned and there is the eigenvalue cluster of this look-up table and replace to meet one, makes eigenvalue the group constant but kind of sum reduce, produces the eigenvalue group after simplifying；Template stores module, stores several determinand templates (Template)；And identification module, the eigenvalue group after this simplification is compared with several determinand templates, with the determinand in identification image to be measured.
According to the purpose of the present invention, proposing again a kind of control method simplifying image feature value group, comprise the steps of utilization and by video conversion to be measured, Feature Conversion module is characterized image, characteristic image comprises eigenvalue group；By feature reduction module according to a look-up table, perform simplification program, from eigenvalue group, the eigenvalue cluster that would not exist in look-up table is deleted, each eigenvalue cluster being deleted is then to meet eigenvalue cluster that is pre-conditioned and that be present in look-up table replacement, make eigenvalue group sum constant but kind minimizing, produce the eigenvalue group after simplifying；Stored module by template and store several determinand templates；And by identification module will simplify after eigenvalue group compare with several determinand templates, with the determinand in identification image to be measured.
Preferably, pre-conditioned being determined by diversity, each value indicative group being deleted need to replace with the eigenvalue cluster minimum with its diversity.
Preferably, look-up table is set up less than the eigenvalue cluster of a complexity (Entropy) threshold value with the complexity of eigenvalue cluster.
Preferably, after performing simplification program, in the eigenvalue group after simplification, each eigenvalue cluster all can find the eigenvalue cluster corresponding with after its transposition.
Preferably, after performing simplification program, in the eigenvalue group after simplification, each eigenvalue cluster all can find the eigenvalue cluster corresponding with after its mirror.
Preferably, several determinand templates are according to being present in the eigenvalue cluster of look-up table, and via advancing algorithm, support vector machine algorithm or Principal Component Analysis Method+linearly have identification force analysis calculation to be set up.
The invention has the beneficial effects as follows: from the above, the image capture unit simplifying image feature value group under this invention and control method thereof, it can have one or more following advantage:
(1) this image capture unit that can simplify image feature value group and control method thereof can simplify the kind of the eigenvalue cluster in characteristic image, therefore, it is possible to the memory body demand of image capture unit is greatly reduced, and then reduce its manufacturing cost.
(2) this image capture unit that can simplify image feature value group and control method thereof can reduce the kind of the eigenvalue cluster in characteristic image, also therefore are able to significantly save the calculation resources of image capture unit, improve its efficiency.
(3) this image capture unit that can simplify image feature value group and control method thereof are that the eigenvalue cluster with diversity minimum is to replace each eigenvalue cluster being deleted rather than directly to delete, thus without the accuracy reducing object detecting.
(4)
Although the kind of the eigenvalue cluster in characteristic image is simplified by this image capture unit that can simplify image feature value group and control method thereof, but each eigenvalue cluster in the eigenvalue group after Jian Huaing all can find the eigenvalue cluster corresponding with after its transposition or after mirror, therefore the characteristic of former eigenvalue group can be retained, to perform relevant application.
Accompanying drawing explanation
Fig. 1 is the block chart of the first embodiment of the image capture unit of the simplified image feature value group of the present invention.
Fig. 2 is the flow chart of the first embodiment of the image capture unit of the simplified image feature value group of the present invention.
Fig. 3 is the block chart of the second embodiment of the image capture unit of the simplified image feature value group of the present invention.
One of Fig. 4 schematic diagram of an embodiment being characterized value group.
Fig. 5 is characterized the two of the schematic diagram of an embodiment of value group.
Fig. 6 is the flow chart of the second embodiment of the image capture unit of the simplified image feature value group of the present invention.
Fig. 7 is the schematic diagram of the 3rd embodiment of the image capture unit of the simplified image feature value group of the present invention.
Fig. 8 is the schematic diagram of the 4th embodiment of the image capture unit of the simplified image feature value group of the present invention.
Fig. 9 is the flow chart of the control method of the simplified image feature value group of the present invention.
Drawing reference numeral: 1: image capture unit；3,7,8: digital camera；10,30,70,80: image；11,31: Feature Conversion module；111,311: characteristic image；1111,3111,81: eigenvalue group；12,32: feature reduction module；121,321: the eigenvalue group after simplification；13: template stores module；131: several determinand templates；14: identification module；141: identification result；33: face template stores module；331: several face templates；34: human face recognition module；341,82: limitting casing；41,51: validity feature value group；42,52: invalid eigenvalue cluster；71: form；710,711: square；S21 ~ S24, S61 ~ S65, S91 ~ S94: steps flow chart.
Detailed description of the invention
Hereinafter with reference to correlative type, the image capture unit simplifying image feature value group under this invention and the embodiment of control method thereof being described, for making to readily appreciate, the similar elements in following embodiment illustrates with identical symbology.
Refer to Fig. 1, for the block chart of first embodiment of image capture unit of the simplified image feature value group of the present invention.As it can be seen, the image capture unit 1 of the present invention comprises Feature Conversion module 11, feature reduction module 12, template storage module 13 and identification module 14.The image 10 that image capture unit 1 captures can be converted to characteristic image 111 by Feature Conversion module 11, and this characteristic image 111 comprises an eigenvalue group 1111 being made up of several eigenvalue clusters.
Feature reduction module 12 can perform simplification program according to by look-up table, delete the eigenvalue cluster being not present in look-up table, each be deleted eigenvalue cluster then with meet one pre-conditioned and be present in look-up table eigenvalue cluster replace, with produce simplify after eigenvalue group 121.
It is noted that this pre-conditioned can determine by the diversity between eigenvalue cluster.Each deleted eigenvalue cluster can utilize the eigenvalue cluster minimum with its diversity to replace rather than directly delete the eigenvalue cluster being not present in look-up table.This mode is advantageous in that, the expression ability of the eigenvalue cluster being deleted can be replaced by the eigenvalue cluster minimum with its diversity, the most still can be retained, and can reduce again the kind of eigenvalue cluster in eigenvalue group 1111 simultaneously.So, i.e. can not affect under the situation of object detecting accuracy of image capture unit, be effectively reduced the memory body demand of image capture unit 1.
Finally, the eigenvalue group 121 after identification module 14 then just simplifies is compared with several determinand templates 131 being stored in template storage module 13, to produce an identification result 141, gets final product the determinand in identification image 10.
Although it is noted that the present invention deletes the eigenvalue cluster being not present in look-up table, but each eigenvalue cluster after simplifying still can find the eigenvalue cluster corresponding with after its transposition or after mirror.It is to say, the eigenvalue group 121 after Jian Huaing still can retain the characteristic of former eigenvalue cluster 1111, therefore can also utilize this characteristic, perform the application on various object detecting.
Certainly, several determinand templates 131 being stored in template storage module 13 are also to be set up with the eigenvalue cluster of the condition set by coincidence gate threshold value, so could compare with the eigenvalue group 121 after simplifying, with identification object.
Refer to Fig. 2, for the flow chart of first embodiment of image capture unit of the simplified image feature value group of the present invention.
In the step s 21, image capture unit an image is captured.
In step S22, Feature Conversion module this video conversion being characterized image, this characteristic image comprises an eigenvalue group.
In step S23, simplification program is performed via feature reduction module, delete in this eigenvalue group, it is not present in the eigenvalue cluster of look-up table, each eigenvalue cluster being deleted then replaces, to produce the eigenvalue group after simplifying being present in this look-up table and the eigenvalue cluster minimum with its diversity.
In step s 24, the eigenvalue group after just being simplified by identification module is compared with several determinand templates being stored in template storage module, with the determinand in identification image.
Image capture unit and the control method thereof of the simplified image feature value group of the present invention also apply be applicable to any camera head with object identification function.Such as, digital camera (Digital
Still Camera), digital code camera (Digital
Camera), monitor (Monitor) is even according to mobile phone (Camera
Phone) etc., to perform face, pedestrian, car plate, vehicle or the identification of other various objects.Following embodiment is as a example by digital camera identification face, but the present invention is not limited thereto.
Refer to Fig. 3, for the block chart of the second embodiment of image capture unit of the simplified image feature value group of the present invention.As it can be seen, the digital camera 3 of the present invention comprises Feature Conversion module 31, feature reduction module 32, face template storage module 33 and human face recognition module 34.The image 30 that digital camera 3 captures can be converted to characteristic image 311 by Feature Conversion module 31, and this characteristic image 311 comprises an eigenvalue group 3111 being made up of several eigenvalue clusters.
Feature reduction module 32 can perform simplification program according to by a look-up table, this look-up table is that complexity (Entropy) threshold value according to each eigenvalue cluster determines, complexity (entropy) is then deleted higher than the eigenvalue cluster of this threshold value, complexity then retains less than the eigenvalue cluster of this threshold value, each eigenvalue cluster being deleted then replaces with the eigenvalue cluster minimum with its diversity, to produce the eigenvalue group 321 after simplifying.
In addition, for keeping characteristics value group 3111 characteristic originally, feature reduction module 32 more can retain eigenvalue cluster corresponding with after each eigenvalue cluster transposition or after mirror in the eigenvalue group 321 after simplification, to perform various human face recognition function.
Finally, eigenvalue group 321 after human face recognition module 34 then just simplifies is compared with several face templates 331 being stored in face template storage module 33, produce a limitting casing 341, getting final product the face in identification image 30, after completing identification, user i.e. may utilize the work that digital camera 3 carries out focusing and shooting.
It addition, several face templates 331 are to be present in look-up table, the eigenvalue cluster that i.e. entropy is relatively low, and via advancing algorithm (Boosting), support vector machine algorithm (Support
Vector Mechine, SVM) or Principal Component Analysis Method+linearly there is identification force analysis (Principle
Component Analysis+Linear Discriminant Analysis, PCA+LDA) calculate and obtain.
It is worth mentioning that, when video conversion is characterized image by the image capture unit of known technology, the utilization rate of each eigenvalue cluster that this characteristic image is comprised is not quite similar, and wherein some is frequently used, and wherein has another part seldom to use.For example, a resolution is the image of 1920 × 1080, and each of which point all comprises an eigenvalue cluster, altogether comprises 500 kinds of different eigenvalue clusters, and wherein only has 250 kinds and be frequently used, seldom uses for other 250 kinds.The frequency that eigenvalue cluster uses is then relevant with its complexity or entropy (Entropy).It is said that in general, the relatively low eigenvalue cluster of entropy relatively has systematicness, represent that power is relatively strong, use frequency the highest.
And concept proposed by the invention, utilize the entropy analyzing each eigenvalue cluster just, to set up a look-up table, and will the most relatively have systematicness according to this look-up table, expression power is relatively strong and uses the eigenvalue cluster that frequency is higher to be retained, and deletes the eigenvalue cluster that complexity is higher.Each eigenvalue cluster being deleted is then with minimum with its diversity and be present in this look-up table and the relatively low eigenvalue cluster of entropy is replaced.So, not only can retain the expression ability of each eigenvalue cluster being deleted, the quantity of eigenvalue cluster kind can also be reduced simultaneously.
Refer to Fig. 4, one of schematic diagram of an embodiment being characterized value group.Can be evident that by figure, in validity feature value group 41, the arrangement of each eigenvalue relatively has systematicness, and in invalid eigenvalue cluster 42, the arrangement of each eigenvalue is the most at random.This invalid eigenvalue cluster 42 is the most at random due to its arrangement, and therefore it represents that power is poor, and uses frequency low.And inventive feature simplifies this threshold value of module, analyzing the entropy of each eigenvalue cluster, the invalid eigenvalue cluster 42 that entropy is not inconsistent unification complexity threshold value is deleted.
Refer to the 5th figure, be characterized an embodiment of value group schematic diagram two.As shown in FIG., its arrangement of invalid eigenvalue cluster 52 is the most at random, and entropy is higher.After invalid eigenvalue cluster 52 is deleted, inventive feature simplifies module and then utilizes relatively low with its diversity, then this invalid eigenvalue cluster 52 of systematicness preferably validity feature value group 51 replacement.
Can significantly be found out by figure, between validity feature value group 51 and invalid eigenvalue cluster 52, only exist the difference of an eigenvalue.Inventive feature simplifies the module then available entropy analyzing each eigenvalue, finds out the validity feature value group 51 minimum with invalid eigenvalue cluster 52 diversity being deleted to replace this invalid eigenvalue cluster 52.And validity feature value group 51 is close with the expression power of invalid eigenvalue 52, invalid eigenvalue 52 therefore can be replaced completely.Therefore, the eigenvalue group after inventive feature simplifies module simplifying will not lose its expression power originally, and image capture unit still can maintain original object identification ability accurately.It addition, the representation of the eigenvalue cluster of different algorithms is different, such as, has different arrow numbers etc., but method proposed by the invention all can be used to be simplified.
It is noted that owing to the quantity of eigenvalue cluster kind and the memory body demand of image capture unit are breezy relevant, can significantly reduce the memory body demand of image capture unit hence with the eigenvalue group after the method simplification of the present invention.On the other hand, due to the minimizing of eigenvalue cluster kind, it is also possible to be effectively saved the calculation resources of image capture unit, make image capture unit more efficiently.And via actual test, under the premise not affecting detecting quality, the memory internal body (Internal of digital camera
RAM) 34% is decreased, the size (ROM of ROM
Size) 36% is decreased, the size of memory body needed for therefore the present invention can actually reduce image capture unit.
Refer to Fig. 6, for the flow chart of the second embodiment of image capture unit of the simplified image feature value group of the present invention.
In step S61, capture an image by digital camera.
In step S62, by Feature Conversion module, this video conversion being characterized image, this characteristic image comprises an eigenvalue group.
In step S63, by feature reduction module according to a look-up table, to delete the eigenvalue cluster of entropy high complexity threshold value, retain the entropy eigenvalue cluster less than this complexity threshold value, each eigenvalue cluster being deleted then replaces with the eigenvalue cluster minimum with its diversity, to produce the eigenvalue group after simplifying.
In step S64, the eigenvalue group after then just being simplified by human face recognition module is compared with several face templates being stored in face template storage module, with the face in identification image.
In step S65, digital camera is utilized to carry out the work focused and shoot.
Refer to Fig. 7, for the schematic diagram of the 3rd embodiment of image capture unit of the simplified image feature value group of the present invention.The present embodiment is with class Hull feature algorithm (Haar-Like
Features) as a example by.As shown in FIG., Haar-Like algorithm is the image 70 utilizing a form 71 scanning digital camera 7 to be captured, then the square that form 71 is divided into two or more than two is mutually added and subtracted, and is reconverted into feature signal.As shown in FIG., square 710 is positioned at the position of face forehead, the signal shallower for color that therefore square 710 is captured, and square 711 is positioned at the position of face eyes, and therefore square 711 captures the signal that color is deeper.And the signal captured by square 711 and square 710 subtract each other and can learn that the signal that both are captured has bigger difference, therefore can determine that this form 71 is probably the position at face place.
It is relatively low that above-mentioned mode belongs to entropy, relatively have systematicness and the comparison method that is relatively often used, certainly also have in form 71, the comparison method simultaneously mutually added and subtracted with multiple squares being positioned at diverse location, but this method less systematicness, the most less used, it is possible to use the relatively low comparison method of entropy replaces.By method proposed by the invention, simplification program can be performed by look-up table, delete some entropy higher, the comparison method being less often used, the most then can be effectively reduced the memory body demand of digital camera 7.
Refer to Fig. 8, for the schematic diagram of the 4th embodiment of image capture unit of the simplified image feature value group of the present invention.The present embodiment is with histograms of oriented gradients algorithm (Histogram
Of Oriented Gradient, HOG) idea of the invention is described.
As it can be seen, user uses digital camera 8 to shoot personage, the Feature Conversion module of digital camera 8 utilizes HOG algorithm that image 80 is converted to the eigenvalue group 81 being made up of several eigenvalue clusters.Feature reduction module then can perform simplification program according to a look-up table, is deleted higher than the eigenvalue cluster of this entropy threshold value by wherein entropy, and the eigenvalue cluster using its diversity minimum replaces.After this is then simplified by human face recognition module, result stores module comparison with being stored in face template in advance, can produce limitting casing 82.After completing the work of human face detection again, user just can carry out focusing and the work such as shooting.
Although it is aforementioned during the image capture unit of the simplified image feature value group of the explanation present invention, the concept of the control method of the simplified image feature value group of the present invention of the present invention is described the most simultaneously, but for clarification, the most separately illustrate flow chart to describe in detail.
Refer to Fig. 9, for the flow chart of control method of the simplified image feature value group of the present invention.
In step S91, utilizing Feature Conversion module that video conversion to be measured is characterized image, this characteristic image comprises an eigenvalue group.
In step S92, by feature reduction module according to a threshold value, perform simplification program, from this eigenvalue group, the eigenvalue cluster that would not exist in look-up table is deleted, each eigenvalue cluster being deleted, then to meet eigenvalue cluster that is pre-conditioned and that be present in this look-up table replacement, makes eigenvalue group sum constant but kind minimizing, produces the eigenvalue group after simplifying.
In step S93, template store module and store several determinand templates.
In step S94, the eigenvalue group after being simplified by identification module is compared with several determinand templates, with the determinand in identification image to be measured.
The detailed description of the control method of the simplified image feature value group of the present invention and embodiment described in time above describing the image capture unit of simplified image feature value group of the present invention, and narration is just not repeated for schematic illustration at this.
In sum, the image capture unit of the simplified image feature value group of the present invention and control method thereof can simplify the kind of the eigenvalue cluster in characteristic image, therefore, it is possible to the memory body demand of image capture unit is greatly reduced, and then reduce its manufacturing cost.The present invention can reduce the kind of the eigenvalue cluster in characteristic image, also therefore is able to significantly save the calculation resources of image capture unit, improves its efficiency.The present invention is that the eigenvalue cluster with diversity minimum is to replace each eigenvalue cluster being deleted rather than directly to delete, thus without the accuracy reducing object detecting.Each eigenvalue cluster in eigenvalue group after the method for the present invention simplifies all can find the eigenvalue cluster corresponding with after its transposition or after mirror, therefore can retain the characteristic of former eigenvalue group, to perform relevant application.Therefore, the shortcoming that the present invention can improve known technology really.
The foregoing is only illustrative, rather than be restricted person.Any spirit and scope without departing from the present invention, and the equivalent modifications carrying out it or change, be intended to be limited solely by appended claims.
Claims (10)
1. the image capture unit that can simplify image feature value group, it is characterised in that it comprises:
One Feature Conversion module, is a characteristic image by a video conversion to be measured, and described characteristic image comprises an eigenvalue group；
One feature reduction module, according to a look-up table, perform a simplification program, from described eigenvalue group, the eigenvalue cluster that would not exist in described look-up table is deleted, each eigenvalue cluster being deleted is then pre-conditioned and be present in the eigenvalue cluster of described look-up table and replace to meet one, makes described eigenvalue the group constant but kind of sum reduce, produces the eigenvalue group after a simplification；Wherein, described pre-conditioned being determined by diversity, each eigenvalue cluster being deleted described need to replace with the eigenvalue cluster minimum with its diversity；
One template stores module, stores several determinand templates；And
One identification module, compares the eigenvalue group after described simplification with several determinand templates described, with the determinand in image to be measured described in identification.
The image capture unit simplifying image feature value group the most according to claim 1, it is characterised in that described look-up table is set up less than the eigenvalue cluster of a complexity threshold value with the complexity of eigenvalue cluster.
The image capture unit simplifying image feature value group the most according to claim 2, it is characterised in that after performing described simplification program, in the eigenvalue group after described simplification, each eigenvalue cluster all can find the eigenvalue cluster corresponding with after its transposition.
The image capture unit simplifying image feature value group the most according to claim 2, it is characterised in that after performing described simplification program, in the eigenvalue group after described simplification, each eigenvalue cluster all can find the eigenvalue cluster corresponding with after its mirror.
The image capture unit simplifying image feature value group the most according to claim 1, it is characterized in that, several determinand templates described utilize and are present in the eigenvalue cluster of described look-up table, and via advancing algorithm, support vector machine algorithm or Principal Component Analysis Method+linearly have identification force analysis calculation to be set up.
6. the control method that can simplify image feature value group, it is characterised in that it comprises the steps of
Utilizing a Feature Conversion module is a characteristic image by a video conversion to be measured, and described characteristic image comprises an eigenvalue group；
By a feature reduction module according to a look-up table, perform a simplification program, from described eigenvalue group, the eigenvalue cluster that would not exist in described look-up table is deleted, each be deleted eigenvalue cluster then with meet one pre-conditioned and be present in described look-up table eigenvalue cluster replace, make described eigenvalue group sum constant but kind minimizing, produce the eigenvalue group after a simplification；Wherein, described pre-conditioned being determined by diversity, each value indicative group being deleted need to replace with the eigenvalue cluster minimum with its diversity；
Stored module by a template and store several determinand templates；And
By an identification module, the eigenvalue group after described simplification is compared with several determinand templates described, with the determinand in image to be measured described in identification.
The control method simplifying image feature value group the most according to claim 6, it is characterised in that described look-up table is set up less than the eigenvalue cluster of a complexity threshold value with the complexity of eigenvalue cluster.
The control method simplifying image feature value group the most according to claim 7, it is characterised in that after performing described simplification program, in the eigenvalue group after described simplification, each eigenvalue cluster all can find the eigenvalue cluster corresponding with after its transposition.
The control method simplifying image feature value group the most according to claim 7, it is characterised in that after performing described simplification program, in the eigenvalue group after described simplification, each eigenvalue cluster all can find the eigenvalue cluster corresponding with after its mirror.
The control method simplifying image feature value group the most according to claim 6, it is characterized in that, several determinand templates described are present in the eigenvalue cluster of described look-up table, and via advancing algorithm, support vector machine algorithm or Principal Component Analysis Method+linearly have identification force analysis calculation to be set up.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110454290.3A CN103188427B (en) | 2011-12-30 | 2011-12-30 | Image capture unit and the control method thereof of image feature value group can be simplified |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110454290.3A CN103188427B (en) | 2011-12-30 | 2011-12-30 | Image capture unit and the control method thereof of image feature value group can be simplified |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103188427A CN103188427A (en) | 2013-07-03 |
| CN103188427B true CN103188427B (en) | 2016-08-10 |
Family
ID=48679379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110454290.3A Active CN103188427B (en) | 2011-12-30 | 2011-12-30 | Image capture unit and the control method thereof of image feature value group can be simplified |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103188427B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015051697A1 (en) * | 2013-10-10 | 2015-04-16 | Harman International Industries, Incorporated | Image processing |
| US9460338B2 (en) * | 2014-12-10 | 2016-10-04 | Intel Corporation | Face detection method and apparatus with lookup table |
| TWI681342B (en) * | 2018-12-11 | 2020-01-01 | 瑞昱半導體股份有限公司 | Object position determination circuit |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3279913B2 (en) * | 1996-03-18 | 2002-04-30 | 株式会社東芝 | Person authentication device, feature point extraction device, and feature point extraction method |
| TWI226589B (en) * | 2003-04-28 | 2005-01-11 | Ind Tech Res Inst | Statistical facial feature extraction method |
| CN1963918A (en) * | 2005-11-11 | 2007-05-16 | 株式会社东芝 | Compress of speaker cyclostyle, combination apparatus and method and authentication of speaker |
| CN101615248B (en) * | 2009-04-21 | 2012-03-21 | 华为技术有限公司 | Age estimation method, equipment and face recognition system |
| CN102110225B (en) * | 2009-12-28 | 2013-10-30 | 比亚迪股份有限公司 | Outdoor face identifying method and system |
-
2011
- 2011-12-30 CN CN201110454290.3A patent/CN103188427B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN103188427A (en) | 2013-07-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11061479B2 (en) | Method, device and readable storage medium for processing control instruction based on gesture recognition | |
| Zhang et al. | Face recognition using Laplacianfaces | |
| US9251402B2 (en) | Association and prediction in facial recognition | |
| Li et al. | Delving into egocentric actions | |
| CN106909888B (en) | Face key point tracking system and method applied to mobile equipment terminal | |
| CN105426829B (en) | Video classification methods and device based on facial image | |
| KR102255215B1 (en) | Method of detecting object in image and image processing device | |
| CN111160269A (en) | Face key point detection method and device | |
| CN108171133B (en) | Dynamic gesture recognition method based on characteristic covariance matrix | |
| US8421877B2 (en) | Feature data compression device, multi-directional face detection system and detection method thereof | |
| KR20160061081A (en) | Method for face recognition, recording medium and device for performing the method | |
| Calonder et al. | Compact signatures for high-speed interest point description and matching | |
| CN103188427B (en) | Image capture unit and the control method thereof of image feature value group can be simplified | |
| Qian et al. | Robust low-rank regularized regression for face recognition with occlusion | |
| Lai et al. | Proposed FPGA hardware architecture for high frame rate (≫ 100 fps) face detection using feature cascade classifiers | |
| CN105528616A (en) | Face recognition method and device | |
| KR102389457B1 (en) | Image Transformation Apparatus, Method and Computer Readable Recording Medium Thereof | |
| Ullah et al. | Gender recognition from face images with dyadic wavelet transform and local binary pattern | |
| Li et al. | Sample selection for visual domain adaptation via sparse coding | |
| CN105512664A (en) | Image recognition method and device | |
| TWI459310B (en) | Image capturing device able to simplify characteristic value sets of captured images and control method thereof | |
| Huang et al. | Regularized trace ratio discriminant analysis with patch distribution feature for human gait recognition | |
| Bruneau et al. | Aggregation of probabilistic PCA mixtures with a variational-bayes technique over parameters | |
| Megahed et al. | Face2face manipulation detection based on histogram of oriented gradients | |
| CN102194107B (en) | Smiling face recognition method for reducing dimension by using improved linear discriminant analysis |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20161222 Address after: China Taiwan Hsinchu Science Park Hsinchu City Road No. 12 5 floor Patentee after: Glomerocryst semiconductor limited company Address before: Hsinchu City, Taiwan, China Patentee before: Huajing Science & Technology Co., Ltd. |