CN101404091A - Three-dimensional human face reconstruction method and system based on two-step shape modeling - Google Patents
Three-dimensional human face reconstruction method and system based on two-step shape modeling Download PDFInfo
- Publication number
- CN101404091A CN101404091A CNA2008102330087A CN200810233008A CN101404091A CN 101404091 A CN101404091 A CN 101404091A CN A2008102330087 A CNA2008102330087 A CN A2008102330087A CN 200810233008 A CN200810233008 A CN 200810233008A CN 101404091 A CN101404091 A CN 101404091A
- Authority
- CN
- China
- Prior art keywords
- dimensional
- face
- point
- human face
- texture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Processing Or Creating Images (AREA)
- Image Generation (AREA)
Abstract
The invention discloses a method for carrying out the three-dimensional human face modeling by a single human face photo to obtain a realistic human face, and the method belongs to the field of graphics and images. The method utilizes a priori knowledge of a human face structure to estimate the depths of two-dimensional feature points on the photo, thereby obtaining similar three-dimensional coordinates; the similar three-dimensional feature points are taken as control points, and the dirichlet free deformation algorithm is adopted to realize the fitting of the neutral human face to the specific human face. Then, a method which is based on the regional expansion is adopted to realize the texture synthesis and the mapping, and the region which is not covered is repaired by adopting the interpolation or the block filling method. The method has simple calculation, high modeling accuracy and can rapidly utilize a small number of feature points to realize the three-dimensional modeling of the human face on the single photo, and the generated three-dimensional human face model has very strong realistic feeling under various postures due to the accurate estimation of the depths of the feature points. The method can also be used for the three-dimensional reconstruction of other objects.
Description
Technical field
The present invention relates to fields such as computer graphics, Digital Image Processing and artificial intelligence, specifically is a kind of method and system that carries out the three-dimensional face modeling by plane picture.
Background technology
Existing three-dimensional face modeling technique can roughly be divided into based on hardware (spatial digitizer, structured light etc.) with based on two kinds of methods of image modeling.Utilize the 3-dimensional digital scanner to obtain people's face three-dimensional information and can access accurate faceform, but because hardware device cost height, dumb generally only is applicable to some special occasions.Therefore, the method for carrying out the modeling of people's face individual character according to the information in image or the video has huge market outlook.
By to the retrieval of technical literature in recent years, can find that more representative in the human face model building based on image is according to many viewpoints facial image or video sequence, calculate and recover the three-dimensional information of human face characteristic point by the machine vision algorithm, and synthesize specific three dimensional people face by distortion to neutral faceform.Space scattered data being interpolation method (the Proceeding of ACM SIGGRAPH that on american computer association computer graphical special interest group 98 nd Annual Meeting collection, proposes comprising people such as Pighin, 1998:75-84) and people such as Liu at " The Journal of Visualization and Computer Animation ", Vol.12, No.4,2001:227-240 (visual and computer animation, the 12nd volume, the 4th phase, calendar year 2001: what the 227-240 page or leaf) go up to propose passes through corners Matching and Structure from Motion (restructure from motion structure) algorithm, and the latter has realized the human face rebuilding based on single camera; In order to improve modeling accuracy, people such as Fua propose to recover people's face shape with the method that flux of light method (Bundle Adjustment) combines with face database from video flowing, and this method can adapt to the variation at illumination condition.But generally speaking, this class needs a large amount of man-machine interactions to finish Feature Points Matching between image based on the modeling method of plurality of pictures, complicated operation, and be difficult for realizing robotization.
In order to simplify manually-operated and to realize the robotization of modeling, some people's face modeling algorithms based on single image have appearred again in recent years.Generally speaking, under the situation without any the hypothesis constraint, rebuilding based on the 3D shape of single image is an ill-conditioning problem.But people such as scholar Blanz are at " IEEE Transactions onPattern Analysis and Machine Intelligence " Vol.25, No.9,2003:1063-1074 (IEEE pattern-recognition and machine intelligence, the 25th volume, the 9th phase, 2003: the 1063-1074 page or leaf) go up the three-dimensional deformation model that proposes and overcome this problem, they retrained the faceform as priori with the three-dimensional face storehouse, have successfully realized the three-dimensional face automatic modeling based on single image.But also there are many improved places of demanding urgently in this model: iterations is many, operation time length be that it is mainly not enough, in addition, the Model Matching algorithm is strong to the initial value dependence, tends to cause modeling to be failed because of locally optimal solution.On this basis, Romdhani adopted multiple characteristics that the optimization aim function is retrained in paper " Face Image Analysis Using a Multiple FeatureFitting Strategy " (based on the facial image analysis of multiple characteristics match scheme) in 2005, alleviated the local optimum problem of deformation model to a certain extent, but also caused calculating complicated more simultaneously.Generally speaking, finding the solution the low performance that large-scale nonlinear function optimization problem caused is the basic difficulty of three-dimensional deformation model, is not still effectively solved at present.
In addition, the three-dimensional face modeling patent that in recent years some are relevant (comprise Chinese patent application number 200610024720.7 and 200610088857.9) also provides some solutions of carrying out the three-dimensional face modeling from single photo, but these systems are all less than considering how to obtain relatively accurate unique point depth value from a photo, only utilize the two-dimensional signal of unique point, thereby influenced the accuracy of modeling.
Existing various people's face dimensional Modeling Technology mainly concentrates on the improvement to deformation algorithm, find by a large amount of experiments, under the constant situation of initial conditions (group of feature point of specific people's face on the only given photo), the raising that the improvement of deformation algorithm itself is rebuild precision to shape is limited.Carry out the 3D shape reconstruction if the two-dimension human face unique point on the single photo is directly used in the deformation algorithm, the faceform of generation observes the feature that lacks individuality from other visual angle.This is because depth information the unknown of the human face characteristic point on the single photo directly adopts the two dimensional character point can not estimate the side-play amount of other summit on depth direction effectively.If the three-dimensional coordinate of known features point uses identical deformation algorithm, the precision of reconstruction can be largely increased.But, under the situation that only has a photo as input, if it is normally impossible not obtain the depth information of unique point by other instrument.
Summary of the invention
Technical matters to be solved by this invention is that design a kind ofly can be obtained unique point depth value accurately, based on individual human face photo, fast, the human face rebuilding scheme of strong sense of reality, to satisfy in the practical application demand to people's face modeling precision and speed two aspects.
The technical scheme that the present invention solves the problems of the technologies described above is that a kind of Realistic Human face rebuilding method and system based on two-step shape modeling designed as knowledge base in a given three-dimensional face storehouse, specifically comprises, imports a front face photo; Pretreatment module is demarcated the specific human face characteristic point of input, obtains the two dimensional character point coordinate, creates neutral faceform according to the three-dimensional face storehouse, and three-dimensional model editor is demarcated the three-dimensional face unique point; People's face geometric configuration rebuilding module is set up the depth value that sparse statistical model is estimated two dimensional character point according to the priori in the three-dimensional face storehouse, estimate the depth value of two dimensional character point, and the combination of known two-dimensional coordinate and estimation of Depth value is obtained class three-dimensional (Quasi-3D, Q-3D) coordinate of unique point; People's face geometric configuration rebuilding module, is out of shape with Di Likeli Free Transform algorithm all summits to gender bender's face as the reference mark with class three-dimensional feature point, generates the geometric model of specific people's face; Synthetic and the mapping model of texture, employing is based on the synthetic texture of the method for zone broadening, it is mapped to geometric model, non-characteristic area is expanded to increase the scope that facial image covers, image after the expansion is carried out vertical projection, adopt interpolation or piece fill method to repair to the zone that is not covered on how much summits, finish the texture of three-dimensional face, generate three-dimensional face model by texture.
The present invention does not need several input photos, need not the complex processes such as coupling of unique point, and it is few to be used for the required unique point number of specific people's face modeling, therefore has real-time, calculates simply, modeling accuracy advantages of higher.Realization is the result show, under the situation of a few characteristic features point, the present invention can generate sense of reality three-dimensional face model preferably; By three-dimensional face model, can synthesize facial image realistic under different attitudes, the different illumination conditions, possess stronger using value.
Description of drawings
Fig. 1 the present invention is based on the frame diagram of the three-dimensional facial reconstruction system of two-step shape modeling
The process flow diagram of Fig. 2 two steps people's face shape modeling scheme
Fig. 3 is based on the texture building-up process of single photo
Fig. 4 blank spot p
0The synoptic diagram of adjacent 8 directions
Fig. 5 is the people's face three-dimensional reconstruction result example of finishing according to the inventive method based on single photo
Embodiment
It is instrument that the present invention proposes with people's face shape statistical model, makes up the three-dimensional facial reconstruction system, utilizes the space distribution knowledge of the last individual features point of faceform in the face database, and the depth value of the unique point of specific people's face is estimated on the comparison film.According to this thought, the present invention proposes one two step people's face shape modeling (Two-stepFacial Shape Modeling, TSFSM) scheme---at first set up the sparse statistical model in training storehouse according to human face characteristic point, estimate the depth value of test face characteristic then with optimized Algorithm, and the combination of known two-dimensional coordinate and estimation of Depth value is obtained class three-dimensional (Quasi-3D, Q-3D) coordinate (it is referred to as the class three-dimensional coordinate is unknown because of the true three-dimension coordinate figure) of unique point; Then, Q-3D unique point input deformation algorithm is rebuild whole shape, created the 3D shape of specific people's face; At last, people's face geometric model is carried out texture based on single photo.Experimental result shows, because the present invention has reasonably utilized human face structure knowledge, still can reconstruct sense of reality three-dimensional face model preferably by a few characteristic features point on individual picture.
At the drawings and specific embodiments enforcement of the present invention is specifically described below.
Be illustrated in figure 1 as the schematic block diagram of three-dimensional facial reconstruction of the present invention system.This system comprises: pretreatment module, people's face geometric configuration rebuilding module and texture module.
(1) pretreatment module
Pretreatment module is created neutral faceform, is obtained the principal character point of two-dimension human face photo and three-dimensional gender bender's face according to the three-dimensional face storehouse, and neutral faceform is made of a plurality of triangular plates.In the present invention, gender bender's face f is by the people's face f in the three-dimensional face storehouse
iSynthesize:
Wherein, m is the number of three-dimensional face model in the face database, and men and women's property respectively accounts for half usually, gender bender's face is divided into summit, a plurality of space, and comes the organization space point with the form of 3D grid.Select 10 unique points that are used for modeling in embodiments of the present invention, comprise 2 at eyebrow exterior angle, 2 at canthus, 1 on nose, 4 of mouths, 1 of chin (annotate: different embodiment can be different, all is applicable to method of the present invention).Demarcate and carry out manual fine-tuning for the two dimensional character point shape (ASM) of taking the initiative, because the unique point number is few, so be highly susceptible to realizing.Three-dimensional model editor is realized gender bender's three-dimensional feature point demarcation on the face, the user selects a point at random on screen, three-dimensional model editor is launched a ray with this point to infinite distant place, search for the triangular plate that the gender bender is intersected with this ray on the face then, select at last on this triangular plate with the nearest summit of intersection point as user-selected three-dimensional feature point.Because of gender bender's face three-dimensional model is fixed, so under the constant situation of unique point, only need carry out the demarcation of a three-dimensional feature point, the selection that preservation can be finished three-dimensional feature point.
(2) people's face geometric configuration rebuilding module
Adopt two step people's face shape modeling schemes (TSFSM).At first estimate the depth value of two dimensional character point on the input photo, all unique points are done as a whole, and it is combined into a sparse shape vector.Then, adopt an optimized Algorithm that the sparse linear model is carried out coefficient and find the solution, and the data (i.e. the z value of all unique points) of disappearance are unified to estimate.
Be illustrated in figure 2 as the process flow diagram of the present invention's two steps people's face shape modeling scheme, among the figure: the 1st step, the depth value of estimation two dimensional character point, structure class three-dimensional coordinate unique point; In the 2nd step, realize the shape reconstruction of specific people's face with the deformation algorithm according to class three-dimensional feature point.
Detailed process is as follows:
Under the situation of corresponding on the face k the unique point of three-dimensional gender bender, a sparse shape vector is constructed in three-dimensional gender bender unique point linear combination on the face on the known input human face photo and in the face database
Order
Be the set of all sparse shape vectors of three-dimensional face storehouse, wherein m is the number of three-dimensional face in the storehouse.After S carried out major component (PCA) conversion, obtain a stack features value (σ
1..., σ
M ') and proper vector (q
1..., q
M '), according to the PCA principle, can be with the eigenmatrix Q behind the convergent-divergent
s=(σ
1q
1..., σ
M 'q
M ') set up sparse linear model as follows and estimate any sparse shape vector,
s
est＝s+Q
s·β＝s+Δs
Wherein, eigenmatrix Q
sBe known, β is a combination coefficient of estimating the unique point coordinate, also is the key of estimating face characteristic value coordinate.The present invention sets up objective function according to known two dimensional character point coordinate, obtains optimum solution β by calculating extreme value
0, acquire the estimated value s of all coordinates of whole unique points
Est:
Decompose as can be known according to SVD,
Q wherein
s 2DBe Q
sTwo-dimentional version, by getting Q
sIn corresponding to (η is that (η equals at 0 o'clock to adjustment factor, β for x, y) the sparse features matrix that constitutes of the row of component
0Value equals directly to find the solution with least square method; When η much larger than 0 the time, β
0Level off to 0), get 0.1 in the present embodiment.In like manner, Δ
s 2DBe by the two-dimensional coordinate of unique point and (x, y) sparse matrix that difference constituted of component of gender bender's face three-dimensional feature point on the 2-dimentional photo of input.With β
0Substitution sparse linear model can obtain the estimated value s of all coordinates of whole unique points
EstWith known two dimensional character point coordinate (promptly import on the photo demarcated characteristic point position) and s
EstIn estimated summit depth value make up the Q-3D coordinate that generates two dimensional character point.After having obtained the Q-3D coordinate of two dimensional character point, just can carry out of the match of gender bender's face to specific people's face.At first, gender bender's unique point is on the face adjusted according to the Q-3D coordinate, adopted Di Likeli Free Transform (DFFD) algorithm that other non-unique point is adjusted then.The characteristics of DFFD are to have locality, flatness preferably, and therefore, the present invention can generate obvious, the smooth three-dimensional face geometric model of individual character.
(3) texture module
The texture module is created the texture image of three-dimensional face according to individual human face photo of being imported.The three-dimensional face geometric model is carried out texture, can remedy the deficiency of geometric model, improve the sense of reality of people's face.If directly with a front face photographic projection to three-dimensional model, many summits of edge part such as cheek can not be covered to so, the present invention is cut apart people's face according to the principal character point (as 2 at canthus, 1 on nose, 4 of mouths) of people's face earlier, then non-characteristic area is expanded to increase the scope that facial image covers, created the texture maps of three-dimensional face.Below be that example specifically describes with an instantiation.
Shown in 2 pictures before Fig. 3, on a front face photo, with people's face three principal character zones---eyes and eyebrow, nose, face etc. split according to unique point.In addition, find corresponding zone on the three-dimensional face shape, these zones, find corresponding relation between two dimensional image pixel and the three-dimensional vertices, thereby obtain the color value of three-dimensional vertices by linear interpolation method.
Because human face photo is all had powerful connections usually, for non-characteristic area, just can not find direct corresponding relation between three-dimensional vertices and the two-dimensional pixel.In order to increase the area coverage of facial image, the present invention at first expands non-characteristic area to generate texture maps, and the result is shown in the 4th subgraph of Fig. 3.Then, we judge with the normal vector on summit on the non-characteristic area of three-dimensional shape model whether this summit is blocked, thereby judge whether to carry out painted with the color value of respective pixel on the image to it:
Make summit V
i(x
i, y
i, z
i) normal vector be n
i, A
iBe the angle between this normal vector and the Z axle positive dirction, setting threshold t
1(t in the present embodiment
1Get 70 degree), if A
i≤ t
1, give vertex v with the color of pixel value
iOtherwise, with v
iColor value be set to sky.For all colours value is empty summit, and we adopt threshold value t
2The filling of blank spot is carried out in selection based on piece fill method or method of interpolation.
The step of carrying out the filling of blank spot based on the piece fill method comprises that order is that the sub-square (for example size can be 5*5) at center is pat with the blank spot
0, on whole texture maps, search and the immediate sub-piece of this sub-piece color value then, give pat with the whole assignment of the color of this sub-piece
0, threshold value then
The process of method of interpolation comprises: make p
0Be a blank spot, the interpolation rule is passed through p
0The color value on summit carries out interpolation calculation on every side, calculates the interpolation color of the blank spot that needs filling:
Clearly, following formula is exactly to calculate the interpolation color by weighted sum.Wherein, n=8 is meant and p
08 adjacent directions, as shown in Figure 4, weights λ size and p
0The distance on summit is inversely proportional to together.
In an embodiment of the present invention, at first carry out shape with TSFSM according to 10 principal character points of people's face and rebuild, wherein the deformation algorithm adopts DFFD; Carry out texture then, the three-dimensional face geometric model is carried out texture, to remedy the deficiency of geometric model based on single photo.As Fig. 5 is to adopt the reconstructed results of the inventive method to one group of real human face photo.Part 1 is the human face photo and the unique point thereof of input, and part 2 is shape point perspective view on original image of match, can see that the three-dimensional point of estimation can be preferably and original photograph match; The 3rd part is three-dimensional face front, the outboard profile of rebuilding; The 4th part has been demonstrated the process by three-dimensional face and the synthetic new attitude photo of original photo.By reconstruction, can see that the three-dimensional face model that adopts the present invention to reconstruct has very vivid effect in certain angle to these human face photos.Similarly, by setting up the statistical model of other object, the present invention can also be used for these image Reconstruction.
Claims (10)
1, a kind of Realistic Human face rebuilding method based on the two-step shape modeling scheme is characterized in that, comprises the steps:
(1) the specific human face characteristic point of input is demarcated, obtained the two dimensional character point coordinate;
(2) create neutral faceform according to the three-dimensional face storehouse, three-dimensional model editor is demarcated the three-dimensional face unique point;
(3) estimate the depth value of two dimensional character point according to the priori in the three-dimensional face storehouse, the class three-dimensional coordinate of construction feature point;
(4) with class three-dimensional feature point as the reference mark, be out of shape with Di Likeli Free Transform algorithm all summits gender bender's face, generate the geometric model of specific people's face;
(5) the synthetic mapping block that reaches of texture projects to texture maps on the geometric model then at first according to the synthetic texture maps of the human face photo of input;
(6) interpolation or piece filling are carried out in the zone that is not covered by texture on how much summits, finish texture, generate final three-dimensional face model.
2, the method for claim 1, it is characterized in that, described neutral faceform is made of a plurality of triangular plates, the demarcating steps of three-dimensional feature point further specifically comprises: the user selects a point on screen, three-dimensional model editor is launched a ray with this point to infinite distant place, search for neutral faceform then and go up the triangular plate that intersects with this ray, select on this triangular plate the three-dimensional feature point of being demarcated with the nearest summit conduct of intersection point at last.
3, the method for claim 1, it is characterized in that, described step (3) further specifically comprises, individual features point according to all faceforms in the three-dimensional face storehouse constitutes sparse shape vector, Distribution calculation according to two dimensional character point goes out combination coefficient, thereby estimates the depth value of two dimensional character point; The depth value of two dimensional character point coordinate and estimation is made up the class three-dimensional coordinate of generating feature point.
4, the method for claim 1, it is characterized in that described step (5) further specifically comprises, cut apart according to people's face principal character people's face of naming a person for a particular job, non-characteristic area is expanded the scope that covers with the increase facial image according to human face ratio, thus synthetic faceform's texture maps.
5, the method for claim 1, it is characterized in that, described step (6) further specifically comprises, not covered painted summit by texture is the central configuration fritter, if fritter is not by the painted number of vertex height of eye of texture maps half in sum, then integral-filled painted from selecting the immediate fritter of color that this piece is carried out on every side, otherwise finish the painted of this summit with the method for interpolation.
As claim 1 or 5 described methods, it is characterized in that 6, described method of interpolation is calculated the interpolation color of the blank spot that needs filling by not by the color value of texture overlay area by weighted sum.
7, a kind of Realistic Human face reconstructing system based on two-step shape modeling, it is characterized in that, pretreatment module is demarcated the specific human face characteristic point of input, obtain the two dimensional character point coordinate, create neutral faceform according to the three-dimensional face storehouse, three-dimensional model editor is demarcated the three-dimensional face unique point; People's face geometric configuration rebuilding module is estimated the depth value of two dimensional character point according to the priori in the three-dimensional face storehouse, the class three-dimensional coordinate of construction feature point, with class three-dimensional feature point as the reference mark, be out of shape with Di Likeli Free Transform algorithm all summits, generate the geometric model of specific people's face gender bender's face; Synthetic and the mapping model of texture, non-characteristic area is expanded to increase the scope that facial image covers, the image after the expansion is carried out vertical projection, the zone that is not covered by texture on how much summits is repaired, finish the texture of three-dimensional face, generate three-dimensional face model.
8, Realistic Human face reconstructing system as claimed in claim 7, it is characterized in that, described neutral faceform is made of a plurality of triangular plates, the three-dimensional feature point that the nearest summit conduct of the triangular plate that the ray that neutral faceform goes up and three-dimensional model editor is launched intersects is demarcated.
9, Realistic Human face reconstructing system as claimed in claim 7, it is characterized in that, people's face geometric configuration rebuilding module constitutes sparse shape vector according to the individual features point of all faceforms in the three-dimensional face storehouse, Distribution calculation according to two dimensional character point goes out combination coefficient, thereby estimates the depth value of two dimensional character point; The depth value of two dimensional character point coordinate and estimation is made up the class three-dimensional coordinate of generating feature point.
10, Realistic Human face reconstructing system as claimed in claim 7 is characterized in that, the synthetic mapping model that reaches of texture adopts interpolation and piece fill method that the zone that is not covered by texture is repaired.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008102330087A CN101404091B (en) | 2008-11-07 | 2008-11-07 | Three-dimensional human face reconstruction method and system based on two-step shape modeling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008102330087A CN101404091B (en) | 2008-11-07 | 2008-11-07 | Three-dimensional human face reconstruction method and system based on two-step shape modeling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101404091A true CN101404091A (en) | 2009-04-08 |
| CN101404091B CN101404091B (en) | 2011-08-31 |
Family
ID=40538095
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008102330087A Expired - Fee Related CN101404091B (en) | 2008-11-07 | 2008-11-07 | Three-dimensional human face reconstruction method and system based on two-step shape modeling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101404091B (en) |
Cited By (42)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101916454A (en) * | 2010-04-08 | 2010-12-15 | 董洪伟 | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization |
| CN101931825A (en) * | 2010-09-03 | 2010-12-29 | 南京大学 | Method for displaying 3D target in free view video stream |
| CN102054291A (en) * | 2009-11-04 | 2011-05-11 | 厦门市美亚柏科信息股份有限公司 | Method and device for reconstructing three-dimensional face based on single face image |
| CN102074040A (en) * | 2009-11-16 | 2011-05-25 | 索尼公司 | Image processing apparatus, image processing method, and program |
| CN102222363A (en) * | 2011-07-19 | 2011-10-19 | 杭州实时数码科技有限公司 | Method for fast constructing high-accuracy personalized face model on basis of facial images |
| CN102428501A (en) * | 2009-09-18 | 2012-04-25 | 株式会社东芝 | Image processing apparatus |
| CN102426712A (en) * | 2011-11-03 | 2012-04-25 | 中国科学院自动化研究所 | Three-dimensional head modeling method based on two images |
| CN102508910A (en) * | 2011-11-11 | 2012-06-20 | 大连理工大学 | Image retrieval method based on minimum projection errors of multiple hash tables |
| CN103473804A (en) * | 2013-08-29 | 2013-12-25 | 小米科技有限责任公司 | Image processing method, device and terminal equipment |
| CN103927161A (en) * | 2013-01-15 | 2014-07-16 | 国际商业机器公司 | Realtime Photo Retouching Of Live Video |
| CN104794728A (en) * | 2015-05-05 | 2015-07-22 | 成都元天益三维科技有限公司 | Method for reconstructing real-time three-dimensional face data with multiple images |
| CN104899326A (en) * | 2015-06-19 | 2015-09-09 | 四川大学 | Image retrieval method based on binary multi-index Hash technology |
| CN104966316A (en) * | 2015-05-22 | 2015-10-07 | 腾讯科技（深圳）有限公司 | 3D face reconstruction method, apparatus and server |
| CN105184860A (en) * | 2015-09-30 | 2015-12-23 | 南京邮电大学 | Method for reconstructing dense three-dimensional structure and motion field of dynamic face simultaneously |
| WO2016146001A1 (en) * | 2015-03-17 | 2016-09-22 | 阿里巴巴集团控股有限公司 | Three-dimensional modelling method and device |
| CN106021550A (en) * | 2016-05-27 | 2016-10-12 | 湖南拓视觉信息技术有限公司 | Hair style designing method and system |
| CN106127147A (en) * | 2016-06-23 | 2016-11-16 | 深圳市唯特视科技有限公司 | A kind of face depth texture restorative procedure based on three-dimensional data |
| CN104077804B (en) * | 2014-06-09 | 2017-03-01 | 广州嘉崎智能科技有限公司 | A kind of method based on multi-frame video picture construction three-dimensional face model |
| CN106652037A (en) * | 2015-10-30 | 2017-05-10 | 深圳超多维光电子有限公司 | Face mapping processing method and apparatus |
| CN106920274A (en) * | 2017-01-20 | 2017-07-04 | 南京开为网络科技有限公司 | Mobile terminal 2D key points rapid translating is the human face model building of 3D fusion deformations |
| CN107590452A (en) * | 2017-09-04 | 2018-01-16 | 武汉神目信息技术有限公司 | A kind of personal identification method and device based on gait and face fusion |
| WO2018040099A1 (en) * | 2016-08-31 | 2018-03-08 | 深圳市唯特视科技有限公司 | Three-dimensional face reconstruction method based on grayscale and depth information |
| CN108492373A (en) * | 2018-03-13 | 2018-09-04 | 齐鲁工业大学 | A kind of face embossment Geometric Modeling Method |
| CN108564619A (en) * | 2018-04-25 | 2018-09-21 | 厦门大学 | A kind of sense of reality three-dimensional facial reconstruction method based on two photos |
| CN108876837A (en) * | 2018-04-19 | 2018-11-23 | 宁波大学 | One kind being based on L1/2The 3 D human body attitude reconstruction method of regularization |
| CN109064542A (en) * | 2018-06-06 | 2018-12-21 | 链家网（北京）科技有限公司 | Threedimensional model surface hole complementing method and device |
| CN109447043A (en) * | 2018-12-23 | 2019-03-08 | 广东腾晟信息科技有限公司 | A kind of face method for automatic modeling |
| CN109448083A (en) * | 2018-09-29 | 2019-03-08 | 浙江大学 | A method of human face animation is generated from single image |
| CN109598188A (en) * | 2018-10-16 | 2019-04-09 | 深圳壹账通智能科技有限公司 | Information-pushing method, device, computer equipment and storage medium |
| CN109829486A (en) * | 2019-01-11 | 2019-05-31 | 新华三技术有限公司 | Image processing method and device |
| CN109978984A (en) * | 2017-12-27 | 2019-07-05 | Tcl集团股份有限公司 | Face three-dimensional rebuilding method and terminal device |
| CN110276825A (en) * | 2018-03-15 | 2019-09-24 | 南京大学 | A kind of three-dimensional facial reconstruction method based on template deformation |
| CN110428491A (en) * | 2019-06-24 | 2019-11-08 | 北京大学 | Three-dimensional facial reconstruction method, device, equipment and medium based on single-frame images |
| CN110544311A (en) * | 2018-05-29 | 2019-12-06 | 百度在线网络技术（北京）有限公司 | Safety warning method, device and storage medium |
| WO2020042975A1 (en) * | 2018-08-27 | 2020-03-05 | 阿里巴巴集团控股有限公司 | Face pose estimation/three-dimensional face reconstruction method and apparatus, and electronic device |
| WO2020062120A1 (en) * | 2018-09-29 | 2020-04-02 | 浙江大学 | Method for generating facial animation from single image |
| CN111160232A (en) * | 2019-12-25 | 2020-05-15 | 上海骏聿数码科技有限公司 | Front face reconstruction method, device and system |
| CN111179332A (en) * | 2018-11-09 | 2020-05-19 | 北京市商汤科技开发有限公司 | Image processing method and device, electronic device and storage medium |
| CN111742347A (en) * | 2018-02-26 | 2020-10-02 | 高通股份有限公司 | Dynamic illumination for objects in an image |
| CN112086193A (en) * | 2020-09-14 | 2020-12-15 | 巢自强 | Face recognition health prediction system and method based on Internet of things |
| WO2021063271A1 (en) * | 2019-09-30 | 2021-04-08 | Oppo广东移动通信有限公司 | Human body model reconstruction method and reconstruction system, and storage medium |
| CN117422831A (en) * | 2023-10-31 | 2024-01-19 | 北京百度网讯科技有限公司 | Three-dimensional eyebrow shape generating method and device, electronic equipment and storage medium |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108109116A (en) * | 2017-12-14 | 2018-06-01 | 大连理工大学 | Image repair algorithm based on BSCB models |
-
2008
- 2008-11-07 CN CN2008102330087A patent/CN101404091B/en not_active Expired - Fee Related
Cited By (65)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102428501A (en) * | 2009-09-18 | 2012-04-25 | 株式会社东芝 | Image processing apparatus |
| CN102054291A (en) * | 2009-11-04 | 2011-05-11 | 厦门市美亚柏科信息股份有限公司 | Method and device for reconstructing three-dimensional face based on single face image |
| CN102074040A (en) * | 2009-11-16 | 2011-05-25 | 索尼公司 | Image processing apparatus, image processing method, and program |
| CN102074040B (en) * | 2009-11-16 | 2013-06-19 | 索尼公司 | Image processing apparatus and image processing method |
| CN101916454B (en) * | 2010-04-08 | 2013-03-27 | 董洪伟 | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization |
| CN101916454A (en) * | 2010-04-08 | 2010-12-15 | 董洪伟 | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization |
| CN101931825A (en) * | 2010-09-03 | 2010-12-29 | 南京大学 | Method for displaying 3D target in free view video stream |
| CN102222363A (en) * | 2011-07-19 | 2011-10-19 | 杭州实时数码科技有限公司 | Method for fast constructing high-accuracy personalized face model on basis of facial images |
| CN102426712A (en) * | 2011-11-03 | 2012-04-25 | 中国科学院自动化研究所 | Three-dimensional head modeling method based on two images |
| CN102426712B (en) * | 2011-11-03 | 2014-07-16 | 中国科学院自动化研究所 | Three-dimensional head modeling method based on two images |
| CN102508910A (en) * | 2011-11-11 | 2012-06-20 | 大连理工大学 | Image retrieval method based on minimum projection errors of multiple hash tables |
| CN103927161A (en) * | 2013-01-15 | 2014-07-16 | 国际商业机器公司 | Realtime Photo Retouching Of Live Video |
| CN103473804A (en) * | 2013-08-29 | 2013-12-25 | 小米科技有限责任公司 | Image processing method, device and terminal equipment |
| CN104077804B (en) * | 2014-06-09 | 2017-03-01 | 广州嘉崎智能科技有限公司 | A kind of method based on multi-frame video picture construction three-dimensional face model |
| CN106033621B (en) * | 2015-03-17 | 2018-08-24 | 阿里巴巴集团控股有限公司 | A kind of method and device of three-dimensional modeling |
| US10789767B2 (en) | 2015-03-17 | 2020-09-29 | Alibaba Group Holding Limited | Reducing computational complexity in three-dimensional modeling based on two-dimensional images |
| WO2016146001A1 (en) * | 2015-03-17 | 2016-09-22 | 阿里巴巴集团控股有限公司 | Three-dimensional modelling method and device |
| CN106033621A (en) * | 2015-03-17 | 2016-10-19 | 阿里巴巴集团控股有限公司 | Three-dimensional modeling method and device |
| US10410405B2 (en) | 2015-03-17 | 2019-09-10 | Alibaba Group Holding Limited | Reducing computational complexity in three-dimensional modeling based on two-dimensional images |
| CN104794728A (en) * | 2015-05-05 | 2015-07-22 | 成都元天益三维科技有限公司 | Method for reconstructing real-time three-dimensional face data with multiple images |
| CN104794728B (en) * | 2015-05-05 | 2017-09-19 | 成都元天益三维科技有限公司 | The method that face real-time three-dimensional data is reconstructed by multiple image |
| WO2016188318A1 (en) * | 2015-05-22 | 2016-12-01 | 腾讯科技（深圳）有限公司 | 3d human face reconstruction method, apparatus and server |
| CN104966316A (en) * | 2015-05-22 | 2015-10-07 | 腾讯科技（深圳）有限公司 | 3D face reconstruction method, apparatus and server |
| CN104966316B (en) * | 2015-05-22 | 2019-03-15 | 腾讯科技（深圳）有限公司 | A kind of 3D facial reconstruction method, device and server |
| US10055879B2 (en) | 2015-05-22 | 2018-08-21 | Tencent Technology (Shenzhen) Company Limited | 3D human face reconstruction method, apparatus and server |
| CN104899326A (en) * | 2015-06-19 | 2015-09-09 | 四川大学 | Image retrieval method based on binary multi-index Hash technology |
| CN105184860A (en) * | 2015-09-30 | 2015-12-23 | 南京邮电大学 | Method for reconstructing dense three-dimensional structure and motion field of dynamic face simultaneously |
| CN106652037A (en) * | 2015-10-30 | 2017-05-10 | 深圳超多维光电子有限公司 | Face mapping processing method and apparatus |
| CN106652037B (en) * | 2015-10-30 | 2020-04-03 | 深圳超多维光电子有限公司 | Face mapping processing method and device |
| CN106021550A (en) * | 2016-05-27 | 2016-10-12 | 湖南拓视觉信息技术有限公司 | Hair style designing method and system |
| CN106127147A (en) * | 2016-06-23 | 2016-11-16 | 深圳市唯特视科技有限公司 | A kind of face depth texture restorative procedure based on three-dimensional data |
| CN106127147B (en) * | 2016-06-23 | 2019-07-26 | 深圳市唯特视科技有限公司 | A kind of face depth texture restorative procedure based on three-dimensional data |
| WO2018040099A1 (en) * | 2016-08-31 | 2018-03-08 | 深圳市唯特视科技有限公司 | Three-dimensional face reconstruction method based on grayscale and depth information |
| CN106920274A (en) * | 2017-01-20 | 2017-07-04 | 南京开为网络科技有限公司 | Mobile terminal 2D key points rapid translating is the human face model building of 3D fusion deformations |
| CN107590452A (en) * | 2017-09-04 | 2018-01-16 | 武汉神目信息技术有限公司 | A kind of personal identification method and device based on gait and face fusion |
| CN109978984A (en) * | 2017-12-27 | 2019-07-05 | Tcl集团股份有限公司 | Face three-dimensional rebuilding method and terminal device |
| CN111742347B (en) * | 2018-02-26 | 2024-03-15 | 高通股份有限公司 | Dynamic illumination for objects in images |
| CN111742347A (en) * | 2018-02-26 | 2020-10-02 | 高通股份有限公司 | Dynamic illumination for objects in an image |
| CN108492373A (en) * | 2018-03-13 | 2018-09-04 | 齐鲁工业大学 | A kind of face embossment Geometric Modeling Method |
| CN108492373B (en) * | 2018-03-13 | 2019-03-08 | 齐鲁工业大学 | A kind of face embossment Geometric Modeling Method |
| CN110276825A (en) * | 2018-03-15 | 2019-09-24 | 南京大学 | A kind of three-dimensional facial reconstruction method based on template deformation |
| CN108876837A (en) * | 2018-04-19 | 2018-11-23 | 宁波大学 | One kind being based on L1/2The 3 D human body attitude reconstruction method of regularization |
| CN108876837B (en) * | 2018-04-19 | 2021-09-14 | 宁波大学 | Based on L1/2Regularized three-dimensional human body posture reconstruction method |
| CN108564619A (en) * | 2018-04-25 | 2018-09-21 | 厦门大学 | A kind of sense of reality three-dimensional facial reconstruction method based on two photos |
| CN108564619B (en) * | 2018-04-25 | 2021-05-14 | 厦门大学 | Realistic three-dimensional face reconstruction method based on two photos |
| CN110544311A (en) * | 2018-05-29 | 2019-12-06 | 百度在线网络技术（北京）有限公司 | Safety warning method, device and storage medium |
| CN109064542A (en) * | 2018-06-06 | 2018-12-21 | 链家网（北京）科技有限公司 | Threedimensional model surface hole complementing method and device |
| WO2020042975A1 (en) * | 2018-08-27 | 2020-03-05 | 阿里巴巴集团控股有限公司 | Face pose estimation/three-dimensional face reconstruction method and apparatus, and electronic device |
| US11941753B2 (en) | 2018-08-27 | 2024-03-26 | Alibaba Group Holding Limited | Face pose estimation/three-dimensional face reconstruction method, apparatus, and electronic device |
| WO2020062120A1 (en) * | 2018-09-29 | 2020-04-02 | 浙江大学 | Method for generating facial animation from single image |
| CN109448083A (en) * | 2018-09-29 | 2019-03-08 | 浙江大学 | A method of human face animation is generated from single image |
| US11544887B2 (en) | 2018-09-29 | 2023-01-03 | Zhejiang University | Method for generating facial animation from single image |
| CN109598188A (en) * | 2018-10-16 | 2019-04-09 | 深圳壹账通智能科技有限公司 | Information-pushing method, device, computer equipment and storage medium |
| CN111179332B (en) * | 2018-11-09 | 2023-12-19 | 北京市商汤科技开发有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN111179332A (en) * | 2018-11-09 | 2020-05-19 | 北京市商汤科技开发有限公司 | Image processing method and device, electronic device and storage medium |
| CN109447043A (en) * | 2018-12-23 | 2019-03-08 | 广东腾晟信息科技有限公司 | A kind of face method for automatic modeling |
| CN109829486A (en) * | 2019-01-11 | 2019-05-31 | 新华三技术有限公司 | Image processing method and device |
| CN110428491A (en) * | 2019-06-24 | 2019-11-08 | 北京大学 | Three-dimensional facial reconstruction method, device, equipment and medium based on single-frame images |
| US11928778B2 (en) | 2019-09-30 | 2024-03-12 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method for human body model reconstruction and reconstruction system |
| WO2021063271A1 (en) * | 2019-09-30 | 2021-04-08 | Oppo广东移动通信有限公司 | Human body model reconstruction method and reconstruction system, and storage medium |
| CN111160232A (en) * | 2019-12-25 | 2020-05-15 | 上海骏聿数码科技有限公司 | Front face reconstruction method, device and system |
| CN112086193A (en) * | 2020-09-14 | 2020-12-15 | 巢自强 | Face recognition health prediction system and method based on Internet of things |
| CN112086193B (en) * | 2020-09-14 | 2024-08-20 | 信阳昆扬信息技术服务中心(个人独资) | Face recognition health prediction system and method based on Internet of things |
| CN117422831A (en) * | 2023-10-31 | 2024-01-19 | 北京百度网讯科技有限公司 | Three-dimensional eyebrow shape generating method and device, electronic equipment and storage medium |
| CN117422831B (en) * | 2023-10-31 | 2024-11-08 | 北京百度网讯科技有限公司 | Three-dimensional eyebrow shape generating method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101404091B (en) | 2011-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101404091B (en) | Three-dimensional human face reconstruction method and system based on two-step shape modeling | |
| CN108921926B (en) | End-to-end three-dimensional face reconstruction method based on single image | |
| CN102999942B (en) | Three-dimensional face reconstruction method | |
| CN101916454B (en) | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization | |
| AU2020200811A1 (en) | Direct meshing from multiview input using deep learning | |
| CN108876814B (en) | Method for generating attitude flow image | |
| Agresti et al. | Deep learning for confidence information in stereo and tof data fusion | |
| CN103400409A (en) | 3D (three-dimensional) visualization method for coverage range based on quick estimation of attitude of camera | |
| CN110378349A (en) | The mobile terminal Android indoor scene three-dimensional reconstruction and semantic segmentation method | |
| CN112785692B (en) | Single-view-angle multi-person human body reconstruction method based on depth UV prior | |
| CN112530005B (en) | Three-dimensional model linear structure recognition and automatic restoration method | |
| CN107358645A (en) | Product method for reconstructing three-dimensional model and its system | |
| CN106447763A (en) | Face image three-dimensional reconstruction method for fusion of sparse deformation model and principal component regression algorithm | |
| WO2018080533A1 (en) | Real-time generation of synthetic data from structured light sensors for 3d object pose estimation | |
| CN106127818A (en) | A kind of material appearance based on single image obtains system and method | |
| Choi et al. | Balanced spherical grid for egocentric view synthesis | |
| Ye et al. | 3d morphable face model for face animation | |
| Brunton et al. | Wavelet model-based stereo for fast, robust face reconstruction | |
| Takemura et al. | Diminishing head-mounted display for shared mixed reality | |
| Gong | Application and Practice of Artificial Intelligence Technology in Interior Design | |
| Wu et al. | Photogrammetric reconstruction of free-form objects with curvilinear structures | |
| CN111768476A (en) | Expression animation redirection method and system based on grid deformation | |
| Aleksandrova et al. | 3D face model reconstructing from its 2D images using neural networks | |
| Li et al. | Example-based 3D face reconstruction from uncalibrated frontal and profile images | |
| Chen et al. | Generation of 3d caricature by fusing caricature images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110831 Termination date: 20141107 |
|
| EXPY | Termination of patent right or utility model |