CN105844605B - Based on the human face portrait synthetic method adaptively indicated - Google Patents
Based on the human face portrait synthetic method adaptively indicated Download PDFInfo
- Publication number
- CN105844605B CN105844605B CN201610152915.3A CN201610152915A CN105844605B CN 105844605 B CN105844605 B CN 105844605B CN 201610152915 A CN201610152915 A CN 201610152915A CN 105844605 B CN105844605 B CN 105844605B
- Authority
- CN
- China
- Prior art keywords
- block
- photo
- portrait
- test
- training
- 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
- 238000010189 synthetic method Methods 0.000 title claims abstract description 16
- 238000012360 testing method Methods 0.000 claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000000034 method Methods 0.000 claims abstract description 33
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 26
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 26
- 238000001914 filtration Methods 0.000 claims abstract description 20
- 239000008186 active pharmaceutical agent Substances 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 5
- 238000003708 edge detection Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 2
- 239000002131 composite material Substances 0.000 abstract description 7
- 238000007796 conventional method Methods 0.000 abstract description 2
- 239000000284 extract Substances 0.000 abstract 1
- 235000013399 edible fruits Nutrition 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 238000011840 criminal investigation Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000007500 overflow downdraw method Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000003760 hair shine Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20224—Image subtraction
-
- 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
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Processing Or Creating Images (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of based on the human face portrait synthetic method adaptively indicated, mainly solves the problems, such as that existing method synthesis portrait clarity is low and details is incomplete.Implementation step is：Database is handled first, after all images are carried out image filtering, to image block and extracts image block characteristics, obtains a training portrait block dictionary and two photo block dictionaries；Secondly according to whether including that marginal information or face feature information select different dictionaries in test photo block, neighbour's block is found；Portrait block to be synthesized finally is obtained using markov network model, and all portrait blocks to be synthesized are merged to obtain synthesis portrait.Compared with the conventional method, composite result has higher clarity, details more complete to the present invention, can be used for face retrieval and identification.
Description
Technical field
The present invention is to belong to the technical field of image procossing, further relates to pattern-recognition and is led with computer vision technique
Human face portrait synthetic method in domain, can be used for the face retrieval in criminal investigation and case detection and identification.
Background technology
With modern society and multimedia development, the more and more video images of people are recorded, how root
Differentiate according to existing image and certification a person's identity, it has also become one of problem to be solved, wherein recognition of face have
Have direct, friendly and the features such as facilitate, has obtained extensive research and application.The one of important application of face recognition technology is just
It is to assist police's criminal investigation and case detection.But the photo of the suspect of some major cases is very unobtainable in many cases, or
To photo be not that positive or uneven illumination is even, then the police can draw out suspicion according to the description of live eye witness
The portrait of suspect is retrieved and is identified in the picture data library of the police later.Since human face photo and portrait are in imager
All there is larger difference in terms of system, shape and texture, with human face portrait directly using existing face identification method identification effect
Fruit is poor.In view of the above-mentioned problems, there are two types of methods altogether：A solution is by the photo conversion in police's face database
It draws a portrait at synthesis, portrait to be identified is identified in synthesis representation data library later；Another scheme is by picture to be identified
As being converted to photomontage, it is identified in the picture data library of the police later.Human face portrait synthesizes usual base at present
In three kinds of methods：First, the human face portrait synthetic method based on local linear；Second, the people based on markov network model
Face portrait synthetic method；Third, the human face portrait synthetic method based on rarefaction representation.
Liu et al. people is in document " Q.S.Liu and X.O.Tang, A nonlinear approach for face
sketch synthesis and recognition,in Proc.IEEE Int.Conference on Computer
It is proposed in Vision, San Diego, CA, pp.1005-1010,20-26Jun.2005. " a kind of next close by local linear
Photo is converted to synthesis portrait like global non-linear method.This method embodiment is：First by the photo-in training set
Portrait pair and photo to be transformed are divided into same size and the image block of identical overlapping region, for each of photo to be transformed
Photo block finds its K neighbour's photo block in training photo block, is then weighted the corresponding portrait block of K photo block
Combination obtains portrait block to be synthesized, finally merges all portrait blocks to be synthesized to obtain synthesis portrait.But this method exists
Shortcoming be：Since neighbour's number is fixed, lead to composite result there are clarity the defect low, details is fuzzy.
Wang et al. is in document " X.Wang, and X.Tang, " Face Photo-Sketch Synthesis and
Recognition,”IEEE Transactions on Pattern Analysis and Machine Intelligence,
31 (11) propose a kind of human face portrait synthetic method based on markov network model in 1955-1967,2009 ".The party
Method embodiment is：First by the sketch-photo pair and test photo piecemeal in training set, then according to test photo block and instruction
Practice the relationship between the portrait block of the relationship and adjacent position between photo block, markov network model is established, to each
It tests photo block and finds a best training portrait block as portrait block to be synthesized, finally melt all portrait blocks to be synthesized
Conjunction obtains synthesis portrait.But shortcoming existing for this method is：Since each photo block position only selects a trained picture
As block carries out portrait synthesis, composite result is caused to there are problems that blocking artifact and details missing.
Patented technology " the sketch-photo generation method based on rarefaction representation " (application number of high-new wave et al. application：
201010289330.9 the applying date：2010-09-24 application publication numbers：101958000 A of CN) in disclose it is a kind of based on sparse
The human face portrait synthetic method of expression.This method embodiment is：Using having, method generates synthesis portrait or synthesis is shone first
Then the initial estimation of piece synthesizes detailed information using the method for rarefaction representation, finally by initial estimation and detailed information into
Row fusion.But shortcoming existing for this method is：The relationship between the image block of adjacent position is had ignored, synthesis is caused to be tied
There is fuzzy and blocking artifact in fruit.
Invention content
It is an object of the invention to overcome the shortcomings of above-mentioned existing method, propose a kind of based on the face adaptively indicated picture
As synthetic method, to improve the picture quality of synthesis portrait.
Realize that the technical solution of the object of the invention includes：
1. it is a kind of based on the human face portrait synthetic method adaptively indicated, include the following steps：
(1) sketch-photo is divided into trained library and test library to database, and chooses a test from test library and shines
Piece PTe；
(2) photo in training library is subjected to difference of Gaussian filtering, and the photo in library and corresponding filtering will be trained to scheme
As being divided into, size is identical and identical piece of overlapping degree；PTr={ PTr,1,PTr,2,…,PTr,i,…,PTr,N, 1≤i≤N, N are
The total number of block；
(3) it uses training photo block and corresponding filtering image block as two features, obtains the first training photo block
Dictionary Dp1, and training photo block and corresponding filtering image block are extracted respectively and accelerate robust features and local binary patterns special
Sign uses this four features as second of training photo block dictionary Dp2；
(4) portrait in training library is divided into that size is identical and identical piece of S of overlapping degreeTr={ STr,1,STr,2,…,
STr,i,…,STr,N, obtain training portrait block dictionary DS；
(5) test photo is subjected to edge detection and characteristic point detection, obtains the marginal information and characteristic point of test photo
Information；
(6) photo in test library is subjected to difference of Gaussian filtering, and test photo and filtered photo is divided into
Size is identical and identical piece of P of overlapping degreeTe={ PTe,1,PTe,2,…,PTe,i,…,PTe,N, and judge each test photo block
PTe,iWhether marginal information or characteristic point information are had：
If this photo block includes marginal information or characteristic point information, to testing photo block PTe,iExtraction accelerates robust
Feature and local binary patterns feature, according to characteristic distance from training photo block dictionary Dp2Middle searching K similar photo blocks are made
For photo block to be selectedSimultaneously from training portrait block dictionary DSIn
Select portrait block corresponding with neighbour's photo block as wait for selection portrait block
If this photo block does not include marginal information or characteristic point information, by this test photo block PTe,iAnd it is right
The filtering image block answered is as feature, according to characteristic distance from training photo block dictionary Dp1Middle searching K similar photo blocks are made
For photo block to be selectedSimultaneously from training portrait block dictionary DSMiddle selection and neighbour
The corresponding portrait block of photo block, which is used as, waits for selection portrait block
(7) using the image block characteristics of extraction, markov network model is solved by the method for alternating iteration, is obtained every
A test photo block PTe,iMultiple features between weights μi={ μi,1,μi,2,…,μi,l,…,μi,L, 1≤l≤L, L are spy
The total number of sign, while obtaining the corresponding photo block to be selected of each test photo blockWeights ωi={ ωi,1,
ωi,2,…,ωi,j,…,ωi,K}；
(8) selection portrait block is waited for using what step (6) obtainedAnd step
(7) the weights ω obtainedi={ ωi,1,ωi,2,…,ωi,j,…,ωi,K, each test photo block P is obtained according to the following formulaTe,iIt is right
The portrait block S to be synthesized answeredi：
(9) iteration executes step (7)-(8) until obtaining the final portrait block to be synthesized of N blocks, and by these pictures to be synthesized
As block is combined to obtain synthesis portrait corresponding with test photo.
The present invention has the following advantages that compared with the conventional method：
First, the present invention considers the relationship between the image block of adjacent position, while selecting K neighbour each piece of position
Block is rebuild so that composite result is more clear；
Second, present invention employs the method adaptively indicated, different regions is synthesized using different features, is used
Different features weighs the distance between two image blocks relationship, improves the quality of composite result and keeps details more complete.
Description of the drawings
Fig. 1 is the implementation flow chart of the present invention；
Fig. 2 is the comparing result of the synthesis portrait with existing four kinds of methods on CUHK student databases with invention
Figure.
Specific implementation mode
Core of the invention thought is：A kind of human face portrait synthetic method is proposed by the thought adaptively indicated, is made not
Same human face region is synthesized with different features, improves the picture quality of composite result.
Referring to Fig.1, implementation steps of the invention are as follows：
Step 1, test photo P is chosenTe。
From sketch-photo N to marking off trained library and test library in database, and chooses a photo from test library and make
To test photo PTe。
Step 2, two trained photo block dictionary D are obtainedp1And Dp2。
The photo in training library 2a) is subjected to difference of Gaussian filtering：
2a1) construct the Gaussian function of two different scale value σ：
Wherein, G (x, y, σ) indicates the Gaussian function under σ scale-values, and x, y indicate that pixel is corresponding in photo respectively
Horizontal, ordinate value；
Photo 2a2) is subjected to convolution with the Gaussian function of two different scales respectively, obtains the photo after two convolution；
2a3) photo after two convolution is subtracted each other, obtained image is exactly that photo passes through the filtered knot of difference of Gaussian
Fruit；
The photo trained in library and corresponding filtering image 2b) are divided into the block of same size and identical overlapping degree,
It uses training photo block and corresponding filtering image block as two features, obtains the first training photo block dictionary Dp1；
2c) training photo block and corresponding filtering image block are extracted respectively and accelerate robust features and local binary patterns
Feature uses this four features as second of training photo block dictionary Dp2；
The extracting method for accelerating robust features feature and local binary patterns feature, difference bibliography " H.Bay,
A.Ess,T.Tuytelaars,L.Gool.SURF:Speeded Up Robust Features.Computer Vision and
Image Understanding,110(3):346-359,2008 " and " T.Ojala, M.T.
Multiresolution Gray-Scale and Rotation Invariant Texture Classification with
Local Binary Patterns.IEEE Transactions on Pattern Analysis and Machine
Intelligence,24(7):971-987,2002”。
Step 3, training portrait block dictionary D is obtainedS。
Portrait in training library is divided into the block of same size and identical overlapping degree, obtains training portrait block dictionary
DS。
Step 4, the information of test photo is obtained.
Test photo is subjected to edge detection and characteristic point detection, obtains the marginal information and characteristic point letter of test photo
Breath；
Step 5, photo block to be selected is obtainedWith wait for selection portrait block
Photo in test library is subjected to difference of Gaussian filtering, and test photo and filtered photo are divided into size
Identical piece of P of identical and overlapping degreeTe={ PTe,1,PTe,2,…,PTe,i,…,PTe,N, and judge each test photo block PTe,i
Whether marginal information or characteristic point information are had：
If this photo block includes marginal information or characteristic point information, to testing photo block PTe,iExtraction accelerates robust
Feature and local binary patterns feature, according to characteristic distance from training photo block dictionary Dp2Middle searching K similar photo blocks are made
For photo block to be selectedSimultaneously from training portrait block dictionary DSIn
Select portrait block corresponding with neighbour's photo block as wait for selection portrait block
If this photo block does not include marginal information or characteristic point information, by this test photo block PTe,iAnd it is right
The filtering image block answered is as feature, according to characteristic distance from training photo block dictionary Dp1Middle searching K similar photo blocks are made
For photo block to be selectedSimultaneously from training portrait block dictionary DSMiddle selection with it is close
The corresponding portrait block of adjacent photo block is as waiting for selection portrait block
Step 6, the weights μ between multiple features is solvediWith the weights ω of photo block to be selectedi。
Using the image block characteristics of extraction, markov network model is solved by the method for alternating iteration, is obtained each
Test the weights μ between multiple features of photo blocki={ μi,1,μi,2,…,μi,l,…,μi,L, 1≤l≤L, L are characterized total
Number, while obtaining the corresponding photo block to be selected of each test photo blockWeights ωi={ ωi,1,ωi,2,…,
ωi,j,…,ωi,K}。
Described is as follows by the method solution procedure of alternating iteration：
6a) each test the weights between the multiple features of the equal random initializtion of photo blockAt the beginning of photo block to be selected
Initial value is ωi；
6b) according to Euclidean distance formula calculate photo block to be selected and test photo block multiple features between it is European away from
From, obtain test photo block and photo block to be selected between relationship：
Wherein, d represents the distance between two features, x1And x2Respectively represent the abscissa of two feature vectors, y1And y2
Respectively represent the ordinate of two feature vectors；
6c) according to 6b) described in Euclidean distance formula, calculate adjacent position wait for selection portrait block pixel value between
Euclidean distance, to obtain the relationship of adjacent position waited between selection portrait block；
6d) by 6b) and result 6c) be brought into Markov model；
Markov model 6e) is utilized, according to the initialization weights between multiple featuresTreat the first of selection photo block
Beginningization weightsIt optimizes, the weights ω of the photo block to be selected after being optimizedi；
6f) according to Markov model and 6e) optimization come photo block to be selected weights ωi, optimize multiple spies
Initialization weights between signThe weights μ between multiple features after being optimizedi；
6g) iteration executes 6b) to 6f), until the weights ω of the corresponding photo block to be selected of each test photo blockiNo longer
Change or reach preset iterations, obtains the weights μ between multiple features of each test photo blockiWith wait for selection shine
The weights ω of tilei。
Step 7, portrait block S to be synthesized is solvedi。
Selection portrait block is waited for using what step 5 obtainedThe weights ω obtained with step 6i, each survey is obtained according to the following formula
Try the corresponding portrait block S to be synthesized of photo blocki：
Step 8, iteration executes to obtain final synthesis portrait.
Iteration executes step 6 to step 7 until obtaining all portrait blocks to be synthesized, and these pictures to be synthesized that will be obtained
As block is merged to obtain synthesis portrait corresponding with test photo.
The effect of the present invention can be described further by following emulation experiment.
1. simulated conditions
It is Inter (R) Core (TM) i5-3470 3.20GHz, memory 16G, WINDOWS that the present invention, which is in central processing unit,
In 7 operating systems, the MATLAB 2012b developed with Mathworks companies of the U.S. are emulated.Database uses in Hong Kong
University of liberal arts CUHK student databases.
2. emulation content
Experiment 1：Synthesis of the photo to portrait
With the method for the present invention and the existing method LLE based on local linear, the method MRF based on Markov random field,
Method MWF based on markov weight field and it is based on multiple features fusion method MrFSPS, in the CUHK of Hong Kong Chinese University
On student databases carry out photo to draw a portrait synthesis, experimental result such as Fig. 2, wherein：
Fig. 2 (a) is original photo；
Fig. 2 (b) is the portrait of the method LLE synthesis based on local linear；
Fig. 2 (c) is the portrait of the method MRF synthesis based on Markov random field；
Fig. 2 (d) is the portrait of the method MWF synthesis based on markov weight field；
Fig. 2 (e) is the portrait based on the MrFSPS synthesis of multiple features fusion method；
Fig. 2 (f) is the portrait of the method for the present invention synthesis.
By testing 1 result as it can be seen that since the present invention is by means of the thought adaptively indicated, the difference of facial image can be made
Region uses different character representations, can preferably weigh the distance between two image blocks relationship so that composite result is excellent
In other human face portrait synthetic methods, the advance of the present invention is demonstrated.
Claims (3)
1. it is a kind of based on the human face portrait synthetic method adaptively indicated, include the following steps：
(1) sketch-photo is divided into trained library and test library to database, and chooses a test photo from test library；
(2) photo in training library is subjected to difference of Gaussian filtering, and the photo in training library is drawn with corresponding filtering image
It is divided into that size is identical and identical piece of overlapping degree；PTr={ PTr,1,PTr,2,...,PTr,i,...,PTr,N, 1≤i≤N, N are block
Total number；
(3) it uses training photo block and corresponding filtering image block as two features, obtains the first training photo block dictionary
Dp1, and training photo block and corresponding filtering image block are extracted respectively and accelerate robust features and local binary patterns feature,
Use this four features as second of training photo block dictionary Dp2；
(4) portrait in training library is divided into that size is identical and identical piece of S of overlapping degreeTr={ STr,1,STr,2,...,
STr,i,...,STr,N, obtain training portrait block dictionary DS；
(5) test photo is subjected to edge detection and characteristic point detection, obtains the marginal information and characteristic point information of test photo；
(6) photo in test library is subjected to difference of Gaussian filtering, and test photo and filtered photo is divided into size
Identical piece of P of identical and overlapping degreeTe={ PTe,1,PTe,2,...,PTe,i,...,PTe,N, and judge each test photo block
PTe,iWhether marginal information or characteristic point information are had：
If this photo block includes marginal information or characteristic point information, to testing photo block PTe,iExtraction accelerates robust features
With local binary patterns feature, according to characteristic distance from training photo block dictionary Dp2Middle searching K similar photo blocks are used as and wait for
Select photo blockSimultaneously from training portrait block dictionary DSMiddle selection
Portrait block corresponding with neighbour's photo block, which is used as, waits for selection portrait block
If this photo block does not include marginal information or characteristic point information, by this test photo block PTe,iAnd it is corresponding
Filtering image block is as feature, according to characteristic distance from training photo block dictionary Dp1Middle searching K similar photo blocks are used as and wait for
Select photo blockSimultaneously from training portrait block dictionary DSMiddle selection is shone with neighbour
The corresponding portrait block of tile, which is used as, waits for selection portrait block
(7) using the image block characteristics of extraction, markov network model is solved by the method for alternating iteration, obtains each survey
Try photo block PTe,iMultiple features between weights μi={ μi,1,μi,2,...,μi,l,...,μi,L, 1≤l≤L, L are characterized
Total number, while obtaining the corresponding photo block to be selected of each test photo blockWeights ωi={ ωi,1,
ωi,2,...,ωi,j,...,ωi,K}；
(8) selection portrait block is waited for using what step (6) obtainedIt is obtained with step (7)
The weights ω arrivedi={ ωi,1,ωi,2,...,ωi,j,...,ωi,K, each test photo block P is obtained according to the following formulaTe,iIt is corresponding
Portrait block S to be synthesizedi：
(9) iteration executes step (7)-(8) until obtaining the final portrait block to be synthesized of N blocks, and by these portrait blocks to be synthesized
It is combined to obtain synthesis portrait corresponding with test photo.
2. according to claim 1 based on the human face portrait synthetic method adaptively indicated, which is characterized in that step (2)
In to training library in photo carry out difference of Gaussian filtering, carry out as follows：
(2a) constructs the Gaussian function of two different scale value σ：
Wherein, G (x, y, σ) indicates that the Gaussian function under σ scale-values, x, y indicate that pixel is corresponding horizontal, vertical in photo respectively
Coordinate value；
Photo is carried out convolution by (2b) with the Gaussian function of two different scales respectively, obtains the photo after two convolution；
(2c) subtracts each other the photo after two convolution, and obtained image is exactly that photo passes through the filtered result of difference of Gaussian.
3. according to claim 1 based on the human face portrait synthetic method adaptively indicated, which is characterized in that step (7)
In markov network model solved by the method for alternating iteration, steps are as follows：
(3a) each tests photo block PTe,iWeights between the multiple features of equal random initializtionAt the beginning of photo block to be selected
Initial value is
(3b) according to photo block select with test photo block the distance between multiple features, calculate test photo block with it is to be selected
Select the relationship between photo block；
(3c) according to the distance between the pixel value for waiting for selection portrait block of adjacent position, calculate adjacent position waits for selection portrait
Relationship between block；
The result of (3b) and (3c) are brought into Markov model by (3d)；
(3e) utilizes Markov model, according to the initialization weights between multiple featuresTreat the initialization of selection photo block
WeightsIt optimizes, the weights ω of the photo block to be selected after being optimizedi；
(3f) according to Markov model and (3e) optimization come photo block to be selected weights ωi, optimize multiple features it
Between initialization weightsThe weights μ between multiple features after being optimizedi；
(3g) iteration executes (3b) to (3f), until the weights of the corresponding photo block to be selected of each test photo block no longer change
Or reach preset iterations, obtain the weights μ between multiple features of each test photo blockiWith photo block to be selected
Weights ωi。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610152915.3A CN105844605B (en) | 2016-03-17 | 2016-03-17 | Based on the human face portrait synthetic method adaptively indicated |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610152915.3A CN105844605B (en) | 2016-03-17 | 2016-03-17 | Based on the human face portrait synthetic method adaptively indicated |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105844605A CN105844605A (en) | 2016-08-10 |
| CN105844605B true CN105844605B (en) | 2018-08-10 |
Family
ID=56588264
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610152915.3A Active CN105844605B (en) | 2016-03-17 | 2016-03-17 | Based on the human face portrait synthetic method adaptively indicated |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105844605B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108154133B (en) * | 2018-01-10 | 2020-04-14 | 西安电子科技大学 | Face portrait-photo recognition method based on asymmetric joint learning |
| CN108932536B (en) * | 2018-07-18 | 2021-11-09 | 电子科技大学 | Face posture reconstruction method based on deep neural network |
| CN110069992B (en) * | 2019-03-18 | 2021-02-09 | 西安电子科技大学 | Face image synthesis method and device, electronic equipment and storage medium |
| CN111179178B (en) * | 2019-12-31 | 2023-06-13 | 深圳云天励飞技术有限公司 | Face image stitching method and related product |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279936A (en) * | 2013-06-21 | 2013-09-04 | 重庆大学 | Human face fake photo automatic combining and modifying method based on portrayal |
| CN104517274A (en) * | 2014-12-25 | 2015-04-15 | 西安电子科技大学 | Face portrait synthesis method based on greedy search |
-
2016
- 2016-03-17 CN CN201610152915.3A patent/CN105844605B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279936A (en) * | 2013-06-21 | 2013-09-04 | 重庆大学 | Human face fake photo automatic combining and modifying method based on portrayal |
| CN104517274A (en) * | 2014-12-25 | 2015-04-15 | 西安电子科技大学 | Face portrait synthesis method based on greedy search |
Non-Patent Citations (5)
| Title |
|---|
| Face Sketch–Photo Synthesis and Retrieval Using Sparse Representation;Xinbo Gao等;《IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY》;20120831;第22卷(第8期);第1213-1226页 * |
| Face Sketch-photo Synthesis under Multi-dictionary Sparse Representation Framework;Nannan Wang等;《2011 Sixth International Conference on Image and Graphics》;20111231;第82-87页 * |
| Robust Face Sketch Style Synthesis;Shengchuan Zhang等;《IEEE TRANSACTIONS ON IMAGE PROCESSING》;20160131;第25卷(第1期);第220-232页 * |
| 基于人脸画像的伪照片合成及修正;李伟红;《光学精密工程》;20140531;第２２卷(第５期);第1371-1378页 * |
| 基于局部约束邻域嵌入的人脸画像-照片合成;胡彦婷;《计算机应用》;20151231;第35卷(第2期);第535-539页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105844605A (en) | 2016-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Walia et al. | Digital image forgery detection: a systematic scrutiny | |
| Chen et al. | Facial expression recognition in video with multiple feature fusion | |
| Cho et al. | A probabilistic image jigsaw puzzle solver | |
| Li et al. | Linestofacephoto: Face photo generation from lines with conditional self-attention generative adversarial networks | |
| Güçlütürk et al. | Convolutional sketch inversion | |
| CN105844605B (en) | Based on the human face portrait synthetic method adaptively indicated | |
| CN103544504B (en) | Scene character recognition method based on multi-scale map matching core | |
| CN104794479B (en) | This Chinese detection method of natural scene picture based on the transformation of local stroke width | |
| Wang et al. | A benchmark for clothes variation in person re‐identification | |
| CN108154133B (en) | Face portrait-photo recognition method based on asymmetric joint learning | |
| CN106980825B (en) | Human face posture classification method based on normalized pixel difference features | |
| CN103984954B (en) | Image combining method based on multi-feature fusion | |
| CN112017162B (en) | Pathological image processing method, pathological image processing device, storage medium and processor | |
| WO2023165616A1 (en) | Method and system for detecting concealed backdoor of image model, storage medium, and terminal | |
| Gangan et al. | Distinguishing natural and computer generated images using Multi-Colorspace fused EfficientNet | |
| CN115240280A (en) | Construction method of human face living body detection classification model, detection classification method and device | |
| das Neves et al. | HU‐PageScan: a fully convolutional neural network for document page crop | |
| Shivakumara et al. | A new RGB based fusion for forged IMEI number detection in mobile images | |
| Zhu et al. | BDGAN: Image blind denoising using generative adversarial networks | |
| CN105869134B (en) | Human face portrait synthetic method based on direction graph model | |
| Satwashil et al. | Integrated natural scene text localization and recognition | |
| CN102110303B (en) | Method for compounding face fake portrait\fake photo based on support vector return | |
| Sharrma et al. | Vision based static hand gesture recognition techniques | |
| CN109685076A (en) | A kind of image-recognizing method based on SIFT and sparse coding | |
| CN106023120B (en) | Human face portrait synthetic method based on coupling neighbour's index |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220712 Address after: 518057 2304, block a, building 2, Shenzhen International Innovation Valley, Dashi 1st Road, Xili community, Xili street, Nanshan District, Shenzhen, Guangdong Province Patentee after: SHENZHEN AIMO TECHNOLOGY Co.,Ltd. Address before: 710071 No. 2 Taibai South Road, Shaanxi, Xi'an Patentee before: XIDIAN University |